Distributed Dispatching Of Active Distribution Network Based On Short-Term Load Forecasting

With the further deepening of power reform,the traditional integration of power transmission and power sales has been broken.Many power sales companies have appeared participated in the power market.Power companies have begun to have power purchase and sales attributes.How to achieve a balance between the economics of grid operation and the power demand of electricity companies under the condition of ensuring reliability and safe and stable operation of the power grid has become an urgent problem to be solved.At the same time,since the electricity market under the current electricity reform process still uses the main mechanism of medium and long-term market transactions and spot market transactions,accurate short-term load forecasting can provide an important reference for the safe operation of power grids and the purchase and sale plans of electricity sales companies.The active distribution network system of a large number of new power sales companies,accurate and rapid load forecasting and a scheduling model that takes into account the conflicts between various stakeholders are the key to solving the current problems.The main research of this research is to solve the problem of load forecasting and scheduling under active transfer network.The specific research content is as follows:(1)Preprocessing of load data.First of all,the collected data is identified,eliminated,and supplemented to eliminate the influence of abnormal data;second,the data is normalized and one-hot encoding processing to eliminate the influence of inconsistent dimensions and variable types;finally,it will affect Load forecasting accuracy factors such as temperature and humidity are analyzed to determine the degree of influence,and the daily load curve,weekly load curve,and monthly load curve is analyzed to explore the periodicity and regularity of the load curve.(2)SOFM clustering electricity consumption analysis based on load feature extraction.First,the data after load preprocessing is de-redundant through the RST algorithm,and the de-redundant data is decomposed by CEEMDAN to calculate the sample entropy and then the load feature is extracted;finally,the data is sent to the SOFM network for cluster analysis,and the classification results.As the input vector of the cross entropy loss function in the load forecasting model.(3)Attention-LSTM load forecasting model based on combined loss function.First,the Attention mechanism is introduced in the LSTM network to ensure that the input factors can be given reasonable weights;second,the cross-entropy loss function is combined in the mean square error loss function as the loss function for model optimization,where the mean square error loss function can effectively ensure the direction of convergence,The cross-entropy loss function can improve the accuracy of the probability distribution stage model;finally,the model built in this paper is compared with the LSTM prediction model and the Attention-based LSTM prediction model for simulation experiments,verifying that the short-term load prediction model established in this paper can effectively improve the prediction accuracy and speed.(4)Active distribution network dispatching considering the interests of electricity sales companies.First,establish a game model for the problem of coordinating the economics of the power grid and the power conflict of the electricity sales company,give priority to the use of renewable energy,reduce the communication volume and improve privacy through distributed computing,and establish a partitioned architecture between the distribution system operator and the virtual micro-grid Secondly,based on(3)medium load forecasting data,the distributed dispatching model of active distribution network is used,and the solution is solved by the alternating direction multiplier method;finally,the effectiveness of the model is verified by the simulation experiment of adjusted IEEE33 node.