A New Method for the Daily Electricity Consumption Forecast of Energy Intensive Enterprise

Abstract

Electricity consumption forecast is very important for both suppliers and large consumers. The electricity consumption of a large enterprise is different from the regional grid, especially for an energy intensive enterprise (EIE). This paper investigates the daily electricity consumption forecast of an EIE by the operating condition. By observation, the electricity consumption is related to maintenance duration and production quantity. And through verification, the relationship is linear. Therefore, the production plan and maintenance schedule can be considered as the input of the forecast model. After the preliminary selection of the feature set, the Nonnegative Least Squares (NNLS) method is applied to build a linear regression model with nonnegative coefficients. In addition to NNLS, a new feature selection method based on correlation analysis and greedy search is applied to select the relevant input features on the available items of the production and maintenance schedule, which has greatly improved the efficiency of feature selection, and decreased the influence of multicollinearity and local optimum. By using the real data from a typical modern large steel corporation, numerical test results show that results obtained by the NNLS are rational, and the improvement of forecast results are obvious in several evaluation criteria.