- 1 Introduction
- 2 Examples
- 3 Pipe Selection Variables
- 4 Current Standards
- 5 CIBSE Technical Bulletin
- 6 Limits imposed by velocity versus common pipe sizes
- 7 Pipe velocity
- 8 Pressure Losses in Pipework
- 9 Water Quality
- 10 Categories of Pipework
- 11 The Peripheral Pipework Zone
- 12 Boost Flow and DHW Delivery Times
- 13 Diversity
- 14 Safety Margins
- 15 Differentiating Return Pipe Design
- 16 Pumping
- 17 Differential Pressure Limitation
- 18 Commissioning Valves
- 19 Gravity Circulation
- 20 Standardising First Fix Brackets
- 21 Calculating Temperature Drops
- 22 Laminar and Turbulent Flow
- 23 Summary
- 24 Article Compiled By
This article is aimed at clarifying the design principles that need to be understood in order to achieve both reliable and efficient heat networks (i.e. pipework), be they big or small.
The current standards, relating specifically to pipe sizing within buildings, are limiting the efficiency heat networks can achieve, down to losses in translation from old standards and confusion between pipework covering long distances, and internal pipes feeding properties.
Here we examine the current standards, and the known science in order to lay bare the problems with current guidance and reinstate the correct science.
District heating has got a bad name because of both the recorded and experienced efficiencies from real life installations. New SAP proposals assume only 50% efficiency, and data from metering companies often shows far lower efficiencies annually - 25% is indeed possible. It is mainly because heat networks at the building level are oversized, overheat buildings, and destroy plantroom and heat loss efficiencies through elevated return temperatures.
Examining if and why the regulations prevent consultants from signing off on their Professional Indemnity is key to improving efficiency in the future, when it comes to the sizing of smaller pipework within buildings connected to a heat network.
See also our working examples.
Pipe Selection Variables
- Inside and Outside Diameters
- Material (smoothness)
- entering water temp
- flow rate
- Maximum velocity (erosion)
- Continuous working pressure vs temperature
- Peak working pressure vs temperature
- Insulation thickness
- Ambient temp
- Surrounding material (thermal absorption)
- Volume (for startup delay calcs)
- Power capacity
- Convection factor
- Heat loss
- Friction heat gain
- Exit temperature
- Time to cool at no flow
Other issues affecting pipe calculations/selection:
- Minimum static head
- reduced bore of fittings
- ease of installation
- pump selection
CIBSE and BSRIA
Extracts from CIBSE A guide on pipe sizing and the (lack of) logic behind 200pa/m
1.A1.3 Pipe and pump sizing 1.A1.3.1 Pipe sizing The following considerations should be taken into account when selecting the appropriate pipe size for a given design flow rate: Pipework noise - Pipes must be sized such that the velocity of the water running through them will not be high enough to cause either vibration induced noise or erosion of the pipe material. Erosion of relatively soft metals such as copper can occur at elbows if the water velocity is excessive. Table 1.A1.4 indicates recommended maximum water velocities. Table 1.A1.4: Recommended range of maximum water velocities. Pipe diameter (mm) Copper Steel 15-50mm 1.0 m/s 1.5 m/s Over 50mm 1.5 m/s 3 m/s Air and dirt settlement: Small air bubbles or particles of debris carried by the flowing fluid may settle out in the pipe at low velocities. Ideally, full load design velocities should be maintained at a value greater than 0.5 m/s. Where full load design velocities may fall below this value additional dirt or air removal devices should be considered. BSRIA Guide BG29/2011 (Brown, Parsloe, 2004) provides recommendations on the maintenance of system cleanliness.
The minimum velocities from CIBSE were most recently quoted as 0.75m/s for pipes up to 50mm and 1.25m/s for bigger pipes. This is where guidance crosses into the domain of another Institute, with pipes under 50mm covered by CIPHE guidance.
Pump energy: Pipes must be sized such that the energy consumed by the pump is not excessive. Smaller pipes will have a greater resistance to flow and will therefore incur a greater pump energy consumption compared to larger pipes. Pump energy consumption will be roughly proportional to the average pressure loss per metre (expressed as Pascals per metre, Pa/m) of the straight pipe lengths in the system. As a general rule, to minimize the life cycle energy consumption of a pipework system (i.e. the embodied energy of its pipes plus the pump life cycle energy consumption), pipes should be sized based on a criterion of not exceeding 200 Pa/m.
This last paragraph is very unhelpful when it comes to designing for thermal efficiency as well as pumping efficiency, especially when calculations show its possible to lose more than a hundred times as much energy through thermal loss in a building than pump energy. The inference that smaller pipes consume more energy that large ones takes no account of relative lengths.
The 200Pa/m rule is based also on recent analysis (see BSRIA report below) that showed over the life of a system, pump energy far exceeds embedded energy from the manufacturing process hence in energy terms it’s best to keep Pa/m as low as possible. But this analysis does not take into account heat losses from pipework. As such it cannot be justified as a general rule for sizing specific pipes to achieve efficient systems, and only leads to oversized pipework if not viewed in the correct context and considered alongside thermal efficiency of a network - where the resulting heat losses will be considerably higher than pump energy.
The typical pressure drop over an instantaneous HIU at its limits is 50kPa, or 50,000 Pa. With pipes sized to pump energy figure of 200Pa/m, that equates to 250 metres of pipe - per property - before pressure / energy loss in pipework starts to overtake losses in an HIU.
The following chart shows the pressure drops and velocities for standard copper tube, with the minimum 0.5m/s limit as well as the 200Pa/m limit. The green areas are pipe sizes that comply to both limits at peak flow.
The real problem lies in that there are ranges of outputs where none of the commonly available pipe sizes comply. How, for example can one satisfy a peak load flow of 13 litres/minute. 22mm tube would exceed the 200Pa/m limit, while 28mm pipe does get close to the 0.5m/s minimum flow. Snookered, unless one takes the pump energy limits in a wider context, and follows the upper limits of 1 to 1.5m/s (in yellow).
Then take into consideration the guidance of 0.75m/s and it becomes impossible to sign off pipe design below 42mm, going purely by this guidance, although one could sneak through 35mm at exactly 42 litres/minute peak load (140kW). This is the defining point on a 'graph' when the Institutes collide.
Further, a fixed limit does not take into account temperature drops. If the figure was derived when standard flow and return temperatures were 82/71C, which they were, then compared to water in a heat network today running at 80/40C, pumping was only abut 25% as efficient as now in delivering load. Adjusting for temperature drops would imply a maximum average pressure drop of 800Pa/m for a modern heat network.
These widely quoted velocity limits were in CIBSE Guide B (Heating) back in the 1980s and were used to write the BSRIA commissioning guide in 1989 – referencing CIBSE. CIBSE dropped them when guide B was updated in around 1990, and thereafter CIBSE have referenced the BSRIA commissioning guide for velocity limits (without realising that the values came from them originally).
In 2010 when this was pointed out, rather than re-instate the velocity limits in Code W and relevant CIBSE guides, it was argued that such choices should be left to the designer, leaving a circular reference between two standards and no actual guidance.
Originally, the table included minimum velocity limits to avoid air and dirt settlement. The minimum values were (and still are) widely ignored since designers, contractors and manufacturers are not keen to use smaller pipes. (certainly not copper). Hence, in most systems today, the velocities in terminal pipes are well below the old CIBSE minimums, resulting in pipes that are prone to air and dirt settlement.
During the 2010 re-write it was decided to not mention minimum velocity limits in the CIBSE guide.
Regarding the 200 Pa/m limit, the BSRIA reference to look at is Energy Efficient Pumping Systems BG12/2011.
With the current guidance on pipe velocities as it is, this has left pipe sizing limits (as far as indemnity is concerned) ruled purely by pump energy concerns and a perceived 200 Pa/m limit - rather than sizing pipes to allowable velocities and looking at pumping energy as a whole.
The suggestion that system with a velocity lower than 0.5m/s should fit dirt and air separation fails to address one simple fact - it takes velocity in pipes to move dirt and air to the dirt separator. Fittings 35mm pipework to a single dwelling and putting a dirt separator in the plantroom, simply does not work, ye this is what guidance advises.
With a loss of proper guidance and consultants tied to over-sizing, heat networks continue to perform poorly through excess heat loss, with dirt and air collection from low velocities, and a higher capital cost than needed.
The CIBSE Guide B1 relates specifically to heating system design, rather than water supply.
CIBSE Technical Bulletin
CIBSE Heat Networks Code of Practice
This is a recent document aimed specifically at heat networks, and should contain the most up to date guidance.
The consultation document for the standard provides a table of typical velocities on page 36:
This table does not cover any pipework used on a network to feed properties. The statement that higher pressure drops can be acceptable in non-critical side branches is unclear in its definitions. Does this, for example, cover laterals or feeds into properties? Velocity is mentioned nowhere else in guidance.
These figures are typical velocities, so there are no limits, either maximum or minimum provided.
No mention is made of the use of hammer arresters to cut down transient pressure waves implying the only mechanism to keep these within limits is to increase pipe size to reduce velocity. Testing pressure on site once installed is not an option - its too late to change pipe size by this point leading to pipes been potentially oversized in design.
Pipe sizing in larger and more complicated buildings is perhaps best done by using a simplified tabular procedure. BS 6700 gives examples of this but for more detailed data readers should refer to the Institute of Plumbing’s Plumbing Engineering Services Design Guide.
The last publication of Plumbing Engineering Services Design Guide from the IOP (Institute of Plumbing - now the CIPHE, or Chartered Institute of Plumbing and Heating Engineers) was in 2002 (I just so happened to author sections in that guide on thermal storage and plate heat exchange, so know it well).
Page 56 covers heating systems and pipe design.
Determine the pressure drop in PA/m for the flow rate in the pipe size concerned ... If the resultant velocity >1m/s or pressure drop >400Pa/m, then increase pipe size.
This is giving a limit of 400Pa/m, however gives no explanation into the thinking behind it, other than to comment that it is worth going through the procedure to ensure your pump can achieve the head - so the thinking may well be coming from an pumping head perspective rather than limits on pipe.
Elsewhere in the guide under pipe sizing hot and cold systems (page 16) it states;
Velocities can be allowed to increase to 1.0 to 1.5 metres/second, and possibly higher where pipes are not routed in non occupied areas. Noise is a major consideration and velocities above 1.5m/s in pipework passing through occupied areas, in particular bedrooms, should be avoided. Erosion and corrosion are less of an issue (than noise). Where velocities exceed 2.5m/s erosion and/or corrosion can result from abrasive action of particles in the water.
The following charts are also provided regarding velocities. The top line represents the an area of the IOP guidance in CIBSE territory now days when it comes to district heating.
This implies a limit on pipework of 2.5m/s before erosion starts to become a issue.
The following chart relates to hot water services. The higher velocities are a reflection that pipes see peak velocities from water use for only a fraction of the time. Note, limit of 8-10 m/s for pipes feeding taps. In other words, high velocities are fine for short periods.
However, the third edition of CIBSE Guide G has been produced in collaboration with the Chartered Institute of Plumbing and Heating Engineering (CIPHE) and may include updated guidance.
The 1997 ASHRAE (American Society of Heating, Refrigerating and Air-Conditioning Engineers) Fundamentals Handbook page 33.3 table gives following information for max water velocity to minimise erosion:
|Normal operation hours per year||Velocity m/s|
Same page also says "velocities on the order of 3 to 5m/s lie within the range of allowable noise levels for residential and commercial buildings"
This American standard introduces a principle missing from UK guidance (except CIPHE above) - the fact that higher velocities are fine for shorter periods. In the context of district heating we need to look at the hours per year for various functions. In the context of pipes feeding properties, DHW runs for less than 365 hours a day for example but requires higher velocities, so should be treated differently to central heating that will be on for 1500 or more hours and use very little flow.
Copper Tube Standards
BS 6700 gives the following maximum water velocities
|Water Temperature °C||Maximum Water Velocity (m/s)|
The Copper Development Association quote the max velocity as 'about 3.0m/s for cold water and 2.5m/s for hot'.
From the CDA:
In general the critical velocity at which erosion-corrosion occurs is higher in hard water than soft water. For example, with some hard water services the maximum velocity at which water can be distributed, without any adverse effect, is about 3 metres per second (m/s) but with certain soft waters the maximum safe water velocity is likely to be nearer 2 m/s. However, it should be noted that for any given water, the critical velocity would be higher on cold water services than hot water systems.
The following is from a major copper tube manufacturer: Recommended flow velocities regarding copper tubes and water have been in place for many years and were presumably derived from research carried out into the subject a long time ago. The fundamentals of the materials involved, i.e. the composition of the copper and the density of the water have not changed over time, so there is no reason to move away from current recommended velocities. In our view to increase water velocities would only increase the risk of erosion corrosion issues and is not something we would consider safe to do. Research was probably carried out by the British Non-Ferrous Metals Technology Centre, this organisation is no longer in existence and we do not have copies of any of the research. Regards Copper Development Association 5 Grovelands Business Centre, Boundary Way Hemel Hempstead, HP2 7TE, UK
The following documents have been provided by the CDA as oldest known records.
Plastic Tube Manufacturers
Polybutylene: 10m/s if fittings used do not reduce bore
This limit of 10m/s is the highest quoted from a manufacturer, however is also understood that cavitation can start occurring at this velocity, suggesting a maximum realistic limit of 7.5m/s.
In essence, pump limits will rule completely.
Limits imposed by velocity versus common pipe sizes
On pipe sizes feeding individual properties, it does seem that the choices are somewhat limited in reality.
If you assume copper pipe, there is little room to move based on flow. The real driver is the assumed temperature drop at peak load. 50C on DHW (75-25) is probably most common.
22mm is the only pipe size that satisfies 0.75 - 1.5m/s limits (min/max) between a typical peak DHW load of 50 to 100kW, which basically covers all properties.
If you want more water to taps, then an HIU+cylinder running off 22mm primaries can drive 5 bathrooms if you fancy, so never a need to go to 28mm.
If you want to reduce primary temperatures, and flow goes up accordingly, 22mm will still deliver around 40kW, without exceeding any velocity limits.
The temptation to increase to 28mm on long runs is simply dropping below minimum velocity limits and will result in air, dirt, longer startup delays, & thermal inefficiency.
There is a point, for 1 bed flats, where both 22 and 15 are within limits - so 15mm flow and 22mm return would be ideal for 1 bed flats where pipe runs are long enough to cause undue delays when an HIU starts up from cold - pump head permitting.
Note the point around 14 litres per minute peak flow rate. No pipe size is strictly within crude limits, but taking into account the number of hours per year spent at peak design load, exceeding velocity limit on 1.5 at this moment is acceptable - in fact desirable for clearing debris.
Below is the same graph where velocities are allowed up to 3m/s. This would be on pipes where peak load is based on instantaneous DHW. Once too many outlets are been fed, the diversity of DHW drops off, so these higher velocities would only be used on short laterals feeding a few properties.
The limits designers should work to regarding pipe velocity concern the following factors:
- Upper limits imposed by pressure drops and pipe lengths (variable)
- Upper limits imposed by erosion
- Upper limits imposed by pumping energy
- Lower limits imposed by the need to clear air and settled dirt (horizontal pipes)
The application of limits is complicated by the fact system flow rates vary, with low flows experienced most of the time with generally predictable patterns (weather, time of day), but unpredictable peaks and spikes (diversity). It is further complicated by a failure in guidance to distinguish between the function and orientation of various sections of pipework in a heat network, treating all pipes equally.
Failures of pipe due to excess velocity over time are caused by erosion. The corrosion rate of copper in most drinkable waters with no flow is less than 2.5 µm/year, at this rate a 15 mm tube with a wall thickness of 0.7 mm would last for about 280 years. In some soft waters the general corrosion rate may increase to 12.5 µm/year, but even at this rate it would take over 50 years to perforate the same tube. (extract from Wikipedia)
Erosion is the process of abrasive wear on the walls of the tube by suspended particles within the water flow. The partner to erosion is the corrosion of exposed metal, building up a protective oxide layer.
At the point of failure, velocity is such that any permanent corrosion deposits (oxides) are eroded away, exposing bare metal. Local areas of turbulence in the water flow adjacent to the tube wall, accelerate suspended particles to higher velocities, resulting in greater impact forces between those particles and oxides lining the tube wall. Repeated erosion at the same point causes failure.
Failure points centre on a point of peak turbulence, so a key factor in the peak velocity is the presence of obstructions within pipework that will cause turbulence, and hence erosion. Such causes may be:
- Snots of solder
- Reduced bore where pipe is cut with a ring cutter and not internally deburred.
- Tight 90 degree elbows
This photo (upside-down as one can tell from the point where solder has collected at top of photo) is an example of erosion after an tight elbow. One can almost picture the start point of turbulence, at the edge of the circular cut copper pipe. The sharp edge to the pipe causes a point of turbulence, and the shiny metal distinctly starts along this line marking the point at which the blue copper oxide is cleaned away by the turbulence. Once eddy currents form they will focus erosion, the pits start to form, and these in turn causing further turbulence immediately downstream - the flow of turbulence can be seen in the photo. An analogy is that of a tornado, where negative pressures suck objects up, and debris within the eddy is effective at generating further debris through interaction with objects on the ground.
Possibly the single most important advice to installers using copper pipe is to debur the inside of the pipe after cutting. Good quality tube cutters are fitted with a tool specifically for that.
Averaging Peak Velocity Theory
Erosion rates are related to velocity over time. They are not a 'single incident'.
A pipe that experiences 5m/s all its life, will no doubt fail from erosion before a pipe that normally sees 1m/s with a once a week spike to 6m/s for a few minutes.
The rate of erosion is physically linked to kinetic energy in particles and water. Kinetic energy increases with the square of the velocity, so it would be logical to work on the basis erosion rates are linked to the square of the pipe velocity.
So if we are to average pipe velocities to link them to actual erosion it is required to analyse the velocity over time.
Starting with an assumption that 1m/s is the lowest recommended peak pipe velocity - for 90C pipework that is inaccessible for servicing. If we scale velocities to loads, to achieve an average velocity of 1m/s, this should give us a peak velocity.
1m/2 = ok for continuous average. 2m/s = 4 times the kinetic energy as 1m/s. 3m/s = 9 times as much energy 8m/s = 64
Looking at a single 2 bed, three person property with a 50kW HIU. 17 litres/minute peak primary flow rate possible (24 litres/minute to tap temperature). That's a luxury bath tap, or two showers and a basin.
This peak velocity is 1m/s for 22mm pipe - so no problems anyhow.
For 15mm, 17 lpm equates to a velocity of nearer 1.7m/s.
Heating load peak 5kW = 2.5 litres per minute primary, or roughly 0.4m/s. This represents the lower velocity steady state where a limit of 1m/s is in place.
Assuming the worst, the entire days hot water use will be drawn back to back. 3 people using a generous 100 litres per day, gives 13 minutes of full 59kW output DHW.
1440 minutes in a day...
(1 x 1) * 1440 = 1440 (acceptable energy-seconds) (0.4 x 0.4) x 1440 = 231 (heating) (1.7 x 1.7) x 13 = 38 (hot water) total = 269 Percentage of limit utilised = 19%
This calculation shows that for a single dwelling, 15mm pipe would not erode due to velocity.
Going further, taking a suggested maximum of 8m/s... the peak velocity figures quoted in CIPHE for peak velocities on pipes feeding taps.
1440 (acceptable energy-seconds) (0.4 x 0.4) x 1440 = 231 (heating) (8 x 8) x 13 = 832 (hot water) total = 1063 Percentage of limit utilised = 74%
Still within time scaled erosion limits.
One can further average heating load over a year to provide even more safety margin.
A full analysis can be found in the following Excel spreadsheet.
You can't go wrong with 22mm tube
This table, based on the best calculations we can do, details the maximum and minimum pipe sizes for feeding various properties.
The diversified loads are put though the erosion calculation and along with minimum 0.75m/s limit we get a range of possible sizes.
The striking result is that 22mm must be used on most laterals, feeding up to 8 properties. Any smaller and erosion adds up. Any larger and peak loads never reach a high enough velocity to keep pipes clean.
So a pretty good rule to pipe sizing is always use 22mm feeding up to 8 properties downstream, with a second rule being to run calculations for your specific temperatures and loads.
Learning from the Oil Industry
The equations used by the oil industry to work out velocities due to erosion are also based on a squared law.
Ve = C*(1/ρm)0.5 where: Ve = fluid erosional velocity, ft/s (m/s) ρm = gas / liquid mixture density at flowing temperature and pressure, lb/ft3 (kg/m3) C = empirical constant The 'C' value as recommended by API 14E and ISO 13703 are as follows: English Units: C = 100 for continuous service; C=125 for intermiitent service; for solids-free fluids where corrosion is not anticipated or when corrosion is controlled by inhibition or by employing corrosion resistant alloys, values of C = 150 to 200 may be used for continuous service; values of 250 may be used for intermittent service
For water at 1000 Kg/s, and C=250 as we have intermittent flow without corrosion anticipated, this calculates an erosional velocity of 8m/s.
Minimum Velocity Limits
Lower limits on velocity exist to provide a high enough flow in pipework to:
- clear air
- clear debris
- clear bacteria
- maintain a minimum flow through check valves to reduce seat wear from oscillation
Without the use of lower velocity limits, pipes end up oversized to the point they never see higher velocities, resulting in permanent pockets of air, and the collection of sludge. In turn this leads to bacterial growth, corrosion and poor system performance.
Most pipework is horizontal, with slight rises and falls due to layout or imperfections in construction. As such, air and dirt can find points to collect. Vertical pipes, such as risers, introduce different rules. Air naturally rises through water, and dirt drops. The direction of flow also becomes important, with water flowing down requiring higher velocities to clear air, and water flowing up will struggle to clear significant solid debris.
The important point to note on minimum velocities is they are not required at all times. The function of clearing air need only happen following commissioning when the system if full of air, and then periodically, to clear any air that has collected. Likewise with clearing debris.
These functions are best served by periodic high velocities, than steady low velocities.
The biggest danger with minimum limits is applying them at all times to all pipes, as this implies by-passes are required on all pipes to cover the eventuality of no load, and no flow. This argument is used both to justify the existence of bypasses and excuse the use of HIUs that maintain a trickle flow at all times, resulting in thermally inefficient systems while not performing the tasks minimum flows are designed for. There is no technical reason why water cannot stand still for periods, and even go cold, providing load response times are met when needed.
The specific argument that a constant low trickle flow keeps systems clean is one used to justify keeping keep warm modes turned on on HIUs is incorrect. Such systems may rarely experience high velocities (high Reynolds numbers) in pipework, as there is always heat on hand and design peak loads may never be approached. By comparison, systems that are left to go cold to improve efficiency, draw a much higher flow at startup, resulting in significant pipework velocities - and a functional system to move debris and locked air. This is the same reason that modern air purging functions on pumps employ a start-stop-start-stop method. There is a need to circulate chemicals in a system, however that function is satisfied by DHW demand and any anti-Legionella functions used. Systems are further assisted by periods of no-flow by allowing air and dirt to settle out in vertical risers towards air vents and drains, Sections of a riser where flow is allowed to stop (i.e. no bypasses or constant trickle flows), act as built in air and dirt separators on a network.
Flushing though decreased temperature
It is possible to increase operating velocities by reducing supply temperatures. HIUs will require more flow rate at the reduced temperature to satisfy demand. If this is ever to be used as a means of clearing air, for example, it should be combined with performance monitoring so as not to go to low and sacrifice function.
Pressure Losses in Pipework
The following is a useful online calculator with various materials.
Throughout this article we make use of our own Heat Network Calculator to perform more complicated network sizing and pressure loss calculations based on diversified loads.
Maintaining water quality is one of the key defenses against erosion, removing particles of all sizes from the system water using properly maintained dosing and separation systems. The following is generally accepted:
Conductivity max 10 μS/cm pH-value 9-10 Hardness 0.1 tHº (100-150 Mg/l CaCo3) Appearance clear and sediment free O2 max 0.02 mg/litre
See page 93 from Danfoss Guidance
Categories of Pipework
In order to make standards more scientific and applicable to modern networks, it is helpful to look at each section of pipework in terms of its function rather than assume all pipework follows the same rules.
The first two categories are:
- Pipework within buildings
- Pipework between buildings, underground
The next two main categories are:
- Flow Pipework - carries hot water to a load, split into two further sub-categories
- Permanently hot (core)
- Can go cold when not used (periphery)
- Return pipework - carries cooled water back to plantroom for reheating
These categories can then be subdivided into the following:
- Transport pipework - for moving water between plantroom and buildings or risers
- Risers - between floors
- Laterals - branch from risers to feed a row of properties
- Feeds - branch from laterals into properties
All flow pipework needs high levels of insulation to prevent unwanted losses from the network, however return pipework insulation requirements will be lower as the temperatures are lower. Where there is a fixed budget for insulation, flow pipes should be given preferential treatment as they contain more heat to lose.
Feeds, Laterals and Risers can often be left to go cold when not in use, providing DHW response times can be met when needed. Providing DHW in time requires water volumes to be lower, so the time it takes to clear the water is reduced. In turn this implies small bore pipes, and higher pressure losses than average. Where there is a fixed 'budget' of DP (pump pressure), flow pipework that requires rapid purging should be given preferential treatment. This is covered more in the next section.
The Peripheral Pipework Zone
Think of a heat network as a living animal, or a plant. You have a heart, arteries and veins, and capillaries, the latter being in the periphery of the circulatory system.
"Blood flows from the heart through arteries, which branch and narrow into arterioles, and then branch further into capillaries where nutrients and wastes are exchanged. The capillaries then join and widen to become venules, which in turn widen and converge to become veins, which then return blood back to the heart through the great veins."
Then a heat network.
Distinction needs to be made between the main network pipework and the peripheral pipework feeding properties. The reason is because thermal losses, within a buildings fabric, increase with the number of branches kept hot at all times, to the point where buildings overheat, and running costs escalate.
The defining point between the two network types is the point at which permanent circulation to pipework is hot at all times stops, and pipework is instead allowed to go cold when not in use - the peripheral region. This definition does however require hot water cylinders, or keep warm modes in HIUs to be turned off or timed.
The size of the peripheral region will depend on the type of hot water systems deployed. Where instantaneous heaters are deployed (standard HIUs) then the requirement to minimise delays in hot water production require the volume of flow pipework in the region to be low enough that the delay is satisfactory. The speed at which HIUs flush the primary pipework of colder water is also a defining factory, with the total time for delivery of heat to the HIU been the volume in the peripheral flow pipework to the most distant HIU, divided by the flow rate of flushing.
In the case of 15mm copper tube carrying 20 litres/minute, a water velocity figure of 2.5 metres/second would put the peak pressure drop per metre at 6000 Pa/m.
An HIU with flushing rate to 20 litres/minute, and a peak delay of 30 seconds, would allow a peripheral flow pipe of 10 litres.
A primary flow of 20 litres per minute will deliver 63kW of DHW load on a 70-25C primary drop, so is above typical peak design flows.
For 15mm copper tube, 20 lpm is equivalent to 3600 Pa/m (3.6 kPa/m), but a velocity below the 2.5 metres/second threshold. Such flow rates will be experienced less than 1% of the time. Flow connections between laterals, and HIUs can therefore be run in 15mm copper providing the impact on overall pumping duty is not significant.
The impact on overall pumping duty of using sections of flow pipework with a higher than average pressure drop can be countered by increased diameter return pipework in the feeds to HIUs and laterals where pipework temperatures are typically below 25C so heat loss is insignificant and we are not concerned with flushing cold legs.
Sample Calculations based on 15mm and 22mm feeds
These two calculations demonstrate a potential 'sweet spot' in sizing for 15mm. Delivery times are between 18 and 27 seconds depending on the temperature of pipes following use, with efficiencies in the 80%s.
Switching to 22mm.
The additional delay is only 2 seconds, leading to the conclusion that although 15mm can be used under certain circumstances, the saving in delay does not justify preference over 22mm and the lower resulting pumping costs of using 22mm unless you are at the margins.
To take advantage of the gain from dropping to 15mm pipe, the DP allowance for feeds into properties needs to be as high as possible, as the inability of 15mm tube to supply more than one property limits its use to flat feeds. The one exception may be when cylinders are used, as peak flow requirements on laterals are a fraction of what they are for instantaneous systems.
A point of note is that the higher pressure drop introduced at peak flows makes if more difficult to maintain the correct differential pressure across HIUs. Pumps may need to be more responsive as there is less residual pressure to work with. It is preferable to measure DP at a riser rather than the end of an index lateral as pressure fluctuations along the laterals result in poorer pump control.
Boost Flow and DHW Delivery Times
One of the most significant figures needed in calculations involving instantaneous HIUs, is that of Boost Flow, or rate of purge. This is the rate at which the HIU draws cold water from the primary system until delivered heat reaches temperatures high enough to deliver hot water.
It is a figure that can be ascertained from independent HIU test results. For our own range of electronic HIUs the maximum setpoint is 20 litres/minute. This flow rate drops back to normal as the flow temperature rises (so as not to overshoot).
Mechanical DHW systems using purely thermostatic control (not flow proportional) will have an inherent boost flow as the valve opens until temperature takes effect. However, this is academic, as once in use the thermostatic control maintains setpoint at all times (working purely on temperature), thereby maintaining primary temperatures.
The other significant figure a designer needs to know before embarking on pipe selection is the clients peak delivery time limit. We use a figure of 30 seconds for a starting point, as this is comparable to a combination boiler. Drawing on experience from private contracts, it is still quite rare for customers to request a secondary return system in order to speed up tap delivery. Most of the population are quite happy waiting a few seconds for heat to come out of a tap, especially if it meant considerably higher hot water bills if you wanted to speed it up. The peak delay time is also a peak - most likely experienced by the first property on a cold leg to run a tap in the morning. It takes time for pipes to go cold, and while there is occasional use in adjacent properties, the lateral pipework will remain above 50C, and delivery times much lower, if not instantaneous.
The maximum delivery time for DHW is the volume of flow pipework that needs clearing, divided by the Boost Flow.
10, 20, 30 Limit
10 litres of (cold) flow pipe, and an HIU Boost flow of 20 lpm, gives you a 30 second peak delivery time.
The 20 lpm limit is applied for 22mm pipework (the smallest bore typically used) to keep velocities within reason (between 1 and 1.5m/s), and high enough to ensure regular movement of sediment in primaries.
This rule implies a limit of 8.4 metres of 22mm copper tube between a riser (left hot) and an HIU, before delay times start potentially going over 30 seconds.
Note that very few HIUs can achieve a 20 lpm boost flow. Purely thermostatic HIUs, as well as electronic HIUs have ability to draw high flow at startup, however thermostatic HIUs remain hot at all times so never see startup. Flow dependent HIUs draw flow relative to tap use and the thermostatic control side has less effect. Such HIUs have a much lower boost flow, leaving electronic systems as the only way to both close off flow when not in use, and then exercise a controlled purge on start-up.
Boost Flow and Water Quality
With a need to introduce occasional high velocities and turbulence in water to clear settled debris, the boost flow function is a means to ensure that. An HIU sized at 50kW may very rarely see a load over 12kW when not fully occupied, and may never see pipe velocities needed to to prevent sediment collecting. A high boost flow overcomes that by drawing the higher flow rates (of cold water) from the primaries on startup.
Pipes are sized to flow rates, resulting from peak loads, and loads are calculated by diversity, so the proper use of diversity is a key part to sizing pipes efficiently.
Diversity is a way of expressing the probability of everybody opening their taps at the same time. As such, it depends on numbers, with a more consistent average arising the more properties you look at. So, for one property there is a good chance you will see the odd occasion when two occupants are running taps, so 100% diversity.
Its easiest to think of diversity sizing as choosing the size of a supermarket car-park. The question is, what is the probability all customers will arrive to shop at the same time. One never takes the assumption every customer will want to shop at the same time, or the car park would need to be far larger than the shop. Nor does one assume everyone will perfectly space their shopping to make use of only a few parking spaces. In reality, there are times when more people shop than other times, and only experience can tell us the best balance. Diversity in district heating is similar, but based on having enough capacity to service known times of popular demand.
Diversity curves are typically measured from a field trial conducted at one point in time on a certain test installation. Better curves are derived from realtime data, catagorised by site type, but this kind of information is only just becoming a reality.
The other way to derive diversity is from probability theory, as described in the CIPHE Building Services Design Guide.
Diversity is the estimate of the maximum hot water use at a given moment, so we can work out peak loads. so the basic thinking is if we know the length an outlet runs for, and the time during which it is likely to be run, we can calculate a probability of how many run at the same time.
This is why consultants oversize - we are talking probabilities - and there is always a chance everyone will run there taps at the same time. A bit like winning the lottery, but possible. So - how much do you oversize plant-rooms and pipes as insurance against winning the lottery (for 2 minutes once upon a time)? The real question is, what is an acceptable probability that performance will drop to the point where tap temperatures (40C) cannot be met, when its also -5C outside?
Probability theory is quite involved, which is why actuaries are paid so highly to predict probabilities in the insurance business, so it is no wonder that diversity is best measured from real world data, and why I will not attempt to go any further in regards to calculations (until we have a decent set of field data to test against).
Existing Diversity Curves
There are a number of standards concerning diversity, some more generous than others, and sometimes abnormal load patterns take over (e.g. sports facilities). It does seem apparent from a growing database of metered data that loads are in reality at the low end of the spectrum. The regular over-sizing of systems as a diversity safety factor is generally down to a lack of data on contract relevant real world patterns, but this is something that is changing as more and more data from the field is accumulated and categorised.
The following graph shows the energy requirements for various numbers of homes, and various diversity curves. These include:
- DS439 - Danish Code of Practice
- SDHA F101 - Swedish District Association
- REDAN - Danfoss Redan
- COHEAT - Using EST (Energy Saving Trust) Data
Scaling Diversity to Peak Dwelling Load
A common misunderstanding about diversity is how it is scaled to suit different dwellings. The curves are based on peak loads per property of around 37kW. From a domestic heating engineers point of view this 18 litres/minute to taps, equivalent to a large combination (combi) boiler, and is capable of running a reasonable bath tap, two standard showers, or one luxury shower. The standard solution to everyone's problems, normally.
While this load is fine for normal dwellings, when a heating engineer is thinking to service a 3 bed house, with potentially 5 occupants, two bathrooms, and kitchen, in London (where to buy a property one may expect high end fittings), the obvious option would be to fit a hot water cylinder. A larger combi may not be the first option usually down to gas supplies and flue, but a cylinder would allow one to deliver water to all outlets simultaneously from a much smaller boiler. However, one needs to still ask what size of combi would do the job, to check it could work. In district heating we ask what size of HIU, and how to use standard codes of practice to determine loads over multiple dwellings ?
One must not just take the size of a specified HIU and scale diversity relative to 37kW. Calculations are driven by the number of outlets, the type of outlets, the number of occupants, and available mains water capacity.
If ones thinking is I have two luxury showers, each requiring 20 litres/minute. An 80kW HIU can drive over 40 litres per minute at shower temperatures. Voilà. Pipes are then sized on this peak load, and with the low diversity in small numbers, the pipes feeding properties will also be large to accommodate the luxury loading.
In reality, 40 litres/minute is more than most domestic mains water supplies will deliver, at the pressures required to hit shower fitting specifications, especially if water meters are fitted. Furthermore, the full shower capacity (when so much is available) is not used that frequently. Its nice to be able to turn up the shower to blast oneself with body jets, but what are the real chances of two people in a property both doing so at the same time (in different showers that is)? Scale this up to adjacent properties, and it becomes clearer that smaller delivery pipes would cope in all but the rarest of loads.
One also needs to think of the nature of not delivering peak DHW load. An HIU specified thus at 80kW, with a nominal supply temperature of 55C, will deliver higher outputs at tap temperatures (38C from shower head). The use of an oversized HIU will simply result in higher return temperatures than a smaller HIU pushed to the limits, and running below setpoint at peak load.
On a larger HIU, at peak load (within nominal range) the 55C supply will be mixed down with a significant amount of cold water to leave the shower head at the desired delivery temperature. An HIU sized to peak load at a delivery of 42C, will result in a higher portion of the flow going through the HIU, and less mixing at the outlet. This higher flow through the HIU can cool primary water more effectively, delivering exactly the same shower experience, as well as more efficiently than an oversized HIU to be safe.
Any single output rating for an HIU is meaningless. Calculations on sizing of HIUs, and the pipework around them, always needs to be done with reference to manufacturers performance figures at the design point. All heat exchange manufacturers provide software for this purpose, and HIU manufacturers will provide specific calculations on request, providing the primary flow rates and temperature drops needed for network pipe sizing.
There are however two limits on how low one can drop the target temperature at peak DHW load:
- Pressure drop through an HIU will increase, and will become the limiting factor
- Some point of use mixing taps for care use require a certain minimum difference between set delivery temperatures and hot supply temperatures, typically 8C higher.
Certainly, where an HIU can deliver peak loads at temperatures in excess of 48C there is rarely a need to up-size.
The other disadvantage of over-sizing is it may cause more problems that it fixes. Certain HIUs have relatively high minimum flow rates, and it may be more common to experience no hot water when at low flow from a sink tap (a common event), than not being able to achieve the full 20 litres per minute simultaneously from both shower (a rare event).
In terms of heat exchange efficiency, heat exchangers have a lower operating level also, below which efficiency drops off and calculated performance is lost. This is a result of laminar flow, rather than turbulent flow, within the heat exchanger, and water passes through the gaps rather than conducting to the metal plates. The volumes of larger plates is also larger, and to a higher extent they store some little heat (enough for a couple of seconds delivery). These effects combine and result in higher return temperatures at low flows.
The bottom line when scaling systems in individual properties to meet instantaneous hot water demand, is if it impacts on pipe-sizing to the extent that pipe volumes require keep warm modes to be turned on for reasonable response. This can introduce a step-change in network efficiency and a significant increase in running costs. At this point, hot water storage should be looked at instead of (or preferably in addition to) instantaneous systems, as a means to provide high-end DHW performance without impacting on network efficiency.
This is the system devised by the Institute of Plumbing to work out the size of pipes in buildings based on the diversity of tap use. Unlike the district heating diversity charts, it takes into account the specific type and number of outlets in a property.
Once you start looking at lots of dwellings, statistical methods based on real world data become more accurate than scaling up sizing methods for individual properties. On a single system we need more margin for error to account for unusual activity, however at a large scale, things get very predictable and safety margins can be much smaller.
Peak Load Calculator
The following calculator can be used to examine various peak load potentials. Outlets are added and then dragged to the start time (note mouse release must be over outlet you are dragging). A boiler input can also be added to calculate a thermal store size requirement.
The calculator cover a three hour peak load period. It provides four graphs covering
- peak mixed flow to taps
- energy used
- buffer energy
Forgive us, this is an old Flash based tool, with the old company info - please ignore telephone. It has proved invaluable to us over the years for rapid generation of load profiles in a graphical format easy to understand.
When sizing pipework there is a tendency to up-size for 'safety' reasons, or to reduce pump energy further. This is often at the expense of thermal efficiency, and should be avoided given there are other less costly methods of delivering more heat than designed for:
- Temperature difference
- Differential pressure
It is also incorrect to allow additional margin for exceptional weather events. Instead, systems should be designed to meet exceptional weather events (1 in 10 years cold, or -10C, for example) and the safety margins built in through the use of methods that do not sacrifice efficiency at other times.
The supply temperature into a network is generally run to maximise efficiency of the boiler of heat source, as one would expect, however are capable of running at higher temperatures than the most efficient temperature. Boilers, for example, can often achieve 90C if asked to do so, but would normally be set nearer 75C. These lower temperatures also help reduce heat loss from pipework.
The sizing of pipes is done at peak loads, and is based on the flow rate required to supply that load. By increasing the supply temperature to a heat network one directly reduces the flow rate required to deliver loads, and the capacity of the network.
Energy (kW) = Flow (Kg/s) x Temperature Drop (C) x Specific Heat Capacity of Water (4.2 kJ/Kg.K)
With peak load return temperatures, for example of 40C, increasing the design flow temperature from 70C to 90C would reduce the required flow rate through pipes (and hence pipe sizing) by 40%.
Given than pipework is generally rounded up, and the peak load we are designing for is very rarely encountered (-10C or 10 year low), it makes sense to size pipework close to the peak temperatures the heat source can deliver. For 99% of its life, the temperature of the system can be turned down.
Differential pressure across networks, provided by pumps, determines the flow through pipes. Increasing pump head will increase flow and the capacity of the network.
Unlike pipes, pumps will fail at some point, and typically a system will have redundancy in the form of an additional pump. The failure of a pump is a 1 in 10 year event (hopefully), on the same scale as -10C weather, and the odds of the two events coinciding are that long that systems will always have some extra pump capacity available. Pumps are known to fail at start-up after not been used, or when overheated through inadequate flow; neither conditions arise at peak load.
Pump controls should therefore be setup to enable the extra pump to assist, if ever needed. Duty/Assist, rather than Duty/Standby.
Lower limits on water velocity do not apply in vertical pipework. Movement of air upwards, and dirt downwards is not assisted in any form by increased water velocity, but rather inhibited.
On systems where riser flow pipes are kept permanently hot, there is little loss in thermal efficiency by going up in riser pipe size, as distances are relatively small (4 properties per floor means under 1m of riser per property). Riser return pipes (should) lose very little heat as temperatures are low, and over-sizing has no other cost than in capital outlay for larger tube.
Where a riser is within the peripheral zone, and left to go cold, the need for rapid purging of cold water prevents pipes from been oversized - or volumes will be too great and delivery times longer than expected. Such risers should be down-sized if limits permit.
Differentiating Return Pipe Design
Flow pipes are there to deliver heat. In a perfect system, all this heat is extracted and then the primary water returns cold to the plantroom to be reheated.
Historically systems have been installed with bypasses resulting in hot return pipes and a need to also insulate them. However, with return temperatures now averaging under 30C potentially, the question of using reduced levels of insulation on return pipes to save cost is one that needs to be looked at closely.
A specific function of flow pipes in peripheral sections of a network, where pipework goes cold, is to deliver heat rapidly, and that requires smaller volumes. The return pipes however need not carry water at speed so impose no local sizing restraints. In other words, they can be bigger and save valuable pump head without incurring significant extra heat loss as a result.
The following graph compares the pressure drops over a network index branch, with the red line representing calculation based on a fixed pressure drop per metre, and the green line showing how pressure gained from over-sizing return pipes can be used to downsize the critical flow pipes. Overall pump energy consumption is the same.
It should be noted, that as designers always round up pipe size, in reality there will generally be spare differential pressure as a result. This can be seen below.
Combining the effects of rounding pipe size as well as differentiating flow and return results in something more like the following:
And finally (one would hope), how it looks in full, further combining the effects of elbows and tees. This graph shows the system at peak load, and also in a no load state.
However this last graph introduces a few more points to note. To minimise pump energy consumption it is highly probable that DP sensed from the top of risers is used to control pump heads. If this is the case, as shown on the graph:
- the minimum riser DP setting - related to pump consumption at lower loads - is minimised by having virtually no pressure drop on the feed and lateral return pipes (big pipes).
- differentiating on return pipe size in the riser and transport pipes has NO effect on providing more DP across HIUs.
A final difference is static pressures. Flow pipes run at higher pressures than the return. This is seen to exaggerated effect nearest the plantroom, where DPs are highest. The flow may see 4 bar higher pressure. Return pipes run downhill in general, and the water they contain is heavier. It is conceivable therefore to envisage a system where the flow consists of small-bore high pressure pipework that is very well insulated, and the return - low pressure, large bore and low-insulated.
Recent changes in boiler flue design, with the introduction of plastic flue, has been a direct result of better boiler efficiencies. The return pipe on a communal system can be thought of as a flue - if the system is efficient enough, and return temperatures are low, then plastic becomes quite attractive.
As it is now possible to run systems where return temperatures never rise above 45C, life expectancy of plastic pipework on the return is extended.
There exists a limit on plastic pipe that kicks in above 70C typically. At this point life expectancy starts dropping rapidly, with 90C been an upper limit that results in short lifetimes.
Mixing copper or steel on flow pipework, with plastic waste pipe on return pipework, is perfectly feasible. One should consider drops in pipework, as one does on waste pipework, with velocities inherently lower. Lower velocity limits can be justified on such a basis, as waste pipe is designed to clear debris. In building design, such next generation building networks could be enabled by suitable pipe routing allowance at building design stage.
Example Pump Range
Assuming maximum DP on system of 1.5 bar (150kPa) based on range for the common Magna 3 pumps.
This range covers loads of
- 100 two-bed properties, based on 20m3/hour at 150kPa and 70C flow.
- 150 two-bed properties, based on 20m3/hour at 150kPa and 80C flow.
- 180 two-bed properties, based on 20m3/hour at 150kPa and 90C flow.
And the equivalent Wilo pump, but duty assist.
These calculations highlight how the use of elevated primary temperatures at times of peak load enable a given pump (and pipe network) to drive more load. By rising the flow from 70C to 90C we have added 80% load. Therefore, pump and pipe sizing should be done with consideration to the available temperatures from specified heat sources saving the highest temperatures available for -10C outside conditions (i.e. exceptional circumstances). Once pipe and pump sizes are fixed, and it comes to operational matters, the flow temperatures can be reduced to target highest efficiency, rather than peak output. Plus, if it cost no extra in efficiency to distribute at 90C rather than 70C, doing so will save pumping energy.
Further calculations and sizing can be done using the online tool:
Controlling Pump Pressures
Pressure losses are highest at peak loads, and that is where pumps are sized.
They do however not need to run at these pressures for much of the time, and as pumps use more energy at higher heads it makes sense to reduce pressures where possible.
This however requires some means of controlling the pump. Some pumps come with built in proportional control that ramps up pressures as loads increase, however it is beneficial to add a double check to the control system based upon sensed differential pressure at the top of an index riser, with enough margin to cover losses through HIUs, feeds to properties and laterals (typically 75kPa).
Moving objects, including water, contain kinetic energy and momentum. It takes time to add or remove energy from an object or fluid, and the time you have decides the forces at play - just as in a car. The name water hammer gets its name for a reason. When you stop water suddenly, by closing a valve, the energy needs to go somewhere.
As water is in-compressible the impulse forces travel to find a point where energy is absorbed. Air in systems is the normal mechanism for absorbing such forces, as air can compress absorbing energy as it does so. This is why we use air in car tyres, expansion vessels and hammer arresters, with the latter been the standard way to protect pipework. Think of them as shock absorbers for your pipes.
The greatest forces can be generated where air is trapped in pipes. The air can allow a significant volume of water to impact at one point, without the ability for pressure wave to be transferred through the water any further. Pressures experienced under such events can easily exceed the rated limits of connections and seals and result in sudden failure (compression fittings are a weak spot for such failures).
With more and more attention been paid now to the removal of air from systems, the opportunities for water hammer increase, and it becomes more important to fit hammer arresters in place of trapped air in pipework and radiators on direct systems. They should be fitted at the ends of pipe runs, as the pressure waves initially travel in the direction of flow (conservation of momentum). The tops of riser flows would be an ideal position for hammer arresters, along with hydroscopic air vents.
The following calculator gives a feel for potential pressure increases from water hammer. One can see how as the time it takes for a valve to close has a considerable effect on the pressures generated, as does the pipe length and velocity. The worst cases of water hammer personally experienced have been down to solenoid valve closure in long mains supplies feeding washing machines, where a reasonable flow can be closed of in less than a second generating a very large pressure wave, that can resound around a buildings pipework. In one instance the pressure wave from a solenoid was heard and seen to activate the flow sensors in hot water systems a number of floors away.
In a heat network, the water hammer induced by pumps stopping is created in reverse. As the pressure source collapses with only a small flow available through a dead pump, the momentum of the water creates negative head behind it. This form of hammer is unlikely to cause pipework to see significant increases in pressure, but may still result in in a low pressure wave that may cause noise. There will be momentum in the pump itself, so closure time will typically be over a few seconds.
In summary, water hammer should not effect the sizing of heat network pipes under 50mm within buildings, however best practice should always allow for the installation of hammer arresters at the end of pipe runs to overcome pressure waves experienced during the filling, venting or rapid circuit isolation of pipework.
Peak static pressures are reduced using mid-stage pumping, potentially enabling the use of lower cost pipework.
Locations for a mid-stage pump would be at:
- the entry to a building (on a large network)
- at the base of a riser
- on an HIU
- on a direct fed UFH circuit
They need not be used on every riser, building or load - just those furthest from the plantroom where DP is lower, or where internal DP requirements are higher than average. Risers close to the plantroom will be withing the DP range of plantroom pumps.
One potentially important benefit of mid-stage pumping is it decouples load from source, making a block of load more independent - so you can have more than one plantroom feeding a network, each with pumps that only need to be sized for the main network transport pipework.
With 150kPa (1.5 bar) been a practical limit on smaller pump families such as the Magna3 noted above, at decent flow mid-stage pumps improves the chances that pumps across a network can remain in this lower pressure class. Where a 3-4 bar pump may be needed with one central pump alone, mid stage pumping may allow the use of two 1.5 bar pumps.
Pumps are much less efficient at low loads, so smaller head pumps will be more efficient at low overnight loads. At lower loads, it may be sufficient just to run the central pump, with mid-stage pumps coming in as required.
Peak DP across loads close to the plantroom are lower potentially removing the need for DP protection anywhere other than direct heating circuits to match TRVs (5-20kPa).
When not in use, a non-return valve around a mid-stage pump, in the same direction of flow, allows it to be circumvented by upstream pumps. It may also be advisable to locate an expansion vessel on the inlet site of a mid-stage pump, to arrest hammer from pump interaction.
Underfloor heating circuits (UFH) fed directly from a district circuit all require a pump in order to push return water back to the flow to reduce temperature. They are also used to overcome internal pipework losses, but if sized with spare capacity they can be positioned to pull from the network and act as mid-stage network pumps. In domestic system design it is common to rely entirely on the UFH pump to pull from thermal stores. With HIUs (rather than cylinders), peak loads into properties are required for instantaneous hot water rather than heating, and an UFH pump will not effect local pipe sizing.
Mid-stage pumping examples
The following examples show pumping at the base of a riser, at full load and at no load (where just the plantroom pump is needed).
In this example, designed to cover very long distances, pumping is alternated between flow and return pipes. The point mid network, where the return pressure is higher than the flow, is an ideal location for additional heat input. Using very high head pumps, rather than mid-stage pumping, makes it almost impossible to create a network that remains at low static head throughout, and allows additional heat inputs to be added on to an existing network.
We generally allow for an additional pump on systems to enable operation to continue in the case of a pump failure. This additional pump, effectively sitting there not needed, but for a couple of days every decade, can arguable be used to cover other rare events, such as -10C heating conditions.
The chances of a pump failure coinciding with -10C outside are so small that it is not worth stepping up pipe sizes to cover such an eventuality alone. Doing so would be guaranteeing a less efficient system for life - to insure against an event that will likely never happen. The use of the additional pump to assist load should be treated as the safety margin in system design, so there is no need to automatically round up where other calculations leave pipe size selection on the margin when looking at peak predicted loads and pressure drops.
It is always worth doing an additional calculation on design with the redundant pump pulled into use. This will act to demonstrate the true limits of the system and see if it is capable of dealing with, for example, -15C conditions, or a higher DHW diversity factor.
Differential Pressure Limitation
There are three reasons why DP limits to be considered:
- To protect valves from excess pressure that prevents them from closing.
- To improve the accuracy of valve control.
- To regulate systems where certain loads (nearer source) draw excessive flow and starve other loads.
We are typically concerned with HIU controls, or TRVs on direct heating systems.
In general, HIUs always have options to cover higher DP. With the extremely low heating loads now designed for in buildings, combined with higher temperature drops, the DP required for central heating is far lower than historically required and more attention needs to be paid to selection of the appropriate DP valve selection, or TRVs. A DP much over 5kPa may not suitable for a standard TRV, and will require TRVs with lower Kv valves and more precise setting.
In a modern heat network, there is no need for commissioning valves on distribution networks. Any flow limitation should be done at load, with network pipes left clear of valves and obstructions. Ideally there should be nothing between a plantroom and an HIU except full-bore isolating valves where necessary.
Moving water without a pump, using the fact that colder water is heavier than hot water so can initiate flow.
Not normally the domain of district heating, but worth covering, especially when one considers that at least one tower block has run for years with some success using nothing but thermo-syphon effects to drive both hot water and heating.
Back in the old days, gravity systems were normal, with gas boilers regularly running hot water cylinders off a gravity circuit, but now they are almost exclusively found in domestic wood burner systems. Wood burner systems require a means to transport heat away from a wood burner under conditions of power failure (no pumps) or system water starts boiling. With some wood burners over 30kW, designing systems to ensure this is safely transported away under nothing but gravity is an art, rather than a science.
Reducing the laws of gravity to a few equations is not easy, as the forces in play are so low that more factors come into play. These include:
- Temperature differences
- Height difference
- pipe sizing
- sharpness of bends
- angle of pipework
- air locking
The complexity is why one may never see it applied in practice - unless by mistake, which is what we came across when we were asked to replace an old cylinder in a tower block flat, with an HIU. The primary system was down when site surveying and DP tests were omitted, but once installed the system would not work, and it didn't take too long to figure out that there was no DP available on the primaries.
Further investigation up and down the riser showed there to be no pressure anywhere, despite the fact that everyone had services. In fact, the pump for the tower had been completely omitted, and only gravity circulation had been in place for a number of years. A simple addition of a pump inline with the HIU, and later on the raiser, fixed the situation, but that it worked (despite grumbles) for years proves such systems can work.
Had the system been designed with this as a goal, then performance would have been better - direct heating for example would have made a big difference to gravity performance. The fact that it was a tower block, certainly helped the initiation of gravity flow. The hot water cylinders charged up very slowly over hours using a trickle of gravity flow, whereas the HIU, requiring a decent pressure to drive instant load would not be an option in such a design.
Completely eliminating pump power from a system is a nice thought, however the additional pipe sizes required, and higher heat losses that result, work against energy efficiency. It does however have one last implication - it may be possible to maintain trickle flows through risers purely on gravity when there is no load or just background heating.
Standardising First Fix Brackets
Designing pipework for a heat network needs to be considered right up to the point of entry to a heat interface unit in properties. As such, it is worth considering the benefits of standardising the interface used to connect to HIUs. The only standard that currently exists is from the Netherlands, using a rail of six bottom-entry connections, and allows pipework to be run up the back of an HIU as well as downwards. The six connections enable connection of the three main variants of HIU:
- Twin plate, indirect heating
- Single plate direct heating
- Single plate direct with pump and temperature control
At this point we must declare a vested interest. We are one of very few manufacturers that match this bracket so it is in our interests to see this specific interface adopted.
Calculating Temperature Drops
Temperature drops along pipe are calculated by balancing the temperature drop of water under flow, with the temperature loss of the insulated tube to the surroundings, based on LMTD (logarithmic mean temperature difference).
It is an iterative process, improving the estimate with each pass until the result varies less than the precision required typically 0.1 to 0.5 of a degree), and thus best suited to software rather than manual calculation. Our Heat Network Calculator uses this form of calculation as the basis for working drops across a heat network.
Temperature drop under flow
Heat Loss [W] = 4200 x Temperature Drop [C] x Flow [l/s]
We assume that a tube has two ends (which we call "A" and "B") at which the hot and cold streams enter and exit; then, the LMTD is defined by the logarithmic mean as follows:
where ΔTA is the temperature difference between the two streams at end A, and ΔTB is the temperature difference between the two streams at end B. With this definition, the LMTD can be used to find the exchanged heat in a heat exchanger:
Where Q is the exchanged heat duty (in watts), U is the heat transfer coefficient (in watts per kelvin per square meter) and Ar is the exchange area.
We can work out the value of U x Ar from a know heat loss for the chosen pipe and insulation, as provided by the manufacturer.
Laminar and Turbulent Flow
The Reynolds Number
This is a numerical representation of the turbulence within flow.
- Pipework within buildings defined as either peripheral flow pipework, or as transport pipework. The former are the flow (not return) pipes supplying heat to properties that branch from permanently hot feeds such as risers. All other pipework to be considered as transport pipework, where thermal losses give way to pressure loss and pump considerations.
- To avoid unnecessary thermal losses, peripheral pipework should be left to respond entirely to demand including the need for Legionella protection and frost protection. Design should ensure flow stops and temperatures drop when there is no demand. As higher percentage of the flow network should sit in the periphery as available pump head allows.
- In the case of instantaneous domestic hot water generators (HIUs), the peripheral pipework region (volume) is calculated from the rate cold water is purged during start-up, multiplied by the maximum allowed start-up delay for hot water. Calculation is done on the index flat, being the furthest from the available heat, working back from the property until pipe volumes result in excessive delivery times.
- Each length of pipe within a building should be sized according to the calculated peak load derived from downstream diversified loads. The selection of diversity curve used should be made according to a recognised standard, or documented data concerning similar patterns of use.
- Instantaneous domestic hot water generators should be tested independently to enable the following to be determined
- purge flow rates following standby
- return temperature performance for hot water, space heating, and stand-by operation
- In the case of cylinders for domestic hot water, where there is no requirement for rapid response of the network, the maximum start-up delays are significantly longer, and therefore the peripheral region is significantly larger. Their may be little need to maintain temperatures through much, if not all of the network, other than for frost protection. Standing heat losses from cylinders should be included in energy calculations.
- Calculations should be made to estimate annual pump energy use, as well as thermal losses from the entire network. This should be broken down into transport and peripheral pipework, and include HIUs and hot water cylinders. Thermal losses are defined as the thermal energy generated by heat sources, less the energy used in the direct (out the tap) provision of domestic hot water and necessary central heating.
- In the case of copper and steel, peripheral pipework should be sized to provide a velocity between 0.75 and 1.5 metres/second at peak design load. In the case of plastic, manufacturers limits apply, providing noise considerations are met, and the effect of fittings is considered. Where more than one pipe size is within limits, the smaller of the two should be selected.
- It is beneficial to keep pipework velocities in all horizontal pipework (both flow and return) within buildings as high as the limits of materials, and noise permit, in order to help move debris and air. Such pipework should be sized to ensure a water velocity over 0.75 metres/second during peak hot water demand.
- Central pumps should be variable speed controlled, through the use of differential pressure monitoring, to ensure the minimum differential pressure required for operation is met across loads. This is to reduce pumping energy to a minimum at times of low load. A backup differential pressure sensor should be fitted in the location of the pump as a backup in case of signal failure from remote sensors.
- In order to reduce capital requirements for pumping and DP limiting on smaller (single complex) networks, distribution pumps should be sized to provide a head of 140kPa at an all time peak load (-10C outside/10 year local low). Where pressure loss across a network becomes too great to achieve this, one should examine the use of:
- higher supply temperatures at peak loads
- reducing return temperatures, and hence peak network flow rate by either
- direct central heating
- better performance plate heat exchange into properties for both hot water and heating
- higher peak pump heads
- mid-stage pumping
- hot water storage to reduce peak loads. Index properties most distant from plant room should be looked at initially, as well as those with abnormal DHW loads.
- Systems should have certified testing of water quality, including pH level, on a regular basis. In the period during and following commissioning of systems, checks should be made on a monthly basis, and for a period of 3 months following final hand-over.
- pH levels in heat networks should be maintained between 9 and 10.
- Side stream filtration should be fitted to remove smaller particles and air.
- Hammer arresters should be fitted at the end of long runs, and risers.
- Automatic air vents and anti vacuum valves should not be fitted on network pipework within occupied buildings. This is to prevent excessive water hammer caused by the exit of air from a system, and to prevent the ingress of air under negative head condition. This further prevents the rapid discharge of system water when there is a breach in the network.
- Thermal storage is typically required to buffer demand on a network from heat sources. This allows supply to be sized to meet average demand over a period, typically 24 hours, with thermal storage sized to meet peaks in demand, and an allowance for the startup time of heat sources.
- Where space limits make the desired volume of thermal storage an impracticability, their size can be reduced by increasing boiler size. The correct balance of store and boiler sizing needs to be struck to minimise capital costs.
- Where hot water cylinders are used exclusively along certain branches, one should consider timing and/or remote instigation of cylinder reheat to coincide with each other, and thereby enable peripheral pipework to remain cold for longer.
- Where hot water cylinders are used, return temperatures, and hence peak network flow rates for hot water, can be significantly reduced, through the use of HIUs to recover hot water cylinders, instead of coils. Where instantaneous HIUs are partnered with a hot water cylinder, it will significantly reduce peak network flow rates, while improving delivery to hot water outlets and introducing the option of electric backup and assist. As such they can be used to cover properties on a network with exceptional DHW demand, without a negative impact on efficient pipe sizing.
- Where, at a later date, the peak DHW demands may rise above initial design, an allowance in properties for the later addition of a hot water cylinder should be considered.
- BESA Standards for HIUs
- Copper Development Association, CIPHE
Article Compiled By
Richard Hanson-Graville, Technical Manager, Thermal Integration Limited.
- Masters in Mechanical Engineering from Cambridge University
- MOD (Defence Research Agency) Graduate Training
- Technical Director for leading UK manufacturer of thermal stores and plate heat exchange systems since 1996
- Fellow of the Institute of Plumbing (now CIPHE) since 1998
- Author in the above's Plumbing Engineering Services Design Guide 2002 (Thermal Storage, Plate Heat Exchange)
- Co-designer of the most efficient Heat Interface Unit tested to date under standards adopted by BESA designed to improve the efficiency of heat networks
- Heat Network Calculator software