Research On Non-intrusive Load Disaggregation Technology Of Smart Meter

Smart meter is an important terminal device of smart grid,with the popularization of the concept of ubiquitous power Internet of Things,consumers are gradually becoming more concerned about electricity usage information,this Put forward higher requirements on the level of intelligence of the electric meter.Through the efficient combination of intelligent electric meters and non-intrusive load identification technology,a non-intrusive load sensing system of intelligent electric meters is constructed to analyze electricity consumption details and dig more electricity usage rules,this can not only improve user experience and promote good interaction between the grid and users,but also promote the upgrade of the smart energy industry and advocate a green and low-carbon lifestyle.The article first puts forward the design framework of the non-intrusive load sensing system of intelligent electric meters based on demand analysis.A modular hardware architecture is designed according to the scalable concept of a new generation of smart meters,and realize the high-efficiency communication and high-speed data sharing with the "dual core" of the intelligent electric meter.In view of the diverse user needs and the limited capacity of the embedded platform,a reasonable data storage management strategy and a flexible multi-task implementation plan are designed.Subsequently,the dimension of load identification characteristics was expanded by mining the non-electrical load characteristics associated with user behavior and habits.On the basis of event-based,the cumulative sum of changes in the time series of change points is used as the basis for the judgment of load event switching.Under the framework of the forward distribution model,the regression tree is used as a weak classifier for training,use negative gradients to fit the training residuals from the previous round,and then the gradient boosting decision tree is constructed,so as to improve the performance of the strong classifier of the final linear combination,which is used as an algorithm model for load identification and applied to an embedded platform.In order to further optimize the effect of the identification algorithm model and obtain the optimal parameter combination of the algorithm model,this paper proposes a global parameter optimization algorithm based on optimized tabu search and particle swarms,which can achieve better balance on efficiency and accuracy in global parameter optimization.Finally,in order to realize the effective management of a large number of users’ electricity load information and a non-intrusive load sensing system,this paper designs a practical framework with strong compatibility.According to demand analysis,a database storage management strategy is proposed,which covers a variety of data tables.It can not only record and save electricity consumption information,but also extract and summarize electricity consumption habits from it,which will be fed back to the load identification algorithm as a more accurate training sample.The designed system communication mechanism interacts on a level-by-level basis,with clear division of labor and clear interfaces between levels.The data encryption layer is designed and implemented by advanced encryption algorithms,which effectively guarantees the security of data information and the practicability of the system architecture.The application of non-intrusive load identification technology on smart meters not only has positive practical significance,but is also an important part of advocating the concept of low-carbon life and building a green and efficient energy society.The expansion of load identification features,the optimization of load identification algorithms,and the architecture management of load sensing systems involved in this article have laid a good foundation for the promotion of non-intrusive load identification technology for smart meters.