Short‐term load forecasting is essential for reliable and economical operation of the power system. Gaussian process regression (GPR), as a novel machine learning prediction algorithm, has become one of the commonly used algorithms in the field of short‐term load prediction because it can fully consider the nonlinear and uncertain characteristics of short‐term load sequence. However, due to the unreasonable selection of training set and the inaccurate acquisition of hyper parameters, the prediction accuracy of GPR decreases. In order to improve the accuracy of power system load forecasting, a GPR prediction model based on density peak clustering (DPC) and information sharing antlion optimizer (ISALO) is proposed. Firstly, the DPC is used to find similar days from the historical load data to construct a more reasonable training set. Then, the ISALO, an improved ALO by introducing information sharing mechanism, is used to optimize the hyper parameters of the GPR. Experiments show that the DPC‐ISALO‐GPR model has a 3.33 and 1.22% reduction in mean absolute percentage error compared to Back Propagation Neural Network (BP) and support vector machines, which is suitable for engineering practical applications. © 2020 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

中文翻译：

---

基于高斯过程回归与密度峰值聚类和信息共享的Antlion优化器的短期负荷预测

短期负荷预测对于电力系统的可靠和经济运行至关重要。高斯过程回归（GPR）作为一种新型的机器学习预测算法，已经能够充分考虑短期负荷序列的非线性和不确定性特征，因而成为短期负荷预测领域中的一种常用算法。但是，由于训练集的选择不合理以及超参数的获取不准确，使得GPR的预测精度下降。为了提高电力系统负荷预测的准确性，提出了一种基于密度峰值聚类（DPC）和信息共享蚁群优化器（ISALO）的GPR预测模型。首先，DPC用于从历史负荷数据中查找相似的日期，以构建更合理的训练集。然后，ISALO，通过引入信息共享机制改进的ALO用于优化GPR的超参数。实验表明，与反向传播神经网络（BP）和支持向量机相比，DPC-ISALO-GPR模型的平均绝对百分比误差降低了3.33和1.22％，适用于工程实际应用。©2020日本电气工程师学会。由Wiley Periodicals LLC发布。©2020日本电气工程师学会。由Wiley Periodicals LLC发布。©2020日本电气工程师学会。由Wiley Periodicals LLC发布。