Study On Energy Consumption And Control Optimization Of Chiller Based On Improved LSTM

As a major energy-consuming system in buildings,central air conditioning accounts for more than 30% of the total building energy consumption.Among them,the chiller unit,as the main component of the central air conditioning system,has a large energy-saving space.Therefore,this article predicts and analyzes the energy efficiency of the chiller unit during operation,and achieves energy consumption reduction by optimizing and controlling the operating parameters.This article focuses on the energy consumption analysis of a single chiller unit’s operational principles and proposes an improved long short-term memory network to address the issue of long-term dependence on flat data.Using data collected from the cooling source system of a cigarette factory in Chongqing,an energy efficiency prediction and optimization model is established,which combines operating parameters and environmental factors to predict the energy efficiency of chiller units.To address the coupling between multiple chiller units and the complexity of operation,a combination model is proposed,which combines graph convolutional neural networks(GCN)with long short-term memory networks(LSTM)to extract spatial and temporal features between multiple chiller units.Through this combination model,the energy efficiency of chiller units is accurately predicted.An optimization algorithm based on population mechanism is proposed to address the problems of premature convergence and easily getting trapped in local optima of the traditional tree species optimization algorithm.Furthermore,the search trend ST is linearized on this basis to improve the algorithm’s global search capability and enhance convergence speed,thus achieving a balance between global and local optimization capabilities.For the single chiller,this paper uses the improved LSTM prediction model to predict its energy efficiency,and the error percentage between the actual value and the predicted value is mainly within 3%.The root mean square error(RMSE)was 0.212.Results show that,Under 80% of working condition load,this method can reduce the cooling water flow rate by 2.5% thus achieving energy-saving goals.For the interconnection mode of multiple chiller units,this paper uses the improved tree species algorithm to optimize the load of chiller units,achieve load allocation among multiple chiller units,and provide optimal strategies to reduce the overall load rate.Specifically,the paper uses the tree species algorithm to find the optimal chiller unit load allocation scheme and realizes a balance between optimal load allocation and energy-saving goals by adjusting the operating parameters of the chiller units in real-time.Experimental results show that the improved tree species algorithm proposed in this paper has good performance and application value in load allocation and parameter optimization among multiple chiller units.Compared with traditional methods,this method can better balance energy efficiency and load allocation,improve system performance and energy-saving effects.