A novel scheme of control commands for smoothing PV power by using ANN

This paper proposes a novel scheme of control commands for smoothing photovoltaic generation (PV) power by using ANN. The proposed method decides the power command as follows. The moving average is performed to smooth the selected one-day PV data. The smoothed PV data is shifted by a half period of moving average to compensate for time delay. The smoothed-and-shifted PV data are used as the target outputs for artificial neural networks (ANN). The structure of ANN is decided by the optimization method. The power command for smoothing PV power is decided by using ANN whose structure is led by learning and optimization. The proposed method has improved the time delay compared to the conventional method. The proposed method has reduced the energy storage capacity by 254 MJ compared to the conventional method, although it has increased the power in LFC band by 39.2 kW compared to the conventional method. Energy control has performed to reduce the energy storage capacity while maintaining the power smoothing effect. As a result, the proposed method with energy control has implied a possibility that it can reduce the energy storage capacity by 5.3% compared to the conventional method with energy control.