Research On Investment Forecast Of Wind Power Project Based On Deep Learning Neural Network

As a clean energy with continuous improvement of technology level,development quality and consumption utilization,the development of wind power is very import to alleviate the energy crisis and environmental pollution caused by China’s rapid economic growth.In this context,improving the level of wind power project investment management plays an important role in ensuring the sustainable and healthy development of wind power.However,in the whole process of wind power project investment,investors focus on the implementation stage of the project construction,and pay less attention to the investment control in the early stage of the project construction.And the wind power project investment forecasting is the key point in the decision-making stage.Its impact on the project cost reaches 70%to 90%,which directly determines the investment management effect of the whole wind power project.Therefore,the research of wind power project investment forecasting method has important practical significance.First of all,based on reviewing a lot of previous related research,this thesis summarized the research of domestic and foreign scholars on project investment forecast,the influencing factors of wind power project investment forecast and neural network in forecasting field,and introduced the basic theories and forecasting methods of wind power project investment in detail.Then,based on the date space of a consulting enterprise,by analyzing the wind power project structure,the influencing factors of wind power project investment forecast were comprehensively extracted from two aspects of relevant literature researches and actual engineering data,and the preliminary selection process of influencing factors of wind power project was completed.On the basis of the expert experience,the fuzzy threshold method was used to preliminarily screen the influencing factors of wind power project before data collection,which reduced the difficulty of data collection.After the completion of data collection,the random forest algorithm(RF)was proposed to extract the key influencing factors of the wind power project to ensure the effectiveness and simplicity of the model input variables.Next,this thesis put forward a wind power project investment forecasting model,which was the combination of convolution neural network and support vector machine(CNN-SVM).The combination model combined the advantages of CNN and SVM,and had good performance in model applicability and prediction accuracy.Finally,the empirical research was carried out.In this process,298 wind power project data was collected,the training of CNN-SVM was completed after the data preprocessing work,and the wind power project investment forecasting was realized based on CNN-SVM.Meanwhile,the model without identification of key influencing factors,the single forecasting model and BP model were selected to forecast the wind power project investment.The results of each model were compared and analyzed,which proved that the RF-CNN-SVM model had strong applicability and superiority.Based on the analysis of influencing factors of wind power project investment,this thesis proposes the screening method of feasible influencing factors and the identification method of key influencing factors,and combines CNN and SVM to get CNN-SVM model.It provides a new forecasting method for wind power project investment,and also provides an objective basis and reference for wind power project investment decision-making.Moreover,the model established in this thesis offers a good reference for the investment forecasting of other engineering projects.