Voltage control in low voltage networks by Photovoltaic Inverters PVNET.dk

Transcription

1 MAKING MODERN LIVING POSSIBLE Voltage control in low voltage networks by Photovoltaic Inverters NET.dk

2 A consequence of the price reduction of solar photovoltaic () systems is that the installed capacity will also increase in the years to come. Thus, more distributed generation will be installed in the public low voltage distribution network (LV network). This results in a paradigm shift in the operation of the LV networks, since the power flow becomes bidirectional. Questions about the operation of the LV networks are therefore being asked, in order to prepare for the future increase in solar capacity. Danfoss Solar Inverters, Technical University of Denmark (DTU) and the two DNOs EnergiMidt and Østkraft have taken the opportunity, to answer some of these questions, in order to facilitate a smooth grid integration of solar systems into the LV network on a large-scale. This is done through various research and demonstration projects and by participating in standardization work and similar activities. Solar generation in the Low Voltage Network Several technical challenges may appear when increasing the capacity of solar in the LV network: Overvoltage at the end user Overloading the infrastructure (distribution transformer and cables) Voltage unbalance Voltage harmonics and flicker Protection system failure Based on a survey among Danish distribution network operators (DNOs), the most urgent issues are overvoltage, overloading of the infrastructure and voltage unbalance. The scope of this investigation is to analyze the two first and most important subjects as they are highly inter-dependent. The 3rd and 4th items on the urgency scale has been studied by Danfoss Solar Inverters in reference [1] and [10] and is not covered here. A comprehensive analysis conducted by the Danish energy association, Dansk Energi, including more than 1100 LV feeders, shows that around 0.4% of all the investigated feeders will experience overvoltage when the solar capacity is around 0.7 kw per user on all the feeders [2]. The amount of 0.7 kw per user corresponds to a total capacity of 3500 MW, which is expected to be installed in Denmark by year 2030 [3]. Doubling this amount of solar to a total of 7000 MW will result in approximately 0.6% of the feeders having problems with overvoltage [2]. A solar capacity of at least 5000 MW is regarded as an optimum in the future energy-mix [4], [5] and 7000 MW can cover around 20% of the yearly Danish electrical energy demand and around 100% of the peak-load. Roughly speaking, there are approximately distribution transformers in Denmark, thus installing 7000 MW solar capacity corresponds to installing 100 kw per distribution transformer, which is also the typical smallest distribution transformer used in Denmark. Some LV networks will of course reach a higher amount of installed solar capacity, thus the probability of having this type of problems will increase in local feeders. Several solutions have been suggested until now in order to cope with the overvoltage phenomena at high solar penetration in the LV network: Voltage control using reactive power generation from inverters Curtailment of active power from the solar systems Network upgrade of cables and distribution transformers Battery storage and energy buffer at solar systems Voltage control at the secondary side of the distribution transformer by on-load tap changers (Seasonal) changes of the tap position of the distribution transformer Energy management systems / load controls at users Voltage control through reactive power injection from inverter is one of the easiest to implement in the LV network because of the versatility of the inverter [6]. Today s inverters have the following methods included for voltage control: constant power factor (PF), constant reactive power, Q(U) control and PF(P) control, more about this later. Curtailing the active power from the solar systems seems a very easy way of mitigating the overvoltage and has been investigated by researchers and DNOs [7]. However, the owner of the solar system cannot estimate the impact of this control scheme upon the economic aspects of the investment and should therefore be avoided. An interesting conclusion is: At first glance it seems that local or central regulation of reactive power comes first among the possible strategies. Active curtailment would then be activated when reactive compensation is no longer sufficient to avoid upper voltage constraints [7]. Danfoss Solar Inverters recommends the same approach. The typical approach applied prior to inverters contributing to the voltage control involved increasing the grid capacity by upgrading the distribution transformers to a larger power rating or by reinforcing the LV feeders by addition of parallel lines or replacement of old lines with higher capacity ones. The last listed methods in are not covered here. According to the study Connecting the Sun conducted by the European Photovoltaic Industry Association (EPIA) [8] and research done at Fraunhofer [9], the cost of the described approaches, when applied to a typical LV network (250 kva distribution transformer with 122 users and 180 kw solar capacity) are: Curtailing the active power to 70% of installed capacity: 3600 /year Reinforcing the network: 3000 /year Reactive power injection: 1500 /year considering the maximum tolerated loading of the transformer and cables is 150% of the rated power and also supposing the voltage at the distribution box (PCC) is kept below 1.03 p.u. when a new solar system is installed [9].

3 Methods Exceeding the prescribed 10 minutes average voltage limits defined by EN50160 and/or overloading the LV network components are two possible scenarios when installing many solar systems in a network. To analyze these phenomena, computer simulations have been performed using a commonly accepted power system simulation tool: Power Factory DIgSILENT. A gradually increasing amount of solar capacity has been integrated in a model for a typical distribution network and its influence over the voltage magnitude and the loading of components has been observed using one year hourly samples of consumption and solar power generation. The simulation model includes only three-phase symmetrical consumers/ generators thus removing the possibility of voltage unbalance. A simple inverter model is implemented by using an RMS three-phase AC current source and two voltage control algorithms. Regarding the orientation of the solar systems, it is assumed that all systems are oriented south with a 45 inclination, see Figure 1. This is considered to be the worst case scenario since during sunny days at noon the highest solar power production will be experienced with south oriented installations. West East The hosting capacity of a LV network is defined by the amount of solar power, which can be installed in the LV network before certain limits are reached. The limits considered in this investigation are the overloading of the LV cable-sections, the overloading of the distribution transformer and the overvoltage at the outermost distribution box (PCC), for each of the simulated 8760 hours. The loading of the LV cable-sections is evaluated by comparing the current in each cable-section with its nominal value. Hence, 100% loading of the cables is equivalent to one specific cable-section being operated at its nominal current. The loading of the transformer is evaluated in a similar manner. The overvoltage is evaluated by reading the hourly maximum voltage recorded at the distribution boxes (PCC). The hosting capacity of the LV network is reached when any bus voltage within the network is exceeding the ±10% voltage criteria from the EN50160 standard. Two types of standard Volt-VAR control schemes are investigated: the power factor depending on active power output of the inverters, PF(P), and reactive power depending on the terminal voltage of the inverter, Q(U), as defined in Figure 3. 1 PF* South Figure 1: Physical orientation of the residential solar systems. Figure 2 shows that by distributing the solar systems in orientation and inclination the hosting capacity can be further increased with more than 10% compared with solar systems only pointing south and with 45 inclination for Inom 16 A/phase 0.90 for Inom > 16 A/phase 0.5 P inv [pu] 1.0 Power factor by active power control as defined in TF based on VDE AR N Distributed orientation Common orientation Q*[pu] Power[W] U m inv [pu] Time [h] Figure 2: Clear sky average hourly power generation for Common/ distributed oriented systems. Two generic Q(U) curves, including the definition of m = ΔQ/ ΔU. Figure 3: Standard voltage control methods for modern inverters, PF(P) and Q(U).

4 Slack bus Line1 Line2 Line3 10 kv P,Q 0.4 kv Aggregated distribution networks 4x 6x 4x 5x 4x 4x Figure 4: Generic distribution network with solar generation at each user. The ring connection is open. The indicated number of inverters in the boxes means a lumped inverter model. A generic LV grid model developed at DTU, with 71 users and a 100 kva distribution transformer is used in the simulations, being considered as representative for most LV networks in the area of ØSTKRAFT Net A/S, see Figure 4.

5 All users are equipped with solar systems, of equal size and orientation. Furthermore, a simple MV network is implemented in order to observe the voltage variations in the 10 kv network as well. The energy consumption for each of the 71 users is based on time-series containing 8760 hourly values for a year of generic consumption, see Figure 5. The solar generation is based on synthesized hourly irradiance by the syst software, taking both clear sky and covered sky into consideration. All loads and inverters are assumed being connected through three phases. The PF for the loads is being held constant at 0.95, inductive. Trees Don t Grow to the Sky The following results are for the worst case scenario where all solar systems are facing south with 45 inclination, as previously mentioned. The hosting capacity of the LV networks can be increased by additional 10-15% if the solar systems are evenly distributed in orientation as in Figure 1. All results are summarized in table II. The results presented here are for one LV network only, but we still believe in the general result. The results in Figure 6 show that without voltage control the overvoltage phenomena starts for a solar capacity of 1.5 kw per user (total 107 kw). By applying a standard PF(P) control scheme with PF equal to 0.95 at nominal power, the overvoltage condition is avoided up to a solar penetration level of 1.8 kw per user. By applying Q(U) control, the overvoltage condition is mitigated up to 2.0 kw per user (total 142 kw). The overvoltage issue is not solved by upgrading the distribution transformer on the contrary increasing the size of the distribution transformer has a slight negative effect. The loading of the distribution transformer is documented in Table I. More than 140 kw solar power can be installed on the 100 kva distribution transformer before it becomes overloaded. This is in agreement with the 150% hosting capacity in [9]. Normalized electrical power consumption for a typical Danish household. A typical Danish household consumes 3.44 MWh per year. Table I: Frequency of loading of the 100 kva distribution transformer when no voltage control is applied and increasing penetration of solar capacity. Solar Number of hours for a loading level at: penetration [kw] 100% 110% 120% 130% Specific energy production for a 1 kw solar system for the town of Brædstrup, DENMARK. A typical solar system produces around 900 kwh per installed kw. Figure 5: Yearly consumption (top) and solar production (bottom) for a typical household and solar system. X-axis is time of day [h], Y-axis is time of year [month] and z-axis is average power [kwh/h]. A solar penetration of maximum 178 kw is possible without overloading the LV cables, when no voltage control is applied. Applying voltage control will decrease the hosting capacity of the LV cables due to the additional reactive current. Figure 7 shows the yearly losses in the LV cables and distribution transformer, when the solar capacity is increased from 0 kw to 178 kw. The losses are seen to drop with approximately 700 kwh when installing 36 kw to 71 kw solar, no matter of how the voltage is controlled, or not. Up to 107 kw can be installed without generating more losses compared with the no solar case. When exceeding the 107 kw solar capacity, the losses for the Q(U) voltage control methods start to increase faster compared with the PF(P) case. This is because the amount of reactive power is required by the grid voltage and the Q(U) curve.

6 60 / 10 kv transformer Yearly min/max voltage for different cases 10 / 0.4 kv transformer Bus 1 Bus 2 MV_T LV_T Outermost distribution box Load 1.1 Voltage (pu) 2.5% Voltage Increase on the LV lines in the 0.4 kv net Tap position +2.5% 2.5% Voltage Increase lines in the MV lines 1 (d) (c) (b) (a) Voltage across the 10 kv network: 5% (d) (c) (b) (a) 5% Voltage drop on the LV lines in the 0.4 kv net (d) (c) (b) (a) 0.9 (d) (c) (b) (a) Installed : 178 kw Installed : 142 kw Installed : 107 kw Installed : 71 kw Installed : 36 kw Installed : 0 kw Minimum Voltage Limit Maximum Voltage Limit Figure 6: Comparison of voltage control methods. Base case is (a) without voltage control. In case (b) the standard PF(P) is applied and in case (C) the Q(U) is applied, with m = 4. The Q(U) control is also applied in the night, thus increasing the minimum voltage on the feeder. In case (d) the distribution transformer is upgraded to 160 kva but without applying voltage control. Power Losses [MWh] Total active power lossess in transformer and cables (a) Trafo 100 kva. No Volt Var Control (b) Trafo 100 kva. PF(P) (c) Trafo 100 kva. Q(U) (d) Trafo 160 kva. No Volt Var Control Total installed in the network [kw] Figure 7: Energy losses for the four cases, when increasing the solar capacity. Minimum losses of 5.3 MWh per year are reached for a total solar capacity of 71 kw, but up to 107 kw can be installed and still keep the losses below the case with no solar.

7 Table II: Comparison of hosting capacity for the four investigated cases. For the loading of the distribution transformer: a range is given when the sum of overloading hours is higher than 86 hours per year and an approximate value is given when the sum is below 86 hours per year. For the scenario with no solar penetration, the total yearly losses in the network including transformer are 6.0 MWh and the maximum hourly reactive power exchange through the transformer is 15 kvarh, inductive. Case Base case with 100 kva trafo 160 kva trafo, no control PF(P), minimum PF = 0.95 Q(U), low sensibility Hosting capacity [kw] Overvoltage Transformer loading Total yearly energy loss at full penetration [MWh] Maximum hourly reactive power exchange through transformer [kvarh] (ind) 107 n.a (ind) 124 ~ (ind) (ind) Scaling up the result in Figure 7 and Table II to the Danish situation with around distribution transformers, the hosting capacity of the LV network, without reinforcing, is in the range 7500 MW 8700 MW solar. By installing 5000 MW solar capacity, the yearly network losses can be reduced with around 50 GWh, corresponding to the consumption of approximately households. Recommendations for voltage control in LV feeders Based on these results, if overvoltage is observed at the user s site the following order of actions may be applied: PF(P) for all inverters on the feeder Q(U) for all inverters on the feeder Increase self-consumption at peak production hours, either by timing the starting hour of local appliances or by using the trigger-signal which some inverters offer Lower active output power of inverters (only in emergency cases for short periods), can be done through the set-up menus, or the P(U) function which some inverters have included Upgrade LV cables, upgrading the distribution transformer might not help Install energy storage devices For Danfoss Solar Inverters of the type DLX (single phase connection) and TLX (three phase connection) the PF(P) voltage control can be activated by selecting one of the following country-codes: Inverter / country-code DLX TLX + / TLX PRO+ EN50438-DK n.a. No voltage control LV1 / Danmark 16A PF(P) with minimum PF = 0.95 LV2 / Danmark > 16A PF(P) with minimum PF = 0.90 LV3 n.a. No voltage control (constant PF = 1.0) If required, the TLX PRO+ can also be configured to Q(U) voltage control, contact Danfoss Solar Inverters at: inverter-application@danfoss.com for more information. Recap A typical distribution transformer can host up to 140% - 150% solar capacity, without suffering from overloading. The hosting capacity of LV networks can be increased with additional 10% - 15% if the solar systems are evenly distributed in orientation. The total hosting capacity for the distribution transformers in Denmark is in the range 7500 MW 8700 MW solar, without reinforcing the LV networks. Thus, a large-scale grid integration of solar systems into the energy system should not be a major problem for the DNOs. However, the transmission system operator, in this case Energinet.dk, will have to include the multi MW solar into their planning and make use of the available primary frequency support and fault ride through capabilities, etc., in modern inverters. The losses in the LV network can be reduced with 10-15% by installing 40-70% solar capacity compared with the nominal size of the distribution transformer. Applying this particular finding to the distribution transformers in Denmark, the yearly savings would correspond to the energy consumption of around households. Overall, it has been found that applying voltage control in the LV networks has a beneficial influence in respect to an increase of the solar penetration. Both PF(P) and Q(U) control types improve the voltage profile at the outermost distribution box (PCC).

8 References [1] R. D. Lazar and A. Constantin, Voltage Balancing in LV Residential Networks by Means of Three Phase Inverters, in proc. European Photovoltaic Solar Energy Conference EUSEC, September [2] Dansk Energi, DEFU rapport RA Solceller og spændingsvariationer i 0,4 kv net, [online] www. danskenergi.dk, July [3] Dansk Energi, Energinet.dk and DONG Energy, Scenarier for solcelle udrulning i Danmark, [online] da.scribd.dk, [4] B. Möller, S.Nielsen, K. Sperling, A Solar Atlas for Building- Integrated Photovoltaic Electricity Resource Assesment, in proc. International Conference on Sustainable Energy and Environmental Protection SEEP, June [5] G. B. Andresen, Solenergi kan blive en vigtig brik i Danmarks grønne omstilling, in Mandag Morgen newsletter, [online] [6] G. Kerber and R. S. H. Witzmann, Voltage Limitation by Autonomous Reactive Power Control of Grid Connected Photovoltaic Inverters, in proc. IEEE Conference on Compatibility and Power Electronics, [7] C. Gaudin, A. Ballanti and E. Lejay, Evaluation of curtailment option to optimize integration in Distribution Network, in proc. CIRED Workshop, [8] European Photovoltaic Industry Association (EPIA), Connecting the Sun, [online] [9] T. Stetz, F. Marten, M. Braun, Improved Low Voltage Grid- Integration of Photovoltaic Systems in Germany, in IEEE trans. on Sustainable Energy, [10] S. B. Kjær, Flicker and Photovoltaic Power Plants, in proc. European Photovoltaic Solar Energy Conference EUSEC, September The NET.dk project studies how to facilitate large-scale grid integration of solar into the existing grid. This is done by examining different types of grid-voltage control, applying SmartGrid functionalities and introducing novel ancillary services integrated into the inverters. The background for the NET. dk project is the already ongoing Photovoltaic Island Bornholm (IB I - III) projects, the EcoGrid EU project, and the Danish Cell Project. The first part of the NET.dk project will establish the theoretical framework for integrating large amounts of solar into the grid. The project will suggest, analyze and assess different solutions. In the second part, the proposed solutions are implemented into solar installations already deployed during the IB projects. Finally, the operation of the network without and with the developed solutions will be verified in a third part, which runs parallel to the first two. The project consortium is formed by: Danfoss Solar Inverters, which is in charge of the project management and also providing the inverter platform and test facilities; Centre for Electric Technology at the Technical University of Denmark, which will develop the required algorithms and test them with a hardware-in-the-loop grid simulator, for making solar systems `SmartGrid enabled and will also be the link to the EcoGrid EU project; EnergiMidt, a DNO which have been in the Danish -business for two decades and making the link to the IB projects and the IEA photovoltaic power system programme task 14; Østkraft, which is the local distribution network operator on the island of Bornholm. The NET.dk project is in part financed under the Electrical Energy Research Program (ForskEL, grant number 10698), administrated by Energinet.dk The full report can be downloaded at and at Questions to this report can be addressed to sbk@danfoss.com Danfoss Solar Inverters A/S Ulsnaes 1 DK-6300 Graasten Denmark Tel.: Fax: solar-inverters@danfoss.com DKSI.PM.208.D1.02 Produced by Metaphor March 2013