Mid-long Term Load Forecasting Based On External Information Aggregation Technology

The mid-long term load forecasting is the foundation in power system planning.The power system is affected by many factors and the relationships among them are complicated,which has brought great difficulties in mid-long term load prediction.When it comes to dealing scenes with multiple factors,the information aggregation technology is often employed to carry out load forecasting.The current information aggregation method has mainly integrated with the internal information from the power grid enterprises,while paying few attention to the information outside.The lack of centralization and integration of the external information may lead to mismatching between the development planning and the actual demand for electricity.Meanwhile,the cumulative effect exists widely in time series.In the data mining algorithm without considering the cumulative effect,a static similarity algorithm is often applied to mine data which are similar to target parameters and other influencing variables from historical information.Such methods without considering the correlation among each factors may result in large deviation.In this paper,a mid-long term load forecasting method based on external information aggregation technology is proposed.This paper is organized as follows:(1)The influence rule of cumulative effect is analyzed,while the dynamic character of time series change is also revealed in the paper.Furthermore,a cumulative effect recognition method based on multivariate fitting degree is proposed.The proposed method chooses the convergence point of fitting goodness to determine whether the cumulative effect exists according to the actual situation.The algorithm is implemented by the example of short-term load forecasting in summer.The verification results show that the algorithm can effectively recognize cumulative effects.(2)The information aggregation method for mid-long term load forecasting is analyzed in this paper.The proposed method has considered all kinds of indicators that affect the prediction of power load demand after carrying out the correlation analysis.Then,through clustering analysis,correlation degree is used as a feature index to filter and cluster data.Finally,principal components analysis is applied to aggregate multiple economic index into a principal component.In this chapter,the process of information aggregation which lays the data foundation for the example analysis in the next chapter is discussed in detail.(3)Based on the principal component obtained from the information aggregation technology which has considered the cumulative effect,the first order Takagi-Sugeno structure is used to predict the annual total electricity consumption.The annual increment and annual growth of the annual principal component data are used as the input data of the model.The annual total electricity consumption is used as the output,and the membership function of the input variables is fuzzed,and the learning and training functions of the neural network are repeatedly corrected.In order to verify the performance of the algorithm,two sets of control experiments are set up in this chapter.The experimental results show the accuracy and effectiveness of the proposed method.In summary,this paper considers the impact of external grid data on power load,establishes a cumulative effect recognition method,analyzes each index separately,and proposes an information aggregation method that considers cumulative effects.Examples show that the method proposed in this paper improves the accuracy of mid-long term load forecasting.