基于矩阵填充的泛在电力物联网电能质量数据修复算法

电网智能优化运行依赖于对系统的泛在感知和完整正确的数据支持,这也是泛在电力物联网感知层必然要达到的最基本要求。在泛在电力物联网应用中,获取完整正确的量测数据是治理电能质量问题的基础。然而,在实际电网采集传输的全环节中,会不可避免地发生数据残缺。针对上述情况,提出基于低秩矩阵填充理论的泛在电力物联网电能质量感知数据补全新方法。首先证明了电能质量数据具有近似低秩的特性,以此为依据,设计多范数联合的秩优化模型,并应用交替方向乘子法将其分解为若干子问题分别求解。同时针对传统交替方向乘子法求解缓慢的问题,提出自适应迭代步长最优选取策略,加快模型求解速度。通过电压暂升、电压中断、脉冲振荡、电压暂降、谐波污染等高频故障场景验证所提方法的有效性,实验结果表明所提方法适用于多场景下的电能质量数据恢复,在缺失50%数据时仍能保证数据矩阵恢复误差在3%以内。

Matrix Completion Theory Based Recovery Algorithm for Power Quality Data in Ubiquitous Power Internet of Things

Automatic optimization and operation of power grid rely on the ubiquitous sensing and complete data support, which are the most basic requirements that must be satisfied in the sensible layer of the ubiquitous power Internet of Things. In the application of power Internet of Things, complete and correct measurement data acquisition is the basis for power quality control. However, in the whole process of data acquisition and transmission, data missing occurs inevitably. Aiming at the above situation, a new method based on low rank matrix completion theory for missing power quality data completion is proposed. This paper first proves that the power quality data have the characteristics of approximate low rank. Based on that, the multi-norm optimization model is designed. The alternating direction method of multipliers is applied to decompose the model into several sub-problems and solve them separately. At the same time, to accelerate the model solution, an optimal selection strategy of adaptive iterative step size is proposed. Effectiveness of the method is verified by high-frequency fault scenarios such as voltage swell, voltage interruption, pulse oscillation and voltage sag. Experimental results show that the method is suitable for power quality data recovery in multiple scenarios, and in the absence of 50% data, the data matrix recovery error is within 3%.