Distributed Photovoltaic Installation Admittance Optimization: An End to End Approach

Distributed photovoltaic installation admittance capacity (PVAC) refers to the maximum capacity of photovoltaics that can be accommodated by the distribution system. The economic and reliable PVAC evaluation is critical for distributed photovoltaic capacity planning and configuration. However, lacking sufficient measurement devices and unachievable to full-scale perception of the practical distribution system impose challenges for PVAC assessment. This paper proposes an end-to-end approach to evaluate PVAC using the data from gird terminals and smart meters, where network topology and line parameters will be unnecessary. Specifically, an end-to-end power flow variables mapping model is proposed, which is essentially using the adaptive online deep learning algorithm to model and train the functional relationship of terminal active power, reactive power and voltage. Based on the above mapping relationship analysis, an end-user data driven photovoltaic installation admittance optimization model is established, which models the PVAC evaluation issue as a convex optimization model with differentiable objective function. Then a projected gradient descent algorithm based on adversarial attacks mechanism is proposed to solve the optimization model to obtain the maximum PVAC. Finally, the effectiveness of the proposed method is verified on IEEE 33-node, IEEE 123-node and actual distribution systems.