Research On Key Technologies Of Autonomous Control For Small Modular Reactor

Small modular reactors(SMRs)are developing towards a higher level of automation and intelligence.Autonomous control is an important trend in the development of SMRs.China has incorporated the research on smart small reactors into the National Energy Development "Thirteenth Five-Year Plan",but the development of the level of intelligence is still in its infancy.The current SMR still needs people to make decisions.Some drastic transient conditions and fault conditions still need to be manually controlled by the operators.Research on autonomous control technology can reduce the need for operators to make judgments,decisions,and manual operations,which reduce the probability of misjudgment and misoperation,the requirements for the number of operators and operating skills,and operating costs.Therefore,it is very necessary to research the key technologies of SMR autonomous control.In this paper,six key technologies for autonomous control of SMR are studied:autonomous control architecture,measured value verification,autonomous decision-making,decision result verification,coordinated control,and intelligent control algorithms,which are verified under fast load-up condition,LOFA,MSLB,LOFA & MSLB accident conditions.Theoretical methods and technical means are provided for the research on autonomous control.In order to transfer operators' functions to machine and realize autonomous control of SMR,a hierarchical three-level autonomous control architecture is proposed,including organization level,coordination level and control level.The organization level replaces the operators' decision-making function.The coordination level replaces the coordination function between multiple operators,and the control level replaces operators' control function.Aim at avoiding the interference of the abnormal measurement value to the autonomous control system,the verification technology of sensor measurement value is studied.A verification scheme for sensor measurement values based on prediction model is proposed,which solves the problem that the abnormal measurement value is difficult to eliminate.Combining BP neural network as the prediction model,the problem that feature parameters are difficult to select is solved.The results showed that the sensor measurement value verification technology accurately predicted the measurement values of various sensors,and it still had good prediction performance when a single input parameter was abnormal.So as to replace the operator to realize autonomous decision-making,the autonomous decision-making technology is studied.The key to autonomous decision-making is the realtime problem.The simple and intuitive multi-attribute utility theory is used to make operation plan decisions,which improves the speed of decision-making.Control goal decision-making is the biggest obstacle to real-time decision-making.A control goal decision-making scheme based on Bayesian optimization algorithm(BOA)is proposed.BOA is used to increase the convergence rate and reduce the number of iterations.The time cost of each iteration is reduced,and the real-time problem of autonomous decision-making is effectively solved.The kernel function,acquisition function,preset observations,and number of iterations of BOA are optimized.The results showed that BOA had fast convergence speed,high optimization accuracy,and real-time decision-making.So that realize the real-time verification of decision results,the decision result verification technology is studied.The machine learning method is used to predict the change trend of key operating parameters in the transient process from the initial working condition to the objective working condition,which can ensure the real-time verification of the decision result.The lack of input data is a key factor that affects the prediction performance.A decision result verification method based on a random forest prediction model is proposed.This calculation method based on classification ideas effectively avoids the i MPact of input data linearization.The K-nearest neighbor and multi-layer perceptrons are used as co MParison models.The results showed that the random forest model had the best prediction performance.For achieving rapid and smooth transition under drastic transient conditions and fault conditions,coordinated control technology and intelligent control technology are studied.A coordinated control strategy is proposed,which draws on the idea of combining manual control and automatic control of traditional nuclear power plants.The coordinated control strategy solves the problem that a single control algorithm cannot meet the control requirements.The operator's operating experience is condensed into fuzzy rules,and the fuzzy logic method is used for fuzzy reasoning,which simulates the manual / automatic control switching process of the operator.Combined with the idea of predictive control,a MLP-PSO controller is proposed.Simulating manual control by the operator,a fuzzy controller is proposed.These two controllers are used to achieve fast control,which solves the problem of slow control at the initial stage of drastic transient conditions and fault conditions.Aiming at the problem of floating control in the later stage of the transient process,a BPNN-PID controller is proposed.The PID coefficients are optimized online through the BP neural network to improve the control performance.The results showed that the coordinated control technology and intelligent control technology proposed in this paper effectively suppressed the overshoot of the controlled quantity,and quickly mitigate the i MPact of huge disturbances under drastic working conditions,so that the SMR can quickly and smoothly transition to the objective working condition.For fast load-up conditions,LOFA,MSLB,LOFA & MSLB accidents,the autonomous decision-making technology,decision result verification technology,coordinated control technology,and intelligent control technology were verified.Simulation results showed that the key technologies of autonomous control proposed in this paper had good decision control performance and fault tolerance.The research and application of the key technologies of autonomous control in this paper have important reference value and practical significance for the future design of autonomous control systems for SMRs.