基于Fisher主元分析和核极限学习机的非侵入式电力负荷辨识模型

Non-intrusive Power Load Identification Model Based on FPCA and KELM

非侵入式电力负荷监测与辨识是实现泛在电力物联网客户侧智能感知的关键技术。针对现有辨识模型存在的特征冗余度高、辨识准确率差、计算效率低等问题,提出了一种基于Fisher主元分析(Fisher principal component analysis,FPCA)和核极限学习机(kernel extreme learning machine,KELM)的非侵入式电力负荷辨识模型。首先,选取电流、功率、谐波含有率等稳态特征作为原始输入变量,运用Fisher得分和主成分分析相融合的Fisher主元分析法剔除可分性较差的无效特征,同时降低有效特征之间的相关性;然后,引入径向基核函数搭建网络结构,并采用遗传算法(genetic algorithm,GA)对惩罚系数等模型参数进行寻优,从而建立核极限学习机分类模型进行负荷识别;最后,通过公开的TIPDM负荷数据集进行算例分析。仿真结果表明,所提模型相比于传统负荷辨识模型具有更好的辨识准确率和计算效率,运用该模型可对常见家用负荷进行有效识别。

Non-intrusive power load monitoring and identification is the key technology to realize customer-side intelligent sensing in the ubiquitous power Internet of things. Aiming at the problems of high feature redundancy, low identification accuracy and low calculation efficiency in the existing identification model, a non-intrusive power load identification model based on Fisher principal component analysis and kernel extreme learning machine is proposed. Firstly, the steady-state characteristics such as current, power and harmonic contents are selected as the original input variables, and the Fisher principal component analysis(FPCA), which combines Fisher score and principal component analysis, is used to eliminate the invalid features with poor separability and to eliminate the correlation among the effective features at the same time. Then, radial basis function is introduced to build the network structure, and genetic algorithm(GA) is used to optimize the model parameters such as penalty coefficient, so as to build the kernel extreme learning machine(KELM) classification model for load identification. Finally, the open TIPDM load data set is used for example analysis. The simulation results show that the proposed model has better identification accuracy and calculation efficiency than the traditional load identification models, and it can effectively identify common household loads.