Photovoltaic Power Impact Analysis Based on Nonlinear Correlation and Time Series Model

The output power of photovoltaic power generation is a multi-variable and coupled nonlinear random process. Traditional correlation analysis methods are ineffective in detecting the nonlinear relationship between photovoltaic power generation and its related influencing factors. To solve this problem, a nonlinear correlation analysis algorithm is proposed based on shrinking and extending the time window of the Chatterjee correlation coefficient. The time window shrinking strategy utilizes time encapsulation windows to partition the entire time series into several subsequences. It then progressively validates the time window pairs within these subsequences. The time window extending strategy selects pairs of time series that meet nonlinear correlation relationships by traversing the minimum window and expanding to the left and right areas. Experimental results show that the proposed algorithm outperforms existing methods in terms of precision, recall, F-Score, and running time. At the same time, the analysis of Anhui Shijiahu photovoltaic power station shows that the proposed algorithm can analyze nonlinear relationships more accurately than existing methods.