基于改进型SVR的电网短期负荷预测

常规的支持向量回归预测模型（SVR）预测算法采用人工经验的方法对RBF核函数参数、不敏感系数和惩罚系数等参数进行选取，其性能会因随机选取的参数而变得随机和不确定。人工鱼群算法的初始参数会对整个算法的优化性能产生较大影响，将粒子群优化算法和混沌机制引入常规人工鱼群算法，对其进行改进，可以提高种群多样性和全局寻优能力，避免优化算法陷入局部最优解。通过实验方法对改进型人工鱼群优化SVR预测模型的性能进行分析。结果表明，所研究的短期负荷预测精度较高，具有较好的工程应用价值。

Short-Term Power Load Forecasting Based on Improved SVR

In conventional SVR prediction algorithm, parameters including RBF kernel function parameters, insensitive coefficient and punish coefficient are selected manually. The random selected parameters lead to performance uncertainty. The initial parameters of artificial fish algorithm greatly influence optimization performance. By applying particle swarm optimization and chaotic theory to conventional artificial fish algorithm, the global optimization ability can be improved. Detailed analysis are performed are testing data of improved SVR load prediction model. Results show that the proposed method has better prediction accuracy and good engineering application value.