The Method System Of Rapid Identification And Fixed-point Repair Of Abnormal Data In Building Energy System

In the current situation of increasingly severe energy consumption,the utilization of solar energy has received more and more attention,and the PVT heat pump system,as a building energy system that can realize cogeneration,is also being studied in depth.In the early stage of this research group,a new type of PVT heat pump system for combined cooling,heating and power was proposed in view of the existing problems of the existing PVT heat pump,and further research was carried out on its optimal control.Accuracy of sensor measurements is the basis for system control and optimal operation.The PVT heat pump system is in the open air for a long time,and the cold and hot working conditions are switched periodically.The sensor is frequently in contact with the liquid and vapor of the high-temperature and high-pressure refrigerant.The internal and external working environment is relatively harsh,and it is easily damaged.The highly integrated setup makes it difficult for faulty sensors to be replaced.Therefore,calibration and repair based on fault diagnosis of PVT heat pump system sensors is of great significance.In this paper,the virtual in-situ calibration(VIC)method is applied to a PVT heat pump system to fix fixed deviations and accuracy degradation of eight temperature and pressure sensors that affect system operation and control.Firstly,First,combined with the PVT heat pump system model,a variety of fault conditions are introduced according to the variable type and sensor factory accuracy.The feasibility of the virtual in-situ calibration(Model-VIC)based on the mathematical physical model in the PVT heat pump system is verified by the simulation data,and the results are obtained that the system error recovery rate of all sensors is above 90%.In order to verify the calibration effect in the actual process,the calibration is further based on the experimental data,but the average error repair rate is only between 20% and 60%,and there are multiple calibration failures.The accuracy seriously affects the repair effect,indicating that Model-VIC has great limitations in the practical application of PVT heat pump systems.Secondly,given the limitations of Model-VIC,this paper proposes a virtual in-situ calibration method based on sparse autoencoders(SAE-VIC).The sparse autoencoder extracts the features between multi-variables in the process of reconstructing the data,reveals the coupling relationship between sensor networks based on unsupervised learning,and can provide accurate constraint features for VIC calibration without building a model,solving the problem of insufficient mathematical model accuracy.Then,the input variables of the sparse autoencoder are optimized based on the correlation coefficient method,m RMR and Relief F algorithms,and the influence of the number of neural units in the hidden layer and the sparse parameters on the calibration is analyzed.Compared with the Model-VIC average error repair rate of only about 50%,under the SAE-VIC calibration based on experimental data,the error repair rate of a single sensor fault is stable above 80%.However,when SAE-VIC calibrates all sensors at the same time,the repair time is more than 20 times(65 minutes)of a single sensor,and it is difficult to achieve the effect of real-time fault repair.Finally,in order to solve the problem of long time and low repair rate for SAE-VIC to solve multiple migration constants,this paper proposes the FDD-SAE-VIC method.Perform fault diagnosis on the sensor before repairing,locate the faulty sensor,and analyze the influence of multi-sensor faults on the diagnostic accuracy.After the structural optimization of the SAE model combined with Softmax fault diagnosis and fixed-point repair,the error diagnosis rate and fixed deviation repair rate can be increased to about 90%,and the accuracy can also be reduced to near the factory accuracy,while the calibration time is controlled at about 200 s.The method truly realizes the real-time online diagnosis and repair of faults,and reduces the adverse effects of sensor deviation on compressor operation and energy consumption.