Research On Optimal Scheduling Of Electricity-Cold-Heat-Hydrogen-Fresh Water Island Integrated Energy System Based On Power Prediction

The development and utilization of Marine energy is an important strategic measure for our country to realize "dual carbon" goal and build a powerful maritime country.It is significant to develop an integrated energy system that can improve the island energy utilization structure and promote the clean transformation of coastal energy.A variety of heterogeneous energy sources,such as electricity,cold,heat,and hydrogen,are coupled in the island integrated energy system.To rationally apply these energy sources,the traditional island integrated energy system optimizes energy allocation and achieves basic supply and demand balance through architecture design and scheduling research.However,there are still a number of problems:first,the system architecture is undiversified,unable to adapt to local conditions to use wave energy generation devices,sea water source heat pump,seawater desalination and other offshore energy conversion devices;Second,the system prediction accuracy is not good,and reasonable optimal scheduling requires efficient prediction methods to provide accurate data support.However,the existing methods fail to fully consider the complex multi-dimensional data of offshore power,and the calculation speed is slow and the data information exploration is not comprehensive enough.Third,the system operation efficiency and the scheduling strategy need to be improved.The traditional optimization methods are mostly single-objective,and the algorithm optimization performance is not outstanding,which cannot realize the synchronous improvement of energy utilization and operation economy.Therefore,aiming to make full use of island energy,improve the power prediction accuracy and reduce the system operation cost,this thesis takes the island integrated energy system as the research object and studies the system architecture,power prediction and optimal scheduling of the island integrated energy system.The main research work and innovation of this thesis are as follows:Firstly,in view of the resource insufficient utilization in the existing island integrated energy system,this thesis proposed a multi-energy flow island integrated energy system architecture covering electricity-cold-heat-hydrogen-fresh water.The architecture included a variety of new energy conversion devices,such as hydrogen energy device,sea water source heat pump,seawater desalination,wave energy generation,while realizing clean energy supply at the source side and diversified energy use at the load side.Furthermore,the working characteristics of various devices including the wave power generation device structure and the heat and water generation ability of hydrogen power device were analyzed,and the mathematical models were established,which laid the foundation for the subsequent optimal scheduling research of the system.Then,aiming at the problems of insufficient data exploration,slow convergence rate and low prediction accuracy in the renewable energy power generation prediction of island integrated energy system by traditional prediction methods,this thesis proposed a combined prediction method of convolution neural network and bi-directional long short-term memory neural network based on environmental variables cross-validation.The effective data set was obtained by preprocessing the historical power data.The dimension reduction and feature extraction of the effective data set were completed by using random forest cross validation and convolutional neural network successively.The short data series after dimensionality reduction was used for the bi-directional long short-term memory neural network prediction model training.Compared with the back propagation neural network method and the long short-term memory neural network method,the proposed combined prediction method effectively improved the prediction accuracy and operation efficiency,and thus provided reliable data support for subsequent scheduling.Thirdly,with the goal of improving the system operation economy and increasing the renewable energy consumption rate,combined with the multi-energy flow island integrated energy system architecture built in this thesis,island integrated energy system optimal scheduling method based on deep reinforcement learning was proposed.Fresh water production and supply was regarded as flexible load and was perceived as energy flow to establish a complete electricity-cold-heat-hydrogen-fresh water constraint system.According to the system working characteristics,three optimization schemes were designed and solved by CPLEX,which verified the rationality of the proposed island integrated energy system architecture.To further optimize the system operation and improve the operation efficiency,the deep reinforcement learning algorithm with strong adaptability and fast convergence speed was used to solve the optimal scheduling scheme.The simulation results showed that the proposed method could improve the system economy and renewable energy consumption rate on the basis of ensuring the energy supply and demand balance.Finally,by collecting the part coastal and island data of Qingdao and Weihai in Shandong,the optimizing operation cloud platform for the island integrated energy system was designed and developed.The developed cloud platform was based on Visual Studio Code editor,the Hyper Text Markup Language(HTML)was used as the carrier of visual page content and the MySQL database was used to store and process data,which had the advantages of easy maintenance and convenient operation.The cloud platform mainly consisted of big data module,intelligent display large screen module and source load prediction and strategy operation module.All modules were connected through data interaction and transmission,which verified the effectiveness and operability of the proposed prediction method and optimization method when the project was implemented.