Power Quality Missing Data Completion In Ubiquitous Power Internet Of Things

Intelligent optimal operation of power grid depends on ubiquitous perception of the system and complete and correct data support,which is also the most basic requirement that ubiquitous perception layer of the Internet of Things must achieve.However,data incompleteness will inevitably occur in the whole process of data acquisition and transmission in the actual power grid,which will interfere with the service of the power grid to varying degrees,affect the normal operation of the service,and even lose the observability of the whole power system in serious cases.Therefore,it is of great significance to study the data completion of missing measurements.In view of the above situation,the representative power quality data in power grid measurement are selected to study,and a missing data completion method based on low rank matrix filling theory is proposed.Firstly,the components of the measured data are divided.On this basis,the approximate low rank characteristics of the measured data are studied in depth.Aiming at the problem that the traditional method of evaluating the low rank of matrix still stays at the qualitative level and can not provide reliable judgement for the feasibility of data recovery,the index of evaluating the low rank of matrix quantitatively is put forward and its mathematical derivation is given.After feature analysis,considering the influence of complex noise on data recovery,a data recovery model considering multi-noise effect is obtained by improving the classical model and adding matrix norm constraint noise to eliminate noise interference.Considering the real-time requirement of data recovery,a fast algorithm for solving the model is designed based on alternating direction multiplier method.After the original problem is decomposed into several optimization sub-problems by using alternating direction multiplier method,the analytical solutions of each sub-problem are obtained respectively,thus avoiding the problem of solving the whole model or nesting iteratively after model decomposition.At the same time,aiming at the problem of slow iteration of traditional multiplier alternating direction method,an optimal selection strategy of adaptive iteration step size is proposed to further accelerate the model solution.Finally,the validity of the waveform data verification method in typical power quality scenarios is selected.It includes high frequency fault scenarios such as voltage sag,voltage interruption,pulse oscillation,voltage sag,and waveforms under different distortion rate harmonics.The experimental results show that the proposed data recovery method is suitable for multi-scene measurement data recovery,and can ensure that the recovery error of data matrix is less than 3% when 50% data is missing.In addition,the proposed method can speed up the model solution,reduce the data recovery time by more than 60%,and achieve fast and accurate data recovery.