Research On Non-gaussian Stochastic Control For Organic Rankine Cycle Systems

Throughout history,all the activities of human society cannot live without energy,With the increasing gradually energy consumption,at the same time,the environment problem is increasingly rigorous,solve the problem of energy is becoming the top priority in today's social development.Improving energy efficiency is the key to solve the energy problem.The data reveal that China's industrial waste heat resources are abundant.Among them,the low-temperature utilization of waste heat energy has great potential,low-temperature waste heat utilization plays an important role in reducing industrial operating costs,improving energy efficiency,and reducing pollution emissions.It plays a decisive role in China's economic development and human living environment.Therefore,in recent years,low-temperature waste heat utilization technology has emerged.Among them,an organic high-energy low-temperature residual heat technology,the Organic Rankine Cycle(ORC),has become a focus of attention in many fields.Compared with the traditional steam lanken cycle,there are many advantages.However,the Organic Rankine Cycle system affected by the external random disturbances inevitably,such as the mass flow rate of flue gas,the temperature of the heat source,etc.,and these random disturbances are all non-Gaussian.The evaporation temperature is one of the important factors in the ORC process.In terms of safety and economy,the evaporation temperature at the outlet of the evaporator should be controlled within a reasonable range.Thus,based on the existing non-Gaussian stochastic system control methods,this project is based on statistical information and further studies the temperature tracking control problem affected by non-Gaussian noise in the framework of information theory.The methods used in this paper are as follows:First,a single neuron controller design method based on Survival Information Potential(SIP)is proposed.A new statistical information—SIP,which can comprehensively characterize the probabilistic properties of nonGaussian random variables—is introduced.In consideration of the tracking error and randomness of control inputs,a new control optimization criterion for nonGaussian random systems is constructed.In the controller selection,a single neuron controller with low computational cost and simple structure is used.Based on the optimization criteria,a simple and feasible,easy to implement optimization method— the gradient method,is chosen to design the optimal control law for a non-Gaussian random system.Then,on the base of the above work,the cumulative SIP based on tracking error and control input is used as the performance index of the control system,and it is possible to effectively handle multi-variable,large delay,non-minimum phase,unstable pole and constraints,etc.The Model Predictive Control(MPC)strategy further proposes a single neuron predictive control strategy for a nonlinear nonGaussian stochastic control system.By minimizing the performance index,the weights of the single neuron controllers are updated,and further analyzed the convergence of the control algorithm.In addition,aiming at the problems of heavy-tailed distribution non-Gaussian noise and control input constraints in organic Rankine cycle systems,an optimal tracking control strategy based on the correntropy criterion is proposed,and based on Kullback-Leibler criterion,the method of adaptive updating of the kernel width in the correntropy with the control process further improves the accuracy of the tracking control.In conclusion,the research on the temperature tracking control method of organic Rankine cycle under non-Gaussian noise not only enriches the theory of non-Gaussian stochastic system control and has certain theoretical significance,but also can be applied to general industrial processes,and has practical implications.