Application of Intelligent Algorithm Prediction Model Based on Particle Swarm Optimization on Power Load Forecast

In Multi-output Support Vector Regression (MSVR), choosing an appropriate combination of the penalty parameter and the kernel parameter is of great significance. To quickly calculate the optimal combination of these parameters, this paper proposes an AMPSO-MSVR model which adapts Adaptive Mutation Particle Swarm Optimization (AMPSO) to optimize MSVR. Firstly, a two-dimensional search space is formed, representing different value choices. Then, the particles' positions are randomly initialized and iterated to find the optimal solution, which represents the combination. Finally, the optimal combination is brought into MSVR to generate prediction values. The experimental analysis results of power load prediction show that the optimized AMPSO-MSVR model reduces the root mean square error by about 0.04 compared with traditional MSVR, and the average absolute error decreases by 0.0618. This verifies the feasibility of the model.