Since available forecast models have a lack of validity and applicability, a new electric load forecast method, considering co-integration relation and hysteresis effect of influenced factors, is proposed. Firstly, by the use of incidence matrix, this paper obtains the close associated factors which affects load change. Secondly, combined with metrology model, X-12-ARIMA, the data of load and influence factors is decomposed into three different parts. Then, based on the seasonal decomposition, it takes lag steps about time lag effects testing and uses Principle Component Analysis (PCA) for de-noising to enhance data purity further. Finally, according to monthly and seasonal load forecasting, numerical examples are studied to verify effectiveness and applicability of proposed method through comparing with other method forecast results and non-linear model forecasting accuracy improvement.