基于混沌粒子群—高斯过程回归的饱和负荷概率预测模型

饱和负荷预测能有效预估区域电网的发展方向和最终规模,为电网规划及电力市场中长期交易提供指导。针对饱和负荷预测不确定性强、时间跨度大的特点,文中采用基于高斯过程回归(GPR)的概率预测模型进行饱和负荷预测,并通过改进混沌粒子群算法(MCPSO)实现以和方差(SSE)最小为目标的模型超参数优化求解;在综合考虑饱和负荷影响因素随机性的基础上,建立了改进混沌粒子群—高斯过程回归(MCPSO-GPR)饱和负荷预测模型,并在多情景下利用上述模型进行饱和负荷预测,同时结合饱和判据得到多情景下饱和负荷的规模和时间。算例分析表明,所述模型不仅具有较高的预测精度,而且可增强预测的弹性。

Forecasting Model for Saturated Load Based on Chaotic Particle Swarm Optimization-Gaussian Process Regression

Saturated load forecasting could effectively estimate future direction and final scale of the regional power grid, providing guidance for planning and mid/long-term transactions of the power market. Firstly, a probabilistic forecasting model based on Gaussian process regression(GPR)is adopted for saturated load forecasting, aiming at its characteristic of strong uncertainty and large time span. Secondly, the optimal solution of model hyper-parameters with the objective of minimizing the sum of squares due to errors(SSE)is realized by a modified chaotic particle swarm optimization(MCPSO)presented. In consideration of the randomness of the factors influencing the saturated load, a saturated load forecasting model based on modified chaotic particle swarm optimization-Gaussian process regression is proposed. Thirdly, in multi-scenarios using the above model while taking saturation criterion into account could forecast the saturated load and obtain multi-scenario scale and time-point. Finally, case studies show that this model not only has high precision, but also enhances the elasticity of forecasting results.