基于压缩感知理论的缺失数据集下线损预测模型

Line Loss Prediction Model Under Missing Data Set Based on Compressed Sensing Theory

线损预测是电网企业进行线损管理的基础,而电力系统中数据收集与传输过程中不可避免出现各种异常状况,导致线损数据缺失,影响线损预测精度。为解决这一问题,应用压缩感知理论研究矩阵稀疏变换方法和矩阵重构算法,实现电网运行缺失数据的补全与重建,利用基于自适应噪声的完整集成经验模态分解建立线损预测模型,完成缺失数据集下的线损预测。某10 kV配电网算例验证结果表明,在数据量较大或数据缺失情况较严重的情况下,基于压缩感知理论的数据恢复方法能比传统方法更好地修复原始数据,恢复原始数据的变化趋势,提高线损预测精度。

Line loss prediction is the basis for line loss management by power grid enterprises, while avoidable abnormal conditions in data collection and transmission in the power system may cause line loss data missing and affect prediction precision. To solve the problem, this paper applies the compressed sensing theory in studying the method of matrix sparse transformation and matrix reconstruction so as to finish completion and reconstruction of missing data. It uses complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to establish the line loss prediction model and finish line loss prediction under missing data set. An actual example of one 10 kV power distribution network has verified that under the condition of large data amount or serious data missing, the data repairing method based on the compressed sensing theory is preferable to repair original data, recover variation trend of original data and improve line loss prediction precision compared with traditional methods.