Data-driven Integrated Energy Forecasting And Research On Optimal Operation Of Integrated Demand Response

With the development of China’s economy,major changes are taking place in the energy production and consumption patterns.At the same time,energy crisis and environmental problems have also been plaguing the development of energy industry.Therefore,the energy transformation is imperative,and the development of a safe,reliable,clean and efficient energy consumption mode has become an inevitable trend.The emergence of regional integrated energy system(RIES)has broken the status quo of independent operation of various energy systems.RIES comprehensively covers various energy systems such as power supply,gas supply,heating supply,cooling supply and electrified transportation in the region,and realizes the coordinated operation of multi-energy sources in source,network and load sectors.Multi-energy coupling increases the uncertainty of energy-using process,which brings new challenges to the accurate and reliable energy prediction of RIES.Meanwhile,the characteristics of multi-energy coupling provide the possibility for the implementation of integrated demand response(IDR)technology.This technology can guide the users to optimize its own energy consumption structure in real time according to the operating state of the system,which is the key to realize the multi-level interaction of the system and tap the potential of user response.Based on the above background,a intergrated energy prediction model is proposed from a data-driven perspective,and on this basis,the optimized operation of RIES considering IDR is studied:As for data-driven intergrated energy forecasting.Firstly,a data-driven RIES prediction model is proposed,and then a unified data analysis and processing method on the supply and demand side is proposed.Finally,it is divided into supply and demand sides for simulation research:On the supply side,taking distributed wind power for example,combining with the historical wind speed and wind power,the XGBoost algorithm is used to forecast wind power.Compared with random forest algorithm,the results show that XGBoost algorithm can reduce the prediction error of wind power prediction better;On the demand side,taking RIES’s multiple load for example,utilizing the coupling of multiple loads and time series memory characteristics of deep long short-term memory(LSTM)network,a combined prediction of multiple loads is proposed,based on deep LSTM network and Adam optimization algorithm.Compared with single load forecasting method and other machine learning algorithms,the results show that the proposed load forecasting model has a lower prediction error and better anti-noise performance.As for optimization model of RIES considering IDR.First of all,a layered interaction framework of IDR is presented,secondly,a park multi-energy flow model is proposed,CCHP,electric heating,electric cooling,multi-energy storage,and IDR model considering multi-energy satisfaction are established.Finally,taking output of RIES energy forecasting model as data support,taking minimazing operation cost of the system as optimal object,and constrained by multi-energy coupling,energy supply and demand,equipment operation,a two-layer economic scheduling model considering IDR is established.Taking an industrial park RIES in northern Chinese as an example,the proposed optimization model was verified by simulation.The influence of IDR technology and electric heating,electric cooling and multiple energy storage equipment on the economic operation cost of the system and the absorption capacity of renewable energy is deeply analyzed.And on this basis,the optimal operation strategy of coupling device is summed up.And then,the optimization effect of IDR on user load curve and the influence of different coupling devices on IDR distribution are analyzed.Finally,the influence of intergrated energy prediction error on the optimization results is explored.