Research On Parameter Identification Method Of Physical Model For District Cooling System

Due to the large scale and complex structure of the district cooling system,its optimal operation and fault detection and diagnosis have always been challenging problems.The establishment of accurate and reliable digital twin model of district cooling system can provide decision support information for the efficient operation and maintenance of the system,and it is also an important prerequisite for realizing the optimal operation of the whole system and fault diagnosis.The establishment of digital twin model depends on model calibration.At present,a large number of studies have been carried out in the academic circle on the model calibration of district cooling system.However,most of these methods are aimed at simple single model.The research on the undetermined methodology of the model considering the nonlinear dynamic coupling of all single models at the whole system level is not perfect,and the attenuation of the actual system performance over time is not fully considered.Parameter identification is often used in model calibration,but the current method has limited application scope.When the model has high-dimensional parameters or the calculation process is discrete,it is often difficult to ensure the convergence performance of the parameters to be identified.In addition,the insufficient number of sensors also leads to insufficient prior information of the model,which reduces the reliability of the simulation results.To overcome the above problems,this study made the following innovations :(1)Aiming at the problem of poor parameter convergence caused by high dimensional parameter identification of district cooling system model,a model parameter identification method based on quantum genetic algorithm is proposed.This method greatly improves the diversity of parameter solution space in the process of parameter identification by encoding state vectors,overcomes the problem that high-dimensional parameters are difficult to global convergence in the process of parameter identification,and the model accuracy is higher than that of traditional genetic algorithm.In addition,the population updating method of quantum genetic algorithm realizes faster calculation and judgment in the optimization process,and improves the computational efficiency of parameter identification of complex models.(2)Aiming at the problem of high uncertainty of model parameter identification results caused by insufficient prior information due to insufficient number of sensors,a parameter posterior estimation method based on Bayesian optimization algorithm is proposed.This method can establish the probability relationship between the empirical knowledge and the unknown parameters of the model under the condition of insufficient or highly uncertain information of the actual district cooling system,and generate the posterior estimates of the missing measuring points to supplement the data in the parameter identification process,quantify the uncertainty caused by insufficient information to the model parameter identification,and improve the reliability of the simulation results.Based on the theoretical model of district cooling system and practical engineering system,the above methods are verified respectively.The results show that for the district cooling system model with discrete calculation process,quantum genetic algorithm can achieve reliable convergence of high-dimensional undetermined parameters.On the theoretical model observation dataset,the error between the parameter values optimized by quantum genetic algorithm and the real value is less than 12 %,and the simulation accuracy of the identified model can reach 98 %.The uncertainty of the simulation results optimized by Bayesian algorithm is reduced by 2 % ～ 7 %,which further improves the robustness of the model.On the actual system operation data set,the relative error of flow simulation is less than 5 %,the accuracy of chiller model is higher than 96 %,and the accuracy of surface cooler model is higher than 97 %.The above verification shows that the parameter identification method proposed in this study can significantly improve the accuracy of the district cooling system model,reduce its uncertainty,and ensure the reliability and effectiveness of the model.To sum up,this study provides a new idea for establishing accurate and reliable digital twin model of district cooling system,provides important technical support for ensuring accurate and efficient operation and maintenance of district cooling system,and has important practical significance for promoting intelligent management of district cooling system.