基于多变量加权一阶局域混沌预测模型优化及应用

鉴于实际应用中多变量因素对混沌预测的影响,提出了多变量时间序列相空间重构方法,以此为基础建立多变量加权一阶局域混沌预测模型。引入等概率符号化极大联合熵求取延迟时间、最小香农熵法求取嵌入维数,实现多变量混沌预测模型子序列重构;对实际序列采用区间邻近点法确定预测中心点的邻近点,避免产生伪邻近点;最后用关联分析确定观测变量。将该模型应用于短期电力负荷预测,分析气温等影响因素与电力负荷的相关程度,引入气温时间序列作为另一观测变量,实验证明相对于单变量预测方法提高了预测精度。

Optimization and Application of Weighted One-rank Local Chaos Prediction Model Based on Multi-variable

In view of the influence of multi-variable on the chaotic prediction in practical application, a method for phase space reconstruction of multivariate time series is proposed, and a weighted one-rank local chaos forecasting model on multi-variable was established. The equal probability-based maximum joint entropy and the minimum Shannon entropy are introduced to get the delay time and the embedding dimension respectively, realizing the sub-sequence reconstruction to the chaotic prediction model. The nearest neighbor point method is used to determine the neighborhood of the prediction center to avoid false neighbors, and the correlation analysis is used to determine the observed variables. The model was applied to short-term load forecasting, and the temperature time series was introduced as another observation variable by the analysis of the impact of temperature and other factors related to electric load. The experimental results showed that the prediction accuracy was improved compared with the single variable forecasting method.