Abstract: To analyze the characteristics of both generation and load sides and the coupling effects of multiple influencing factors in power systems, a causal inference-based methodology to investigate the long-time-scale common-mode mechanisms between power sources and loads is proposed. First, correlations between meteorological factors, renewable energy generation, and electricity load are analyzed. Next, the greedy fast causal inference（GFCI） algorithm is employed to construct a causal graph, identifying causal relationships among the power sources, loads, and various influencing factors. Subsequently, a double machine learning（DML） approach is applied to quantify the dynamic causal effects of key influencing factors. Finally, a comparative analysis is conducted using publicly available datasets to evaluate the effectiveness of different causal inference methods. Case study results indicate that temperature significantly influences electricity load; specifically, due to the cumulative temperature effect under extreme heat conditions, the causal impact of temperature on load decreases by 200 MW per degree increase. The influence of wind speed on wind power output demonstrates segmented nonlinear characteristics, exerting a positive effect in low-speed ranges and turning negative during extreme high winds. The causal effect between solar irradiance and photovoltaic power output is significant at low irradiance levels, whereas it begins to decline once irradiance exceeds 300 W/m². The proposed method can qualitatively identify causal pathways between meteorological factors and source-load dynamics, and quantitatively evaluate the strength of causal effects from various influencing factors.