PSO-BP Based Wind Power Prediction

With the increase in the proportion of wind energy in the power system, its intermittentity and volatility pose a challenge to the stable operation of the power system. Accurate wind forecasts are crucial. This paper proposes a wind energy prediction method for backpropagation (BP) neural networks which rely on the particle swarm optimization (PSO) algorithm improved. First, the year-on-year operational data of the wind farm, including multi-dimensional information, such as wind speed and direction, were standardized for pre-processing. The processed data are then inputted to the BP neural network model which is founded on the trained PSO and the initial weight and threshold are optimized by PSO. Case studies show that this method performs best in root square error, average absolute percentage error and accuracy compared to traditional methods such as BP neural network, support vector apparatus, and chronological series analysis, achieving more accurate and stable wind energy prediction.