Short term electric load forecasting model and its verification for process industrial enterprises based on hybrid GA-PSO-BPNN algorithm - a case study of papermaking process

Abstract

Process industry consumes tremendous amounts of electricity for production. Electric load forecasting could be conducive to managing the electricity consumption, determining the optimal production scheduling, and planning the maintenance schedule, which could improve the energy efficiency and reduce the production cost. This paper proposed a short term electric load forecasting model based on the hybrid GA-PSO-BPNN algorithm. The GA-PSO algorithm is used in a short-term electric load forecasting model to optimize the parameters of BPNN. The forecasting model avoids the shortcoming that the prediction result is easy to fall into local optimum. The papermaking process, as one of the most representative process industries, is selected as the study case. The real-time production data from two different papermaking enterprises is collected to verify the proposed model. Besides the proposed GA-PSO-BPNN model, the GA-BPNN and PSO-BPNN based electric load forecasting models are also studied as the contrasting cases. The verification results reveal that the GA-PSO-BPNN model is superior to the other two hybrid forecasting models for future application in the papermaking process since its MAPE is only 0.77%.