Causality analysis of meteorologically sensitive in offshore wind power generation based on the time-lagged convergent cross mapping

Abstract

Regarding the relationship between wind power generation and meteorological factors, previous studies tend to focus on the correlation between them, but correlation does not imply causation. In this context, we propose to use a combination of time-lagged convergent cross mapping (CCM) and graph network theory to investigate the causal relationship between wind power generation and meteorological factors. The effectiveness of time-lagged CCM is demonstrated by applying it to three scenarios of strong linear correlation, weak linear correlation and no linear correlation, respectively, and comparing it with Standard CCM and time-lagged Pearson correlation coefficient (PCC). Time-lagged CCM can accurately identify the causal relationship between wind power generation and meteorological factors and quantitatively assess the variables' causal intensity. Further, combined with graph network theory, we constructed a causal pattern diagram. From it, we can find that the causal relationship and intensity between wind power generation and meteorological factors also change dynamically throughout the year, following a specific temporal causal chain law. This finding is conducive to establishing a more accurate wind power prediction model, especially in the model’s feature selection and feature relationship analysis.