考虑爬坡特性与预测区间优化的电热水器集群功率区间预测

Interval Prediction of Aggregated Power for Electric Water Heaters Considering Ramp Characteristic and Prediction Interval Optimization

根据电热水器(EWH)集群功率的波形特征,提出了一种考虑爬坡特性和区间优化的EWH集群功率短期区间预测方法。首先,针对EWH负荷功率的不确定性,提出了一种考虑样本分布多源异构特性的结合集合经验模态分解(EEMD)、主成分分析(PCA)和多核相关向量机(MKRVM)的高精度组合点预测模型。然后,为获得期望预测覆盖率下宽度更窄的预测区间,综合区间预测覆盖率、区间平均宽度和累积宽度偏差等区间评价指标,设计了一种核密度估计(KDE)与粒子群优化相结合的改进预测区间优化方法,改善了MKRVM-KDE区间结构性能,避免了参数选择的随意性。最后,采用EWH聚合功率数据对算法有效性进行了验证。结果表明,该预测方法具有较高的预测精度和较好的清晰度,能够提供高质量的预测区间。

According to the waveform feature of the aggregated power for electric water heaters(EWHs), a short-term interval prediction method is proposed, which considers the ramp characteristic and interval optimization. Firstly, in view of the uncertainty of load power for EWH, a combinatorial point prediction model with high-precision considering multi-source heterogeneous characteristics of the sample distribution is presented, which combines with ensemble empirical mode decomposition(EEMD),principal component analysis(PCA) and multi-kernel relevant vector machine(MKRVM). Secondly, to obtain a narrower prediction interval with the expected prediction coverage, the evaluation indices of the interval prediction coverage, interval average width, and cumulative width deviation are combined to design an improved prediction interval optimization method integrating the kernel density estimation(KDE) and particle swarm optimization, which enhances the performance of MKRVMKDE in interval structure and avoids the randomness in parameter selection. Finally, the aggregated power data of EWH is used to verify the effectiveness of the approach. The results show that the prediction method has high prediction accuracy and better clarity,and it can also provide prediction intervals with high quality.