Application CUDA for optimization ANN in forecasting electricity on industrial enterprise

Abstract

Technological progress in the manufacturing sector is characterized by an increase in energy consumption and, consequently, an increase in electricity consumption. It’s necessary to carry out electricities economical consumption to meet the growing demand for electricity. The problem of forecasting of energy consumption is a complex multi-factor problem with nonlinear dependencies. Due to the complexity of the calculations for the solution of this problem requires large computational resources. Therefore there is a need of optimization algorithms to improve the quality of the forecast. This article describes the use of parallel computing on the GPU algorithm neural network training based on CUDA technology, to optimize the energy consumption prediction process in an industrial plant. According to the results of the experiments presented in this paper, the parallel algorithm has reached the required prediction accuracy for a shorter period of time. Applying the proposed algorithm can enable enterprises to get a more accurate prognosis and reduce the costs associated with payment of electricity