A novel seasonal lag grey model for forecasting quarterly electricity consumption in China

Purpose Due to macroeconomic and seasonal impacts, electricity usage is highly uncertain, showing complex random, nonlinear and periodic patterns. To address this, a new seasonal lag grey forecasting model, TVGM(1,1,sin), is proposed to predict small sample series with long-term trends, quarterly changes and random nonlinearity. Design/methodology/approach First, trigonometric functions and time-varying parameters were added to the nonlinear grey model to create the TVGM(1,1,sin) model. Next, optimal nonlinear parameter values and time delay were found using the debugging method and genetic algorithm. Lastly, the model was used to forecast China’s quarterly electricity usage, showing it can effectively capture nonlinear and quarterly trends. Further tests confirm the model’s high accuracy, with a MAPE of just 3.98%. Findings A new seasonal grey TVGM(1,1,sin) model was built by adding trigonometric functions and time trends to the traditional nonlinear grey model. It fits quarterly cycles well, showing high adaptability for predicting complex, small-sample time series with quarterly periodicity. In case studies, it outperformed other models, proving its strong generalization ability. Practical implications This paper offers a scientific prediction model for China’s electricity use, which has seasonal cycles and complex nonlinearity. The prediction results can aid power firms and governments in efficient decision-making. Originality/value The main contribution of this article is to propose a new seasonal grey prediction model that accurately captures nonlinear dynamics, fitting the data sequence better. In addition, due to the presence of nonlinear parameters, the model is endowed with strong flexibility and intelligent algorithms are used to dynamically optimize the nonlinear parameters, further improving the prediction accuracy of the model.