Research On Two-level Decision Model Of Park Energy Use Optimization Based On Power Load Forecasting

"Flexible,open and efficient interaction" is the prominent feature of the new power system in the future.As the main body of the new power system,new energy will also put forward higher requirements for the load side regulation ability of the power system.In addition,the proportion of flexible loads such as electric vehicles and intelligent parks in the power system is gradually increasing,and the randomness of the "source-load"side is further highlighted,which brings new challenges to the safe and stable operation of the power system.In this case,based on intelligent optimization algorithm and other key technologies,the load-side resource energy consumption curve optimization is an effective measure to improve the power grid regulation capacity and reduce the user energy cost,which has important practical significance for the high-quality development of electric energy in China.In this paper,energy use optimization in parks is taken as the background,and the value is clustered and predicted based on the power load data,so as to study the energy use optimization strategy of parks,providing theoretical support for the energy use optimization service of China’s load aggregation parks and other user groups.Firstly,the clustering and load prediction model of users’ electricity consumption behavior in the park is established,and the data results of load clustering and prediction are input into the energy consumption optimization model.In order to extract and integrate the load characteristics of each main user in the park and improve the efficiency of energy consumption optimization,an improved K-means clustering model of electricity consumption behavior was constructed.The error caused by manual operation was reduced by introducing DBI index to determine the initial clustering number,and the atypical days were eliminated to improve the representativeness of the clustering curve.Then,a short-term load prediction model based on LSTM algorithm is established to predict the load value in the next 24 hours.In order to improve the accuracy of prediction,training set and test set are established,and the root mean square error is selected to determine the prediction level.After several training and network update,the prediction accuracy is improved.Then,the clustering and load prediction model of users’ power consumption behavior and the two-layer decision model of user-park operators’commercial performance optimization are constructed.The decision-making body of users in the park at the lower level takes the reduction of energy cost and maintenance of comfort as the goal,formulates the user demand response participation strategy,and solves the problem with linear programming method.The decision-making body of operators in the park at the upper level purchases demand response services from users in the park internally.In order to improve the comprehensive service level and benefit of the park,the optimal scheduling strategy of the park’s demand response pricing,park load,energy storage and other energy-using facilities was developed by using differential evolution method.Finally,in order to verify the validity and practicability of the proposed model,an example is given based on the load data of a certain park.The results show that the improved K-means clustering algorithm has significant clustering characteristics of users’ electricity consumption,with great differences between them.The 1STM-based short-term load forecasting model of the park has high accuracy and small root mean square error after updating the network.The two-layer optimization model can reduce the peak-valley difference rate of the park,relieve the pressure of the power grid,reduce the cost of purchasing power,and both the park and the users can get benefits,which proves the feasibility and effectiveness of the model.