GVM based copy-dynamics model for electricity load forecast

Electricity load forecast, as the core of electricity scheduling, plays a vital role in meeting the basic needs of modern human life. It has been widely studied in the past few decades. However, literature studies have shown that, as a problem of time series forecast, electricity forecast is prone to be influenced by many environmental factors, which result in lacking accuracy and stability in practice. In this paper, the General Vector Machine (GVM), a new type learning machine which was derived from Neural Network (NN) and Support Vector Machine (SVM), is applied into electricity load forecast. Meanwhile, copy-dynamics idea is introduced to electricity load forecast. Results reveal that, based on copy-dynamics, the maximum precision promotion of GVM reaches 71.7%, compared with BP. Hence, GVM and copy-dynamics models have great potential in electricity load forecast.