Abstract: In order to accurately predict the potential of electric energy substitution in aggregate and in segments, a cumulative electric energy substitution forecasting model considering multi-dimensional historical data is proposed. First, the main factors influencing the development of electric energy substitution are quantified based on multi-dimensional data; then, a particle swarm optimization algorithm with inertia weights is used to optimize the traditional support vector machines forecasting model. The validity of the method is verified by the historical data of S province in China. The results show that multi-dimension historical data can improve the prediction accuracy, and the new intelligent forecasting algorithm has the advantage of good adaptation and fast convergence speed.