Abstract: Aiming at the security risk of distribution network caused by the private capacity increase behaviour of distributed photovoltaic(PV) users, this paper proposes a PV capacity expansion detection model based on grey wolf-particle swarm hybrid optimization discrepancy robust extreme learning machine. Firstly, the cosine similarity and dynamic time regularization are used to pre-process the PV power generation data to reduce the impact of regional meteorological differences. Secondly, the global search capability of the grey wolf algorithm and the fast convergence characteristics of the particle swarm algorithm are combined to optimize the implicit layer weights and bias of robust extreme learning machine to improve the robustness of the model to outliers. Altimately, the robustness of the model to capacity expansion is achieved by computing the violation expansion coefficient K to achieve the dynamic diagnosis of capacity increase intensity and time node. Based on actual PV data in Changsha region, the experiments results show that the probability density analysis of the violation expansion coefficient can identify the expansion behaviour as low as 10%, and the expansion time node positioning error is no more than 4%, and three violation expansion users are successfully detected in the actual case.

Key words: distributed photovoltaic, non-compliant capacity expansion detection, grey wolf-particle swarm, robust extreme learning machine, dynamic time regularization