基于云-边协同求解的智慧综合能源园区多能协调碳排放优化模型OA

This paper proposed a multi-energy coordinated carbon emission optimization method for smart integrated energy parks based on cloud-edge collaborative solution to address the issue of difficulty in achieving low-carbon operation within the park due to the distribution of information technology processes. Firstly,a dynamic carbon reduction model for cloud-edge collaboration in smart parks is established based on carbon emission flow theory,taking the uncertainty of equipment response in the system into account.By establishing affine functions and setting noise elements,the uncertainty of renewable energy output and load response can be quantified,and the uncertainty can be embedded into the objective function to improve the stability of cloud edge collaborative solution. Then,to dynamically solve the carbon reduction plan for smart parks,this paper establishes a cloud-edge collaborative solution network based on federated transfer learning theory,realizes parallel training of the cloud edge network,and accelerates the model training process.By combining the convolutional block attention module(CBAM)mechanism,particle swarm optimization(PSO)algorithm,and bidirectional long short-term memory(BiLSTM)model,and utilizing the key feature perception ability of CBAM,the training efficiency of BiLSTM model can be improved.The PSO algorithm is used to assist the neural network in optimization,and the training process of the model is guided by heuristic algorithms,avoiding the problem of the neural network getting stuck in local optima and convergence difficulties.Enable the established CBAM-BiLSTM-PSO cloud-edge collaborative solution network to simultaneously achieve new energy power output prediction,multi-energy load prediction,and the solution and optimization of low-carbon scheduling plans. During the operation of the CBAM-BiLSTM-PSO cloud-edge collaborative solution network,the operation data of the smart park is input through a convolutional network,and feature extraction is achieved through the CBAM module.The extracted feature data is input into the BiLSTM-PSO network of the cloud and edge nodes.Only one BiLSTM-PSO network is set up at the edge nodes of renewable energy to achieve renewable energy power output prediction.Two BiLSTM-PSO networks are set up at the edge nodes of multi-energy loads to achieve load energy consumption plan prediction and optimization,respectively.Set up one BiLSTM-PSO network in the cloud node to achieve iterative solution of scheduling plans.The model mainly includes five parts:feature extraction,source load power prediction,initial scheduling plan solving,load energy consumption plan optimization,and scheduling plan optimization. Finally,the effectiveness of the method proposed in this paper is verified by taking the actual integrated energy system as an example.The results show that the CBAM-BiLSTM-PSO network established in this paper effectively improves the training efficiency of the prediction model and the accuracy of model prediction,and significantly improves the prediction performance of the traditional single model and combination methods.Moreover,after optimization,the carbon emission reduction demand of the park is jointly borne by the source side and the load side,reducing the system regulation pressure.Compared with single cloud computing and distributed computing,it has better optimization effect and shorter training time.Moreover,the carbon emissions of the park are reduced by 8%after optimization,which verifies the effectiveness and superiority of the proposed method in the low-carbon operation of IES.