Improvement of Particle Swarm Optimization Algorithm and Its Application in Power Data Mining

In power data mining, in order to cope with the rapidly growing data volume, particle swarm optimization algorithms should be improved and applied to power systems to improve performance and efficiency. In the process of particle search, the size of the particle is changed to affect the size of the ELM network structure. Such operation can not only expand the solution space of the problem, but also solve the shortcoming that traditional Particle Swarm Optimization-Extreme Learning Machine (PSO-ELM) can only find the optimal network structure in the network structure of the same scale. Experiments show that the PSO algorithm with mutation operator can converge to the optimal value more quickly. In addition, compared with the traditional ELM algorithm, Improved Particle Swarm Optimization-Extreme Learning Machine (IPSO-ELM) can obtain higher generalization performance and prediction progress in short-term power load prediction. The 4 curves respectively represent the MAPE value obtained by IPSO-ELM on the training set $1-4$, and the fluctuation range is all within 4%, which proves that IPSOELM has good stability. By improving the particle swarm optimization algorithm and applying it to power data mining, the efficiency of power system can be enhanced, energy waste can be reduced, and power supply quality and reliability can be improved.