A New Fdd Method For Centrifugal Chillers Based On Nonlinear Models

The chiller systems have played an increasingly important role in people’s life and work.However,as a major component of building energy consumption,it also greatly increases the energy consumption of the building.According to statistics,the energy consumption of chiller accounted for 40% to 50% of the total energy consumption of air conditioning systems.When the chiller fails,not only affect the safe operation of the unit and the thermal comfort of the indoor space,but also increases the energy consumption of the entire system,resulting in energy waste.Thus,real-time monitoring chiller running,and using reliable fault detection and diagnosis(FDD)strategy to prevent the occurrence of failure and ensure the efficient operation of the chiller has always great practical significance.In the past decades,a number of chiller fault detection and diagnosis methods have been conducted.However,the improvement of FDD methods for centrifugal chillers is still required.There are two types of faults in chiller as abrupt faults and incipient faults.Since the abrupt faults can be easily detected,the main objective of this thesis is to develop robust and enhanced FDD methods suitable for the detection and diagnosis of incipient faults.In implementing of this thesis consists of a comprehensive literature review and the methodologies of three proposed FDD strategy.There are basically three tasks to be carried out:First,this thesis develops an online fault detection and diagnosis strategy based on nonlinear radial basis function(RBF)to online detect and diagnose faults of centrifugal chillers.In this task,the RBF is adopted to develop the reference feature parameter(FP)models.The benchmarks of the feature parameters are also provided by their corresponding reference models.Then,the residual of each FP,which is the difference between the current FP and its benchmark value,is calculated.The Exponentially-weighted moving average(EWMA)residual control charts of each FP is used to detect the faults.A rule-based diagnostor is developed to online identify the fault.Seven common faults are taken in account for typical centrifugal chillers.The FDD strategy proposed was validated by using the experimental data of incipient faults from the ASHRAE RP-1043 project and the operating data of a centrifugal chiller in an office building of Hong Kong.The test results show that the RBF-EWMA method has achieved significant improvements in accuracy and reliability by comparing with the previous method with SVREWMA.The proposed RBF-EWMA method is robust for fault detection and diagnosis in centrifugal chiller systems.Second,a successful chiller FDD strategy is highly dependent on three core aspects,a perfect reference model,a sensitive fault detection technique and an effective fault identification method.A perfect reference model play an important role in the performance of chiller FDD strategy.Therefore,in this task,the accuracies of three models,MLR(multiple linear regression),KRG(kriging)and RBF(radial basis function)are tested and compared under various modeling criteria.Through comparisons,the most accurate one for chiller FDD strategy is found.Then,it is suggested to apply to the fault detection and diagnosis strategy in chiller systems.The EWMA residual control chart,which is a more powerful tool in detecting small shifts than t-statistics based,is also adopted as a sensitive fault detection technique.The most sensitive feature parameter method,which is based on both the direction and the magnitude changes of the feature parameters related faults when the fault occurs,is developed to improve the performance of identify faults.Seven common faults in typical centrifugal chillers are considered.Three strategies(MLR-EWMA,KRG-EWMA and RBF-EWMA)which are combinations of three reference models with EWMA and the most sensitive feature parameter method are validated and compared by using two sample datasets with different sizes from ASHRAE RP-1043.Through the comparisons of diagnosis performance,the most perfect chiller reference model is found and suggested to detect and diagnose faults in chiller systems.The results present that the performances of RBF-based strategy has the best diagnosis performances among the three strategies.Third,developing the fault detection and diagnosis(FDD)for centrifugal chiller systems is very important for improving the equipment reliability and saving energy consumption.Due to the accuracy of a reference model is difficult to build.The results of model-based FDD are strongly dependent on the accuracy of chiller models.The accuracy of the chiller models depends on the indefinite model parameters normally chosen with experience.Therefore,optimization of model parameters is very useful to increase the accuracy of chiller models.This paper presents a new fault detection and diagnosis method for centrifugal chillers of building airconditioning systems,which is a combination between the nonlinear least square support vector regression(LSSVR)bases on the DE algorithm and the exponentially weighted moving average(EWMA)control charts for chiller fault detection and diagnosis.In this new strategy,the nonlinear LSSVR,which is a reformulation of SVR model with better generalization performances,is adopted to improve the accuracies of reference feature parameters models in a typical non-linearly chiller system.The differential evolution(DE)algorithm which is a realcoding optimal algorithm with powerful global searching capacity is employed to enhance the LSSVR model precision.The exponentially weighted moving average(EWMA)control charts are introduced to improve the detection faults capability as well as to reduce the Type II errors in a t-statistic-based way.Six classical faults of the centrifugal chillers from the real-time experimental data of ASHRAE RP-1043 project are chosen to validate proposed FDD methods.To validate proposed method,we make comprehensive comparisons between proposed method and two similarly previous studies which were the open literatures for FDD in chillers.The comparison results show that the proposed method has achieved significant improvements in accuracy and reliability,especially at low severity levels(i.e.at severity level 1 of increasing order four severity levels from severity level 1 to severity level 4).For example,in the case of condenser fouling,the proposed method obtained the correct diagnosed rate of 16.67% at severity level 1 at the confidence level of 99.73%,while the standard SVR-EWMA and RBFEWMA was only of 7.7% and 0% at severity level 1 respectively at the confidence level of 99.73%.