Application of hybrid neural networks combined with comprehensive learning particle swarm optimization to short-term load forecasting

Short term load forecasting is one of the key components for economical and safe operation of power systems. In competitive environment of electricity market, electricity utilities require more accurate load forecasting strategies to make better decisions on purchasing or generating electricity. This article offers a new method based on machine learning short-term load forecasting which is made up of a two-level feature selection technique and a new forecast engine. The feature selection part uses irrelevancy and redundancy filters to select best sets of input features. The proposed forecast engine is composed of a support vector regression machine, hybrid neural network and comprehensive learning particle swarm optimization. By applying comprehensive learning particle swarm optimization along with hybrid neural networks, the accuracy of forecasting is improved and its error decreases effectively.The proposed strategy is tested on PJM and AEMO electricity markets. The numerical results show the effectiveness and robustness of this method in comparison with recent short-term load forecasting methods.