Power Load Forecasting Based On Multi-Source Data Hierarchical Association Analysis And Optimized Neural Network

Power load forecasting plays an irreplaceable role in the national energy distribution and national economic development.The power system department can make reasonable arrangements for the transmission and distribution of power grid according to the forecast results,and then make reasonable planning and resource allocation for municipal living electricity,industrial production,etc.,so that social production can be carried out efficiently.Based on the analysis of power load classification and basic characteristics,this paper proposes a power load forecasting method based on outlier detection,multi-source data hierarchical association analysis and group adaptive beetle algorithm optimization neural network.Firstly,for the noise,redundancy and load exception types of power load multi-source data,the outlier detection method based on the Shortest Fork Tree(Shortest Forking Tree,SFT)and mixed indicator local anomaly factor(Mixed Local Index Abnormal Factor,MDILAF)is proposed,The comparative experiment of six outliers detection algorithms(LOF,LDOF,LOCI,COF,LMDOF,RDOS)proves that SFT can replace the traditional K neighborhood information method and effectively eliminate k value sensitivity,and proves that MDILAF recognizes outliers better than other algorithms.Secondly,in order to reduce the complexity of the raw load data and accurately analyze the correlation between multi-source influencers and power load,the Multi-source Data Hierarchical Correlation Analysis method(MSDHCA)combining Dual-Tree Complex Wavelet Transform(DTCWT)and Grey Relational Analysis(GRA)was constructed.Among them,DTCWT has excellent characteristics such as translation invariance,suppression of frequency aliasing,and approximate analysis.The original power load sequence can be transformed to obtain subsequences belonging to different layers,which reduces the complexity and instability of the original data,and overcomes the modalities Aliasing.Considering the influence of meteorological factors on short-term power load forecasting,GRA is used to analyze the meteorological influence factors and load subsequences,reduce the dimension of the data,and extract the main features of meteorology as meteorological influence factors.Finally,in order to solve the problem that BP neural network is easy to fall into local minimum,a group adaptive algorithm based on beetle algorithm(Group Adaptive Beetle Antennae Search,GABAS)is used to optimize the parameters of the neural network and combine the first two methods to build a prediction model.The YangZhong data is used to establish a prediction model for experiments.The results show that the method proposed in this paper can effectively improve the prediction accuracy.