Research And Implementation Of Adiustable Load Management Subsystem Based On Electricity Consumption Behavior And Price Factor

The construction of the smart grid promotes the construction of flexible and interactive smart power consumption,allowing the users on the demand side to actively participate in the operation of the grid.Therefore,carrying out demand-side management of the smart grid,analyzing users’ power consumption behavior,and reasonably regulating users’ peak load,making the complex power market under the smart grid architecture maintain the balance between power resource consumption and power demand in a user-friendly way are the keys to the construction of smart grid.To meet these demands,this thesis studies from two aspects:user electricity behavior prediction and adjustable load control.The meaning of the former lies in accurately estimating the user’s electricity load,efficiently and rationally planning and utilizing electricity resources,which can avoid resource waste.At the same time,the research on external factors of users’ electricity consumption behavior can also improve the prediction effect,so as to adjust the peak and valley of users’ load from human controllable factors such as electricity prices,which also supports the latter research.The meaning of the latter is to smooth the user load curve,reduce the pressure on the grid,and at the same time alleviate the contradiction between supply and demand.Among the key algorithms studied in this thesis,the research on the prediction of users’ electricity consumption behavior is carried out from two perspectives based on the raw data.Firstly,a univariate load forecasting model integrating EMD and GRU is proposed for the easy-to-obtain load data itself.Secondly,for multivariate data including external influencing factors,a multivariate load prediction model based on pre-attention mechanism and convolutional neural network is proposed.The former applies the empirical mode decomposition of the data stabilization method in the field of signal processing to the data processing of the load curve and uses a gated recurrent unit network with a simpler structure,which makes the model converge faster and more efficient.The latter improves the weight calculation of attention,comprehensively analyzes the influence weight of time dimension and factor dimension on power load,and uses a convolution neural network to extract the characteristics of two-dimensional multivariate time series data for the follow-up model to fully learn,so as to better capture its internal change law,and specifically analyze the influence weight of different external factors on power consumption behavior.The superiority of the model in prediction accuracy is proved by comparison with the basic comparison method on domestic and foreign data sets,and the weight analysis in the multivariate prediction model also proves the practical significance of the model in practical application.For the research of adjustable load control,based on improved DQN and multivariate prediction model,a load regulation model is proposed in this thesis.The structure of DQN is improved and integrated with the previous multivariate prediction model to iteratively simulate the dynamic interaction between the power sales side and power demand side,so as to realize the purposes of load reduction and load peak shaving.The electricity price data obtained in the interactive process will be output as pricing suggestions to provide a reference for relevant departments.Through the simulation experiment on the real data set,according to the change of the user load curve,the change of the daily electricity consumption,and the change of the user’s electricity cost before and after the experiment,it is proved that this model can effectively reduce the peak value of the user’s electricity load,and can reduce the electricity cost of users to a certain extent.Further,since the above algorithm models can support the two important business applications of load control and predictive analysis in the new generation of load management systems,this thesis designs and implements an adjustable load management subsystem based on electricity consumption behavior and price factors,and It is provided to the relevant staff of the electric power system through a web application,so that analysts in the electric power field without algorithm experience can conduct more efficient analysis directly through the visual interface and combined with their electric power expertise.This thesis firstly introduces the research background and significance of the whole subject and expounds the research status and related technologies at home and abroad.Then,based on the business scenario,the demand analysis is carried out,the corresponding algorithm is proposed for the key demand,and the experimental process is introduced.Then it introduces the design and implementation of the whole system,as well as the deployment and testing of the system.Finally,the work of this thesis is summarized,and the future research is also prospected.