Local Voltage Control in Converter Based PV Generation

Abstract: As the world is trying to rid itself of its dependence on fossil fuels, the need for renewable energy is rapidly increasing. In the power distribution network (DN), this takes the form of distributed generation (DG) from renewable sources, such as photovoltaic (PV) power production. But as more DG is installed, it can lead to issues in the DN such as overloading and overvoltage when production is high. Today, in order to combat these issues, the DN is reinforced, which might be costly, and take time. This thesis has instead studied an alternative to network reinforcement, active network management (ANM), in order to avoid overvoltage. By implementing PI controllers in converter based DG, the power output from the DG can be curtailed, and thus reducing the voltage locally. By implementing local control of the power output in the DG, the voltage is kept within allowed limits across the feeder. In this thesis, one such ANM method was studied, which implements delta control of the active power produced by PV, in order to avoid overvoltage in a low voltage (LV) DN. Stability of the method has previously been shown analytically. In this thesis, the method was studied numerically in order to analyse the performance of the controllers in different test systems. The test systems were based on the CIGRE LV European distribution network benchmark, and was implemented into DIgSILENT PowerFactory to run load flow calculations, on different cases. In the different cases, properties of the PI controller, the power lines, and the loads were varied. After running all simulations, the performance of the controller for each case could be judged against criteria which were decided upon, to determine if the performance was feasible or not, along with stability of the controller. After analysing the performance in all cases, whether or not the controller was feasible seemed to be related to the loop gain of the entire system. This in turn seemed to be proportional to the number of PV units installed, the resistance in the system, the proportional gain of the PI controller, as well as inverse proportional to the integrating time constant of the PI controller. The loads did not affect the feasibility of the system