采用分形插值的典型日负荷曲线改进预测方法

Improved Forecasting Method of Typical Daily Load Curve Based on Fractal Interpolation

提出了一种结合粒子群算法的改进分形预测方法。针对各年典型日负荷曲线形态相近且具有上移趋势的特点,采用调整向量来描述该趋势,在生成迭代函数系吸引子的过程中利用粒子群算法对调整向量进行优化。针对传统分形预测中迭代初始点经验性选取的问题,提出了利用"时序平移"的思想来计算迭代初始点的方法。结合调整向量优化和时序平移思想,建立改进的分形预测模型。最后,通过实例计算说明了该方法的有效性。

This paper proposes an improved method of forecasting typical daily load curve combined with particle swarm optimization. In point of the typical daily load curve characteristics of similar shape and upward trend year by year,this paper adopts adjustment vector to express the trend,and optimizes the adjustment vector using particle swarm algorithm in the progress of generating attractors of iterative function system. Aiming at the experiential selection problems of the initial iteration point in the traditional fractal prediction,a method based on time-shifting technique is proposed to calculate the initial point for the fractal interpretation iteration. Combined with the adjustment vector opti-mization and time-shifting thought,an improved fractal prediction model is then established. Case study results show that the proposed algorithm possesses advantages of more accurate forecasting results.