Research On Sustainable Development Performance Prediction Of Park-level Integrated Energy Project In Data-scarce Scenarios

In the context of global energy and environmental crisis,China has put forward the goal of sustainable development of "carbon peak,carbon neutralization".Especially in the situation of tight coal supply and multi-site power restriction in 2021,Integrated Energy System(IES)meets the requirements of sustainable energy development.With the development and expansion of IES projects,it is of great significance to strengthen the dynamic prediction of sustainable development performance and analysis of main indicators,as well as dynamic control measures to improve IES project performance,for promoting the sustainable development of IES industry.However,IES project is characterized by multi-agent game,complex technology,integration of heterogeneous energy flow coupling,and lack of performance data,difficulty in collection,uneven distribution of energy types,random interference from power grid faults and other loads.Therefore,it is difficult to carry out performance prediction.This dissertation carries out in-depth research on this issue,firstly establishes evaluation system of IES project sustainable development performance indicators;then,extracts key indicators affecting sustainable performance of the project,in the context of data scarcity,purposefully enhances data,carries out performance forecast of sustainable development of the project,and finally puts forward management enlightenment,with specific contents as follows:(1)Establish IES project sustainable development performance index evaluation system and carry out evaluation.At present,the research on IES project performance evaluation is incomplete,which is different from traditional energy project dimensions such as economy.This dissertation highlights the features of IES renewable energy dissipation,coupled supply and policy support,and defines and calculates five dimensions such as "comprehensive energy utilization efficiency" and 32 indexes;and proposes "IES project sustainable development performance evaluation index"."Monthly Performance Evaluation Label Vector" is used to evaluate the performance of 638 IES projects in Tianjin from January to December 2017.The monthly performance evaluation index is calculated,and then the binary method is applied to form the monthly performance evaluation label value("1" corresponds to the top 50%items,and "0"corresponds to the top 50%items),and the performance database of historical projects is established.(2)XGBoost model is put forward for the first time under the background of data scarcity of IES project,which effectively identifies the key indicators affecting the sustainable development performance of IES project.At the same time,the mechanism of key indicators was studied.At present,there is relatively little research on extracting key indicators of sustainable development performance of IES projects.Under the condition of limited training samples,short training time and lack of samples,this dissertation firstly proposes XGBoost model to extract 9 key indicators affecting sustainable development performance of IES projects.Compared with other methods.XGBoost model has strong generalization ability and controllable model complexity.Major indicators are extracted objectively and accurately.In this paper,ISM(interpretative structural modeling)was used to study the mechanism of key indicators,and an intuitive five-layer multi-order structural model was formed to interpret the optimization path of IES project sustainable development performance.(3)In the situation of IES project performance data scarcity,improved GAN data generation algorithm,which can enhance data targeted,and solved the problem caused by data scarcity.Traditional data reconstruction methods mainly consider the mathematical distribution law of data,easily neglect the regularity of time sequence,correlation and performance change of IES project variables,and data generation accuracy is not enough.In this dissertation,a label value of sustainable performance evaluation is proposed.For conditional value,the GAN of unsupervised network is improved to CGAN of supervised network.The generator learns the mapping relationship between noise distribution and historical data set under predictive conditions.With Wasserstein distance as monitoring index,6220 project data with similar distribution are generated.More performance samples in energy combination mode are added pertinently compared with other methods.CGAN improves the generalization ability of data mining under the background of data scarcity and the correct rate of stability algorithm,and reduces the training and maintenance costs.(4)CGAN-LSTM model is put forward for the first time in the context of lack of IES project performance data,which improves the accuracy and speed of performance forecast for sustainable development of the project.Traditional modeling analysis has been difficult to meet the performance prediction requirements of IES projects.Due to the complex conversion of IES energy coupling,there is a lack of in-depth study on the dynamic mechanism and optimization law of IES energy coupling.In this dissertation,CGAN-LSTM model is first proposed,which can complete long-term memory while deleting current redundant information and be applied to subsequent forecasting to capture the time-domain dependence of input data.Monthly data of 9 key indexes,tag vector of project performance evaluation after one year as output variable,through 12-month historical data set training experiment of 491 IES projects in Tianjin,80 project data sets not used in the training process are validated,and the non-linear relationship between identification of data correlation factor sequence and actual output sequence is found.3.Predict the sustainability performance of the project.Compared with other methods,the model has time adaptability and simple structure.The average response time(ART)of the performance forecast for the same park and year is reduced to 2.73 months and the forecast accuracy is increased to 98.75%.Therefore,management countermeasures are put forward to promote sustainable development performance of IES project and green development of integrated energy industry.