A wind speed prediction system based on new data preprocessing strategy and improved multi-objective optimizer

Abstract

The way to predict wind speed more accurately has always been one of the most important problems. The current combination prediction models have some limitations in data preprocessing, optimization algorithms and so on, which decreased the accuracy of the model. In order to solve the above problems, a new wind speed prediction system is proposed including a new data preprocessing strategy, combined neural network prediction and an improved multi-objective optimizer based on Halton low-discrepancy sequence and multi-distributed disturbance operator. The main contributions include: Firstly, a new data preprocessing strategy (GMM-ICEE) is proposed, which not only retains the characteristics of the wind speed data, but also removes the noise well. Secondly, Halton low-discrepancy sequence is introduced in the population initialization, which can effectively escape from local optimal solution. Thirdly, a new disturbance operator is proposed, which effectively solve the problem of optimal stagnation. Through six groups of experiments and two groups of discussions, it is verified that the accuracy, stability, generalization ability and advanced nature of the wind speed prediction system are satisfactory. Compared with traditional methods, the proposed wind speed prediction system improved by about 10% in prediction accuracy (MAPE) and by about 5% in CPU operation time.