Research And Application Of Non-Invasive Load Monitoring Based On Event Detection And Color Coding

Energy crisis and climate change are two major challenges facing the world today.Improving energy consumption structure to reduce fossil energy demand is the key measure to deal with these two challenges.With the development of China’s economy,the total electricity consumption of residents accounts for an increasing proportion of the total electricity consumption of the whole industry.At present,it has become the third largest electricity consumer except industry and manufacturing.Non-intrusive load monitoring can realize the operation monitoring and energy consumption evaluation of various electrical equipment within the family by analyzing the total load data of the family,which is convenient for users to understand the operation of various electrical equipment within the family in a timely and comprehensive manner,so as to optimize the power consumption structure within the family and achieve the goal of energy saving and emission reduction.This paper mainly studies the event-based non-intrusive load monitoring scheme,and carries out the following research work around event detection,load identification and engineering application:(1)An improved real-time event detection algorithm is proposed for the problems that the standard GLR algorithm requires manual parameter setting,the standard χ~2 GOF algorithm has a high miss detection rate and the traditional event detection algorithm has a low processing efficiency.The algorithm is implemented by the combination of GLR algorithm and χ~2 GOF algorithm,and improvements are made to the processing of noise signals,window length,sliding step and decision threshold of GLR.The experimental results show that the improved algorithm outperforms the traditional algorithm in several indexes,and its operation efficiency is twice that of the χ~2 GOF algorithm for the same size data set,which meets the demand of real-time event detection.(2)In order to improve the identification accuracy of load events,this paper proposes a color coding-based load event identification method:on the basis of retaining the basic waveform of active power,three features of active power(R),reactive power(G)and reactive power trend(B)are fused to construct a color image of load events,and the image is trained and recognized based on AlexNet convolutional neural network with adjusted parameters.The experimental results show that the load event feature can stably and effectively distinguish the load events of different devices,and achieve better classification performance than the traditional features.(3)A non-intrusive load monitoring platform is designed and implemented from practical engineering applications.A hardware platform capable of data acquisition,data processing and data communication is designed based on the JSY1009 single mutual inductive power metering module,STM32F407ZG microcontroller and Wi-Fi communication module ESP8266.Based on the Vue front-end framework and Django back-end framework,a software application platform is designed that can realize three major functions:active power event detection,equipment energy consumption statistics and equipment label update.The design and implementation of this platform provides new ideas for future practical engineering applications of non-intrusive load monitoring technology.