Solar Panel Efficiency in SAP 2009 – Products with High Second Order Coefficients 3rd May 2011 Stuart Elmes MA, Viridian Solar For simplicity, SAP 2009 uses only first-order efficiency value for solar heating panels. This works well for most solar panels that have a relatively straight-line efficiency characteristic, however solar panels with poorer insulation and a more curved efficiency characteristic are unfairly rewarded. A simple enhancement of the current calculation is put forward to remedy this anomaly and make the calculation in SAP Appendix H more accurate. Table 1 shows performance characteristics of two flat plate solar panels2. These panels were selected for comparison due to their similar zero loss coefficients and aperture areas (2.5 and 2.35 m2), but very different heat loss coefficients. Background The UK government publishes a methodology for estimating the energy performance of dwellings, the Standard Assessment Procedure1 (SAP). Given a set of input data such as U values of building fabric and performance of heating systems, the calculation estimates the energy consumption and carbon dioxide emissions from any given dwelling. In the 2005 revision the calculation was modified to include a more refined estimate of energy savings from the use of solar water heating. Where previous versions of SAP assumed a flat rate of energy benefit from solar heating panels (per square metre installed), SAP 2005 captured the diminishing return produced by solar water heating systems as they increase in size. The calculation was also structured to reflect the better energy outcome from higher performing solar panels by using accredited test data from EN12-975 part 1, the European standard for testing the thermal performance of solar heating panels. These tests describe the performance of a solar heating panel using three values: zero loss efficiency (η0), first order heat loss coefficient (a1) and second order heat loss coefficient (a2). The efficiency of a solar heating panel can be calculated at any operating point using an equation in the following form: η = η0 – a1.{(Tm – Ta)/ GT} – a2.GT.{ (Tm – Ta)/ GT}2 where η is the panel operating efficiency Tm is the average temperature in the panel (°C) Ta is the ambient temperature (°C) GT is the incident light energy (W/m2) For the SAP calculation, it was decided to use the zero loss efficiency (η0) and the first order heat loss coefficient (a1) parameters to characterise panel performance. This was understood to be an approximation to real-world performance but, after analysis of published test data from a number of solar panel products, it was decided that this was sufficiently accurate. “Hook-Nosed” Solar Curves There are a growing number of solar heating panels in the market where an attractively low first order heat loss coefficient conceals a poorer overall thermal performance. These products have a solar efficiency curve with high curvature - their efficiency fades more quickly as the operating temperature rises above ambient. The decision to use only the first order coefficient for simplicity means that SAP does not properly reflect the performance of such panels. Although they are still few in number, products with efficiency figures which can exploit this “loophole” are rapidly gaining favour with SAP assessors. Aperture η0 2.35 2.50 0.790 0.795 (m2) Panel A Panel B a1 a2 ceff (W/m2.K) (W/m2.K2) (kJ/m2K) 2.414 3.90 0.0490 0.0125 8.09 6.30 Table 1 Efficiency characteristics and heat capacity Ceff for two similar size flat plate solar panels. Figure 1 Panel A has a better first order heat loss coefficient than Panel B, but by virtue of its higher second order heat loss coefficient is of overall lower performance than Panel B. Figure 1 shows a graph of the efficiency of the two panels at an irradiation of 800W/m2 as the average panel temperature increases from ambient. As can be seen, Panel A has a far greater curvature in its efficiency plot due to its higher second order coefficient, resulting in a generally lower overall performance across the typical operating range for solar water heating. In SAP 2009, only the first order coefficient is used, whereas T*Sol, a commercial solar modelling package takes both first and second order coefficients into account. A solar energy calculation was made using Appendix H of SAP and using T*Sol for each of the two panels. In SAP the floor area of the building was set to 85 m2, producing a daily hot water demand of 100 l/day. The total panel area was set at 5.0m2 for both panel A and B to remove differences except those due to their relative efficiencies. Other settings: South facing, 30 degrees inclination, shading – none or very little, 200 litre twin coil cylinder with 70 litres of solar dedicated volume. In T*Sol the daily demand was set to 100 l/day at 50C, Detached House (evening max) water use pattern. The panel performance characteristics shown in Table 1 were entered for each panel, but the aperture area was set to 2.5 m2 for both types. Other settings: 2 panels (for a total area of 5 m2), Location –London, South facing, 30 degrees inclination (1084 kWh/m2 panel irradiation), 200l twin-coil cylinder. Panel A Panel B SAP Qs (kWh/year) 1,135 1,058 T*Sol (kWh/year) 1,350 1,390 Table 2 Panel A produces 7% more solar energy in SAP Appendix H than Panel B. T*Sol, taking account of the second order efficiency coefficient predicts 3% more energy from Panel B. Table 2 shows the results. Ignoring the differences in total energy between the two methodologies and considering only the relative energy produced by each of the two panels, it can be seen that under SAP Panel A produces 7% more energy than Panel B. By taking the second order coefficient into account, in T*Sol Panel B produces 3% more energy than Panel A. By ignoring the second order coefficient, SAP produces an overall 10% inaccuracy in favour of Panel A. Suggested New Treatment in SAP 2009 A suggested refinement of SAP Appendix H is offered that would be simple to implement, but would remove the anomaly caused by solar panels with high second order coefficients. It is proposed that a new linear coefficient a1* is derived from the a1 and a2 test values. In this way the main body of the calculation can be left as it is, with a1* being used instead of a1. This reduces the testing required before making the change and lowers the risk of introducing unforeseen consequences. In preparation for the derivation of a1*, a representative value for GT is needed, as well as an understanding of the distribution of operating conditions for a solar panel. A study3 conducted in collaboration with the Building Research Establishment describes results from a test rig simulating an average hot water load on a hot water cylinder heated by solar and an auxiliary heat source. A water use pattern equivalent to 100 litres per day of hot water at 60C in a pattern according to EU M324EN tapping cycle 2 was applied to the hot water cylinder, and the heating system set up to simulate a “disinterested” or non-optimising household. Data from this study includes a record of irradiation in the plane of the panel, cylinder and panel temperatures and solar energy collection at three-minute intervals covering a 12 month period. If: η = η0 – a1.{(Tm – Ta)/ GT} – a2.GT.{ (Tm – Ta)/ GT}2 and for simplicity we set b = {(Tm – Ta)/ GT}, and choose GT = 600 W/m2, then: η = η0 – a1.b – 600a2.b2 [1] A straight line that passes through η0 has the form: η = η0 – a1′.b If a1′ = a1 + αa2 η = η0 – a1.b – αa2.b [2] then: [3] A best fit value of α was found by numerical methods such that the weighted difference between [1] and [3] was minimised over the range of operating points from 0.0 – 0.22. For panels A and B the value of α which achieved this was found to be 45. As a check a high performance tube was considered (η0 0.76, a1 2.12, a2 0.0077). A value of α equal to 45 also produced the best fit between the straight line and the curve. a1′ = a1 + 45 a2 [4] The energy collected in the year at different values of irradiation during this study is shown in Figure 3. Figure 5 A straight line approximation to the curve is found where the weighted difference between the two lines is minimised. Weighting values are shown in Figure 4 Figure 3 Annual energy collected by a solar water heating system at different levels of in-plane irradiation It was found that the panel collected around half the energy in the year at an in-plane irradiation below 600W/m2 and around half the energy at irradiation levels above 600W/m2. The representative value of GT used to derive a1* is therefore set to 600 W/m2. The operating point {(Tm – Ta)/ GT} was found for each three minute interval for the year, and the total number of minutes spent at each point summed. The results are shown in Figure 4. This new linear coefficient a1′ is always greater than the value of a1, so adopting this as the new coefficient for SAP would mean that the energy produced by all solar heating systems would be reduced. Consequently, a scaling value, s is found that reduces a1′ in such a way that the energy produced by a solar panel with an average performance is unaffected by the overall change. a1* = s.a1′ = s(a1 + 45 a2) [5] The performance characteristics from solar panels tested by SPF and published on their web site4 were analysed. The sample was randomly selected as the first 100 panels in the list, plus panels A and B. It included 23 evacuated tube collectors and 79 flat plate collectors, a proportion which is broadly representative of the breakdown of the solar market in the UK5. Figure 6 shows a frequency distribution for second order coefficients in the population considered. When equation [4] was used to calculate a1′ for each of the 100 panels, the average value of a1′/ a1 was found to be 1.121. To correct the average panel: s = 1/1.121 = 0.892, and a1* = 0.892 (a1 + 45 a2) Figure 4 Time spent at each operating point by a solar water heating panel during a one year monitoring study The profile in Figure 4 was used to create a weighting factor for each operating point and derive a straight-line approximation to the second order curve that minimised the weighted difference between the two curves. [6] Figure 6 Distribution of second order coefficients from test results of 102 solar panels. Effect of the Proposed Change The comparison calculations performed previously were repeated using the proposed new calculation with a1* in the Appendix H calculation instead of a1 For Panel A a1* = 0.892 (2.414 + 45 x 0.0490) = 4.12 For Panel B a1* = 0.892 (3.90+ 45 x 0.0125) = 3.98 The solar energy yield calculated by Appendix H is as follows: Panel A Panel B Proposed SAP Qs (kWh/year) 1,045 1,054 Panel A now produces 0.8% lower energy yield than Panel B and is therefore in better agreement with more sophisticated solar simulations that take the second order coefficient into account. Conclusion A simple modification to SAP Appendix H is recommended, where a new linear heat loss coefficient is derived: a1* = 0.892 (a1 + 45 a2) This would increase the accuracy of SAP solar energy estimates, particularly with respect to solar panels with a high second order heat loss coefficient. Although the number of these is small relative to the general population of panels, the significant benefits the current “loophole” offers means that they are increasingly favoured by SAP assessors and building designers who are often seeking only slight improvements in building performance to achieve compliance. 1 2 3 4 5 See http://www.bre.co.uk/sap2009 Panel A, Keymark number 011-7S516F, Panel B Keymark Number 011-7S652F See http://www.viridiansolar.co.uk/Assets/Files/BRE_Report_Viridian_Solar_Average_House_Simulation.pdf See http://www.solarenergy.ch/Collectors.111.0.html?&L=6 See solar thermal statistics compiled by HHIC. http://www.centralheating.co.uk/news/category/market-reports