Daily load forecasting of urban gas based on GRA-ABC-BPNN model

Urban gas load forecasting is of great significance for rationally and efficiently deploying gas resources and solving the problem of gas consumption by urban gas users. Herein, the 11 identified influencing factors of the daily gas load were analyzed through the Gray Relation Analysis (GRA) method and screened according to the correlation degree, having the influencing factors with little correlation eliminated one by one, and using the remained influencing factors with high correlation degree as the input of the Back Propagation Neural Network (BPNN). Meanwhile, the BPNN weights and thresholds were optimized with the Artificial Bee Colony (ABC) algorithm. Besides, a GRA-ABC-BPNN forecasting model was established to predict the daily load of urban gas, and the accuracy and effectiveness of the established forecasting model was verified. As shown by the results, the Mean Absolute Percentage Error (MAPE) of the daily load of urban gas forecast by GRA-ABC-BPNN model is 0.552 8%, while the MAPEs of Genetic Algorithm-BPNN (GA-BPNN) model and ABC-BPNN model are 1.491 3% and 0.636 9%, respectively. This indicates that the GRA-ABC-BPNN forecasting model is an effective and accurate method to forecast the daily load of urban gas, and it could provide a new way for daily load forecasting of urban gas.