FAST SIMULATION ALGORITHM FOR PV SELF
Transcrição
FAST SIMULATION ALGORITHM FOR PV SELF
Presented at the 29th European PV Solar Energy Conference, Amsterdam, September 2014 FAST SIMULATION ALGORITHM FOR PV SELF-CONSUMPTION WITH HEAT PUMP SYSTEMS: CALIBRATION METHODOLOGY WITH COMPREHENSIVE DYNAMIC SIMULATION MODEL AS REFERENCE Andreas Witzig1, Thomas Straub2, Matthias Hartmann2, Torsten Sonntag2 1 Vela Solaris AG, Winterthur, Switzerland 2 SMA Solar Technology AG, Niestetal, Germany Corresponding authors: Andreas Witzig, +41 55 220 71 01, [email protected] Thomas Straub, +49 561 9522 3125, [email protected] ABSTRACT: Self-sufficiency of systems with photovoltaic (PV) energy production gets more and more important. Demand side management with the use of heat pump and thermal energy storage has proven to be an efficient and effective means for increasing the PV self-consumption ratio [1, 2]. Accurate prediction of the self-consumption and the resulting self-sufficiency quota requires a comprehensive simulation including heat pump characteristics, thermal balance of the building and hot water demand profiles. These simulations are usually elaborate and time consuming. This work presents a fast but reliable estimate of the system’s PV self-consumption potential. It is based on the wellknown tool Sunny Design [3], which has been extended by the capability to calculate the building and hot water energy demand, and applying heat pump model with realistic characteristic curves. The proposed algorithm is capable of including battery storage and advanced controllers as future extensions. The algorithm is in accordance with the physics-based simulation tool Polysun which includes device physics as well as controller details with a tight coupling between the electrical and the thermal part of the system [4]. The newly introduced model parameters in Sunny Design are calibrated and validated with the detailed Polysun model. Keywords: PV self-consumption, load management, heat pump, simulation, Sunny Design, Polysun. 1. prerequisite of any photovoltaic simulation and in general, it can be recognized that the time resolution has to be at least one hour but accuracy still increases with a smaller time step. Sunny Design applies a five minute time step and Polysun has a variable time step down to a few seconds. The strength of Polysun is that in addition to the electrical calculation, it analyses the hydraulic topology on the thermal system and calculates the fluid flow in all pipes [5], its application for modern energy efficient buildings [6] and combined photovoltaic and thermal systems [7]. The simulation model therefore also includes pumps, three-way valves, heat exchangers and thermal storage tanks with stratification (as displayed in Figure 1) as well as the detailed control algorithms in the system. Sunny Design, the dimensioning and PV yield prediction tool of SMA Solar Technology AG, has been extended with a building and hot water heat demand model and a heat pump model in the course of this work. A fast CPU execution time is crucial for this application. In the course of the presented joint work, a MATLAB script has been developed which is translated into Sunny Design. The intermediate simulation results presented in this work with the “fast algorithm” are typically shown out of MATLAB and are not available in the graphical user interface of Sunny Design. INTRODUCTION 1.1. Motivation PV self-consumption optimization recently received a lot of attention. In many cases demand side management is more effective and cheaper than the usage of battery systems or is used in combination with electrical storage. In particular, the combination of heat pump heating systems with PV rooftop installations is promising and potentially offering a relatively high self-sufficiency quota. It is important to recognize that the coupling between electrical and thermal systems has the advantage to utilizing already existing thermal storage elements (hot water storage tank or thermal inertia of buildings), which typically have a time constant of about one day and experiencing no deterioration also with high number of charging and discharging cycles. Simulation of PV self-consumption with heat pump systems as a part of the planning process today exceeds the services a typical PV system specialist offers and has up to now mainly been done in research institutes and in the academic world. This work contributes by simplifying the planning process and providing ready-touse simulation tools for several stages in the planning process. Namely, the planning software Sunny Design and the comprehensive simulation tool Polysun offer new functionality that facilitates PV self-consumption calculation and optimization. A good agreement between the tools increases planning certainty and result reliability. 1.3. Scientific results This work describes a general procedure to bring simulation models of different detail-levels in agreement and is effective for many other algorithms beyond the specific above mentioned commercial applications. A calibration methodology is presented that allows implementing a very fast yet accurate and easy to use algorithm, based on a more detailed comprehensive 1.2. Starting point Both simulation models are based on a time-stepping algorithm and comprise the yield calculation of photovoltaic modules and inverters. This is an important 1 Presented at the 29th European PV Solar Energy Conference, Amsterdam, September 2014 simulation model. The advantage of using models on several detail-levels is that model parameters in the fast model can be determined by calibration against the comprehensive model. The fast model can be reduced to a very small set of parameters, which makes it very lean and easy to use. An important part of the work lies in the definition of a large number of system details in the comprehensive model. These details are all comprised in the small parameter set of the fast model resulting by the calibration procedure. Publishing the comprehensive model makes the procedure transparent and implicitly lists all the model assumptions based on physical rationale. In de-centralized energy systems, controller settings are key influencing variables and it is important to model them in sufficient accuracy also in the fast model. In consequence, innovative components or controller strategies can easily be accounted for. A concise set of default values is proposed, built into the software tools and discussed in this paper. User can make a quick simulation easily and if required can go into more detail if necessary. systems with heat pump in comparison to the system without heat pump. 2. SIMULATION ALGORITHMS 2.1. PV Self-consumption calculation For the owner of a PV power plant the most important figure is the total energy production per year. Secondly, he will relate the energy production to the energy consumption and he may typically recognize that in summer he is a net energy producer while in winter he is a net energy consumer. Figure 3: Example results for the electricity production and consumption in a typical plus-energy house with a roof-top PV installation. (3) (1) In the view of the house owner, it may seem natural to look at the summation of the energy balance over year or over a month. However, since the power grid is more and more in a focus with a large portion of photovoltaic energy production in summer, it is important to consider the power to and from the power grid at any instant in time. As a general rule, the self-consumption ratio is the highest number if yearly sums are divided (as in the definition of a “plus-energy house”). It becomes smaller if the energy balance is calculated on a monthly basis and even smaller if hourly values are considered. In reality, instant power values are relevant, and in many applications, the resulting self-consumption is considerably smaller than expected. In consequence, detailed electrical consumption profiles are a key input parameter for any simulation in the planning phase of a PV system. Within this study we used a measured domestic electricity consumption profile published in reference [8]. This profile for medium energy consumption has a time resolution of 5 minutes. (2) (5) (7) (6) (4) Figure 1: Polysun simulation used in this work. The system is also available as a company template in the release version of the Polysun software. Components: (1) PV field, (2) inverter, (3) electrical consumption profile, (4) heat pump, (5) floor heating with building model, (6) storage tank, (7) hot water tap. Energy yield 8,707 kWh Self-consumption 2,914 kWh 2.2. Building model For the evaluation of the heat pump power consumption profile, it is required to calculate the building heat demand. The building is modeled with a single node dynamic model where the thermal losses through windows and walls play the major role, followed by ventilation and infiltration losses. Solar thermal gain through the windows and thermal gain from people and equipment inside the building are also accounted for. Note that passive solar gains are considerable, especially for modern well-insulated buildings. They have to be included in order to model the thermal balance of the building accurately. The models for solar thermal gains differ significantly between the fast and the comprehensive algorithm. In the fast algorithm, the passive solar heat gain 𝐺psh of the building is proportional to the global irradiation 𝐺(𝑡) from the weather data. Grid feed-in 5,793 kWh Consumption 9,121 kWh Figure 2: Sunny Design simulation results. Selfsufficiency and self-consumption quotas are calculated and displayed together with a graphical representation of the yearly energy balance. The tool always shows the 2 Presented at the 29th European PV Solar Energy Conference, Amsterdam, September 2014 𝐺psh(𝑡) = 𝐴(𝑡) ∙ 𝑢 ∙ 𝑊𝑊𝑅 ∙ 𝑆𝐻𝐺𝐶 ∙ 𝑆irr ∙ 𝐺(𝑡) with the window transmittance 𝑢, the window-to-wallratio 𝑊𝑊𝑅, the solar heat gain coefficient 𝑆𝐻𝐺𝐶 and 𝑆irr being a scaling factor used for calibration (see Section 3.2). In the comprehensive simulation, the solar gain is calculated by evaluating the solar irradiation on every wall of the building according to the angle of incidence. At a lower position of the sun in winter, there is automatically more solar irradiation in the comprehensive model. In the simplified model, the excess passive solar gain in winter is accounted for by the formula Figure 6: Building heat demand, monthly average, comparison between Polysun and Sunny Design. The differences during the months are due to the difference of the passive solar gain. The irradiation through the windows is calculated geometrically in Polysun and is approximated with a lumped model in the fast algorithm. 𝐴(𝑡) = 1/(1 − 𝑎 ∙ cos(2π (t + 10 days)) with the angle modifier factor a as a fit parameter and a phase shift of 10 days between the winter solstice on December 21st and the start of the simulation period on January 1st. The resulting agreement between the solar gain calculated by the fast and the comprehensive algorithm is shown in Figure 4. As it can be seen in the Figures 5 and 6, the passive solar gain provides sufficient energy to the building for more than four months of the year. In the specific example, no heat demand occurs for mean outside temperatures above 10°C. The fast algorithm assumes that the building heat demand is always met but that the heating system only switches on as soon as a certain energy deficit is accumulated. The level of energy deficit that is allowed is a measure for the thermal inertia of the building. In a similar manner as for room heating, the heat demand of the hot water production is calculated: the accumulated energy deficit is treated separately due to its different temperature level. It is assumed that the hot water demand has always a priority over the building heat demand. The comprehensive model applies a plug-flow technique and accounts for the specific hydraulic topology of the system of choice. It includes pipe details and accounts for thermal losses through the pipes. Furthermore, it assumes that thermal storage tanks are used and – next to the thermal losses of the storage tank – simulates natural stratification and mixing of fluids with different temperature levels. Figure 4: Passive solar gain is the main energy input for a well-insulated modern building. In winter, it covers most of the building heat demand and in summer, additional ventilation losses are required (people open the windows) in order to prevent overheating. 2.3. Heat pump model In the system under consideration, an air-water heat pump is applied as the only heating system installed to cover the building heat demand (monovalent system). The heat power delivered to the building from the heat pump is therefore required to cover the heat demand of the building also during the coldest time of the year. It is assumed that a low temperature heating system (underfloor heating) is applied. An inflow temperature of 35°C is assumed for the heating system. Furthermore, it is assumed that the same heat pump is used to provide the heat for the hot water. A heating temperature of 55°C is assumed. The heat pump efficiency strongly depends on the temperature difference between the primary side of the heat pump (outside temperature) and the secondary side (temperature level of the floor heating or the hot water, respectively). Note that the assumed temperature levels for the required heat are relatively low, which is in favor of the heat pump system and are the state-of-the-art in modern energy-efficient buildings. In the fast algorithm, built-in characteristic curves for a typical heat pump are applied according to Table I and II. The different efficiencies for the higher temperature (hot water production) and the lower temperature level (heating of the building) are reflected in these curves. Figure 5: Building heat demand, daily average, comparison between Polysun and Sunny Design. It can be seen that the Polysun calculation shows higher peaks due to a more detailed building model. 3 Presented at the 29th European PV Solar Energy Conference, Amsterdam, September 2014 Table I: Characteristic curve for heat pump when used for heating (assuming floor heating at 35°C) Tamb °C Pheat kW Pel kW -20.0 -15.0 4.5 2.1 4.9 2.1 -7.0 6.4 2.2 2.0 7.0 10.0 20.0 8.5 10.7 11.8 12.0 2.3 2.4 2.4 2.4 Table II: Characteristic curve for heat pump when used for hot water production (hot water temperature at 55°C) Tamb °C Pheat kW Pel kW -20.0 -15.0 3.2 3.1 4.4 3.1 -7.0 2.0 7.0 10.0 20.0 6.2 3.1 7.5 3.3 9.7 10.5 11.0 3.4 3.5 3.6 Figure 7: Hot water profile, summation over each day of the year. The detailed profiles also show a realistic intraday time variation, which is not visible in this figure. 2.6. Weather data For the calculations in Sunny Design, the characteristic curves are taken from a 20kW heat pump. Pel and Pheat are the electrical power into the heat pump and the heat power at the output of the heat pump, respectively. For heat pump sizes other than the 20kW nominal power, Pel and Pheat are scaled accordingly. In Polysun, heat pumps are characterized with more supporting points, typically originating from the standards EN 255 and EN 14511. If available, more measurement points can be entered in Polysun. In all the available points from the heat pump characteristics, Polysun interpolates with the scattered data interpolation algorithm [9]. Choosing the correct size of the heat pump is in reality crucial because a too big heat pump reduces the efficiency of the system. However, a too small heat pump that is not capable of covering the requirements in the coldest time of the year is a severe malfunction resulting in complaints and expensive service work. In the simulation, the thermal power from the heat pump is adjusted such that the heat pump runs continuously during the coldest 48 hours of the year. Furthermore, an oversizing factor is introduced which takes into account that the weather data represents a typical year and the heat pump system is also required to cover the building heat demand for an extremely cold winter. In consequence, the heat pump size is automatically chosen such that the heat pump is driven in pulse operation, switching on and off several times per day. As a prerequisite, weather data for specific locations are required, including ambient temperature and solar irradiation. Both the fast and the comprehensive algorithm use marching-on-in-time algorithms which span over an entire year in order to reflect the different seasons as well as the natural fluctuations of the weather. Therefore, weather data based on monthly or daily basis is not sufficient as an input, but has to be provided at least in arrays of every hour per year. In any calibration activity or when comparing different simulation tools with one another, it is absolutely necessary to have exactly the same weather data as an input. For the calculations presented in this work, a weather data set for Kassel, Germany, was used which had been generated using the software Meteonorm [10]. 2.7. The art of choosing good default parameters It is important to request the right information from the user of the simplified software and to choose adequate default parameters for the information that is not available in the typical situation when the software is applied. In the case of Sunny Design, it is feasible (and necessary) to ask about the buildings size and insulation standard. However, it is not an option that the user has to care about the window areas and glass type since this is in general unknown to the PV planner. In Polysun on the other hand, more details about the building can be entered, as for example the window-to-wall-ratio for the different directions south, east, north and west and the solar energy transmittance coefficient for the window glasses. Default values are chosen according to well-known standards, if they are available. As an example, heat transmission coefficients according to Table III have been chosen for the five different building types that are offered in Sunny Design. 2.4. Heating controller The heating system controller that drives this operation is modeled in the comprehensive simulation with Polysun as well as in the fast simulation with Sunny Design because for future use of this algorithm, switching on and off the heat pump needs to be modeled in much detail, carrying a high optimization potential. Table III: Different building types offered in Sunny Design. 2.5. Hot water profile Building type The hot water energy demand shows two daily peaks, one in the morning and one in the evening. At weekends, it is assumed that 14% more water is used, and a lower water consumption in summer is assumed, as it is shown in Figure 7. overall heat transmission coefficient Passive house New building, well insulated Average residential building (2010) Average residential building (2000) Building not modernized (<1995) 4 0.17 W/K/m2 0.24 W/K/m2 0.34 W/K/m2 0.50 W/K/m2 0.80 W/K/m2 Presented at the 29th European PV Solar Energy Conference, Amsterdam, September 2014 Secondly, the two variables “irradiance scaling” 𝑆ir and “building shell scaling” 𝑆bs are used to obtain the agreement in the heat demand from the building. While 𝑆ir appears in the formulas as a pre-factor to the global irradiance from the weather data, the second parameter 𝑆bs scales the building shell heat loss and appears in the product with the building shell area and the heat transmittance. A model agreement has to be found for different building types and different locations, i.e. for various building heat demands. In an exemplary manner, Figure 8 shows how a variation of the desired indoor room temperature and the effect of different fit parameters influence the agreement between the comprehensive and the fast algorithms. In a next step, it has to be ensured that the dynamic behaviors of the fast and the comprehensive algorithms are in agreement. The time constant of the building 𝜏building represents the thermal inertia. Furthermore, the fast algorithm has two controller parameters which are the trigger points to switch on the heat pump for heating heating hotwater or for hot water production, 𝜗trigger and 𝜗trigger , respectively. Table IV: Main default parameters Parameter Default value Building floor height Internal heat gain in the building Air change (causing ventilation heat loss) Desired room temperature Hot water temperature Nominal inlet temperature floor heating Cold water temperature Solar heat gain coefficient Solar energy transmittance through windows 2.5 m 1.5 W/m2 0.3 1/h 21 °C 50 °C 35 °C 10 °C 0.85 0.5 Simulation projects done with the comprehensive model Polysun are provided in order to make the choice of default parameters transparent. Having the comprehensive model at hand, it is not only possible to obtain these values but also to recognize the influence they have on the simulation. 3. MODEL CALIBRATION AND VALIDATION 3.1. Consistency of parameters and input data In the comparison between the fast and the comprehensive simulation algorithms, first it has to be ensured that the same input data is used for the weather data, the hot water and electrical consumption profiles, the characteristic curve of the heat pump. A selection of building model parameters from Polysun is offered in Sunny Design, which is summarized in Table IV. In the fast algorithm, several parameters, such as the energy storage capacity of the building and the thresholds for switching on the heat pump due to an accumulated building heat demand have to be determined. In the calibration procedure, the above parameters are chosen such that the heat pump runtime, the number of switching on, and the total thermal energy from the heat pump match between the fast simulation and the comprehensive simulation. The goal is that monthly values match well in order to make sure that seasonal effects are well covered. Figure 8: Calibration of the fast algorithm. The building heat demand depends on the desired indoor room temperature. The pairs 𝑆ir and 𝑆bs have to be chosen such that an optimal fit is found in comparison to the Polysun reference. “Fit-2” shows good agreement with the scalar parameters 𝑆ir=2.8 and 𝑆bs =2.1. 3.3. Results comparison The simplifications in the fast simulation algorithm introduce some model inaccuracy. However, compared to the uncertainties in the input data (hot water consumption, weather data), the agreement between the fast algorithm and the comprehensive model is still good (see Table V and Figure 9). Furthermore, examining the time domain simulation results it is clear that the heat pump is in pulsed operation most of the time, even in winter (Figure 10). In order to obtain a good selfconsumption ratio, the PV field has to be large enough in order to cover the power consumption requirement from the heat pump. The potential to increase the selfconsumption ratio by a smart controller depends on the thermal storage capacity of the heating system. 3.2. Calibration parameters The agreement between two different simulation models can be recognized as a non-linear optimization problem. While the comprehensive model is based on physical formulae and detailed information about the system (such as pipe diameter and length, storage tank dimensions and insulation properties), the fast model has some built-in fit parameters. The calibration of these parameters is an important procedure in the model development and it is important for the acceptance of the new fast algorithm that the calibration procedure can be reproduced. In a strictly methodical approach, a scalar measure is defined to evaluate the agreement between the two numerical models. However, different from a rigorous numerical optimization problem, engineering know-how about the system behavior is also used in the fitting procedure, bringing not only some final results in agreement but also some intermediate variables: A first variable “hot water scaling” 𝑆hw is used to force an agreement in the hot water energy demand. Table V: Result comparison between Polysun and Sunny Design for the building type “New building, well insulated”, 220 m2 heated floor area and an annual hot water demand of 146 m3 (8 people). Variable Unit Polysun Sunny Design Heating demand Hot water demand 5 kWh/a kWh/a 11126.6 6998.2 11208.0 7133.4 Presented at the 29th European PV Solar Energy Conference, Amsterdam, September 2014 New trends include the use of modulating heat pumps for a better match to the PV power production [9] and the application of smart control strategies [11]. The presented algorithm is capable of covering such systems. While the comprehensive simulation model covers these topics, the current version of Sunny Design is optimized for simple user interaction and does not yet comprise this level of complexity. In many regions of the world heat pumps are more often used for cooling than for heating, providing a huge potential for PV self-consumption operation. The presented algorithms are well capable to cover this application range and a very high potential for efficiency optimization is recognized in this field. Figure 9: Result comparison between Polysun and Sunny Design show good agreement for the heat pump operation time. 5. The authors want to thank E. Vrettos and Dr. S. Koch, (ETH Zurich) and Prof. Dr. Zogg, (University of Applied Sciences and Arts Northwestern Switzerland) for fruitful discussion and advice. The work has partially been financed by Swiss Electric Research. (1) (2) 6. [1] REFERENCES P. Trichakis, P.C. Taylor, P.F. Lyons, and R. Hair, Predicting the technical impacts of high levels of smallscale embedded generators on low-voltage networks, IET Renewable Power Generation, Vol. 2, Issue 4, pages 249-262, 2008. [2] M. Castillo-Cagigal, et al., PV self-consumption optimization with storage and Active DSM for the residential sector, Solar Energy 85.9 (2011): 23382348. [3] www.sunnydesignweb.com, www.sma.de/SunnyDesign [4] www.polysunsoftware.com, www.velasolaris.com [5] S. Mathez, Polysun 4: Simulation of systems with complex hydraulics. In the proceedings of the 17 th OTTI Symposium “Thermische Solarenergie”, Bad Staffelstein, Germany, May 2007. [6] A. Witzig, S. Geisshüsler, P. Brönner, E. Kaminsky, Planung von Plus-Energie Häusern mit Polysun, in the proceedings of the 2nd OTTI symposium „AktivSolarhaus“, Lucerne, Switzerland, September 2010. [7] E. Vrettos, A. Witzig, R. Kurmann, S. Koch, G. Andersson, Maximizing local PV utilization using small-scale batteries and flexible thermal loads. In the proceedings of the 28th European PV Solar Conference and Exhibition (PVSEC), Paris, France, October 2013. [8] Dr Ian Knight, Nico Kreutzer: Three European Domestic Electrical Consumption Profiles –IEA / ECBCS Annex 42 FC+COGEN-SIM, The Simulation of Building-Integrated Fuel Cell and Other Cogeneration Systems, July 2006. [9] L. Fierz, A. Witzig, L. Gasser, B. Wellig, Simulation modulierender Wärmepumpen, Schlussbericht Bundesamt für Energie (BfE), Bern, Schweiz, Dezember 2013. [10] www.meteonorm.com [11] E. Vrettos, K. Lai, F. Oldewurtel, and G. Andersson, Predictive control of buildings for demand response with dynamic day-ahead and real-time prices, European Control Conference (ECC), Zürich, Switzerland, July 2013. Figure 10: Electrical input power of the heat pump (blue line, left axis), solar gain (green line, left axis), and ambient air temperature at the air-water heat pump (red line, right axis). Note that the heat pump runs in different states depending on if it is producing heat for hot water (1) or heating for the building (2). The plot shows the operation in the end of January and it can be seen that the heat pump operates continuously at night for several hours but is in pulsed operation most of the time. CPU time is a critical measure since the fast simulation algorithm is offered on the internet and the typical simulation times of a comprehensive simulation like Polysun (18 seconds on a 2.2 GHz i7 CPU with Windows 7) is already too long. The Sunny Design code evaluates in 90 miliseconds and the speed-up therefore is a factor 200. 4. ACKNOWLEDGEMENT CONCLUSION AND OUTLOOK PV self-consumption in combination with heat pumps is a good way to reduce the load on the distribution power grid. However the pulsed operation of the heat pumps in reality naturally puts an upper limit to the selfconsumption ratio. In order to obtain a good match between the PV installation and the heat pump, the entire system dynamics has to be considered. They need to match both in the power levels as well as in the energy balance. There are many approaches to PV self-consumption calculation and this work aims to give some indication to rate these calculations. In order to propose a quick algorithm, one might be tempted to use a one hour time step or even an algorithm that calculates selfconsumption on a daily or monthly basis. However, it has been found that this is not accurate enough and at least a five minute time step should be used, otherwise the electrical consumption profiles and the pulsed operation of the heat pump cannot be modeled adequately. 6