Low-Carbon Economic Dispatch Of The CCHP-VPP Considering Electricity Price-Load Forecasting

To achieve the environmentally-friendly national strategic goal of "peak carbon dioxide emissions and carbon neutrality",the traditional electric power industries should undergo reforms to cope with the multiple challenges of decarbonization,marketization,and energy transition.Designing a dispatch strategy that considers both low-carbon demand and economic cost has become a major concern in integrated energy systems.And traditional power systems exist sourceload uncertainty and regulation technology defects when faced with multi-energy distributed generation units,customer-side demand response mechanisms,and energy storage system scheduling control.Accordingly,this paper proposes an innovative concept of combined cooling,heating and power-virtual power plant(CCHP-VPP),and aims to achieve an optimal low-carbon economic dispatch strategy by combining the day-ahead electricity price and load forecasting technologies simultaneously.This research work has good theoretical references and engineering application values.Considering the interactive operation mechanism of "source-grid-load-storage",an improved system architecture of the CCHP-VPP covering power-to-gas,carbon capture,and day-ahead electricity price and load forecasting is proposed.And the operating rules for the energy and information flows of the CCHP-VPP system are described.The operating characteristics of the sub-modules in CCHP-VPP are analyzed one by one and their power mathematical models with linear variable parameters are built,including multiple types of renewable energy generating units(wind,photovoltaic,and hydropower),boiler units,refrigeration units,customer-side flexible loads,carbon capture systems and power-to-gas systems.Because CCHP-VPP has not yet realized accurate forecasting of high precision day-ahead electricity prices and loads,a novel hybrid deep learning-based model based on convolutional neural network+stacked sparse denoising auto-encoders is proposed.And the complete ensemble empirical mode decomposition with adaptive noise,an improved decomposition method,is introduced to enhance model performance by the decomposition of complex data sequences.The experimental results verify that the proposed hybrid model can effectively improve prediction accuracy and stability,which has a faster convergence speed.Considering the sequential decision characteristics of low-carbon economic dispatching for CCHP-VPP,this type of stochastic dynamic optimization problem is mathematically simplified and described as a Markovian decision process.A low-carbon economic dispatch optimization model based on deep reinforcement learning algorithms for CCHP-VPP is constructed in this work.Then the action space,state space,transfer probabilities,constraints,and multi-objective optimization(reward)functions are designed.For ameliorating local optimum in agents caused by sample selection of prioritized experience replay,a novel deep reinforcement learning-based approach named multi-level backtracking prioritized experience replay-twin delayed deep deterministic policy gradient is developed.The experimental case proves that the proposed algorithm can make a set of scientific and reasonable low-carbon economic dispatching decisions of power generation-distribution-sale-supply-storageconsumption under the random fluctuation of external environmental parameters(forecasted electricity prices and electric loads,natural gas prices,cooling and heating loads and distributed power generation output).The optimal strategies can realize the "source-grid-load-storage" interactions and the strategic goal of "peak carbon dioxide emissions and carbon neutrality".And the dispatching results can attain the low-carbon and economic benefits of maximizing the profit income,minimizing carbon emission,and maximizing the utilization of renewable energy in CCHP-VPP.