To improve the accuracy of power system load forecasting and maintain the safety and stability of power system operation

a combination of self-organizing maps（SOM）clustering based on feature vector and improved radial basis function（RBF）neural network for power load forecasting model is proposed. The samples are clustered by extracting feature vectors that reflect the characteristics of the daily electric load. Data with similar features are used as training samples for the neural network to improve sample regularity. To overcome the effects of gradient descent and local optimum on the network prediction accuracy

the particle swarm optimization（PSO）algorithm is used to modify the neural network particle swarm velocity and position. The validity and good adaptability of the proposed model are verified based on power load data of distribution network in an area.