Research On Key Technologies Of Short-term Load Forecasting Based On Big Data

Power system load forecasting is based on the historical data of power load,economy,society and meteorology.It explores the influence of historical data of power load on future load,and seeks the internal relationship between power load and various related factors,so as to make a scientific forecasting of future power load.This paper mainly studies the load forecasting method based on the load data of grid load characteristics of impact load of large steel mills.The load forecasting work in there is different from other load forecasting work.Because the proportion of steel load in the load composition in there is too large and the overall load base is too small,the load curve does not have a smooth trend,but has a large number of sawtooth waves.It is the fluctuation of load curve that aggravates the difficulty of improving the accuracy of load forecasting.The existing load forecasting system in this area only uses historical data for forecasting,without considering the effect of influencing factors on load,and the accuracy of forecasting is not high.Aiming at this problem,this paper mainly combines big data technology with emerging deep learning load forecasting method,and uses big data technology to mine and integrate multi-dimensional influencing factors,uses the advantages of deep learning algorithm to take the multi-dimensional influencing factors as input,and finally improves the accuracy of load forecasting,which proves the adaptability of deep learning algorithm to the load forecasting in there.The main researches are as follows:(1)This paper summarizes the existing load forecasting methods,evaluates the characteristics,advantages and problems of various forecasting algorithms,and analyses the development trend in the future.This paper introduces the application of big data technology in power industry,and shows the reason why load forecasting method based on big data technology is more suitable to the times than traditional forecasting method.(2)This paper introduces the basic principle and model structure of deep learning,which is the basis of the research on load forecasting algorithm of deep learning in this paper.The advantages of deep learning network compared with shallow network are analyzed.The structure and model of deep learning network are summarized and analyzed,including DNN,CNN,RNN and LSTM network model used in this paper.On this basis,the adaptability of deep learning network to load forecasting is clarified.(3)The overall situation of grid load characteristics of impact load of large steel mills power grid is analyzed.The composition of power grid is divided into steel loads and other loads.The load curves of each part are introduced,and the changes of loads in different date types are compared.The load curves of the system and three typical buses are analyzed,and the changing trend and:fluctuation of load curves are analyzed.Finally,the relationship between influencing factors and load changes is clarified by combining the analysis results of load characteristics with external factors.(4)The actual system model is established to verify the big data based load forecasting method proposed in this paper.Firstly,two sets of deep learning network models are constructed according to the actual situation of grid load characteristics of impact load of large steel mills.Secondly,the input and output of the model are processed,and the feature data set is generated.Finally,through the simulation analysis of the system and the load of the six busbars,the adaptation scenarios of the two models are compared,and the simulation results are compared with the existing load forecasting methods of system,as well as the artificial neural network algorithm and support vector machine algorithm.It is concluded that the prediction accuracy of the deep learning algorithm has been greatly improved.(5)Based on the load forecasting algorithm of deep learning network,the software and hardware design scheme of the load forecasting system is proposed.The development and test of the system are carried out based on the scheme,and the applicability of the system in the actual environment is verified.(6)Summarizes the full text of this paper and prospects the work in the later stage.