Central air conditioner fault diagnosis model migration method
By performing simulation modeling and fault simulation of the central air conditioning system, the target domain and source domain data are obtained, and the pre-trained CNN_1D model is fine-tuned through transfer learning method, the problem of migration of central air conditioning fault diagnosis model is solved, and effective in the target building is achieved Troubleshooting.
Patent Information
- Application Number
- CN202411495277.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2025-02-14
AI Technical Summary
In the prior art, the migration method of central air conditioner fault diagnosis model cannot effectively apply the knowledge or model learned by existing models on a certain building to different but related buildings, resulting in a lack of fault data on the target building and it is difficult to establish an effective fault diagnosis model. .
A central air conditioner fault diagnosis model migration method is proposed. By simulated modeling the target building and the central air conditioning system of the source building, the target domain data and source domain data are obtained respectively, and preprocessed, and divided into training Datasets, validation datasets, fine-tuning datasets, and test datasets. Then, by freezing all or part of the convolutional layers of the CNN_1D model of the pre-trained one-dimensional convolutional neural network, the model is fine-tuned and a fault diagnosis migration model is established.
Through this method, a fault diagnosis model can be effectively established in the target building, solving the problem of lack of fault data in the target building, and realizing the application of knowledge or patterns of the source building to the fault diagnosis of the central air-conditioning system of the target building.
Smart Images
Figure CN119443204A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of air-conditioning fault diagnosis, and in particular to a central air-conditioning fault diagnosis model migration method. Background Art
[0002] According to statistics, the energy consumption of central air-conditioning systems accounts for about 30% to 60% of the total energy consumption of large public buildings. The normal and stable operation of central air-conditioning systems can provide building users with a hygienic, comfortable and convenient living and working environment, and a high-efficiency and high-stability air-conditioning system fault diagnosis model is becoming increasingly important.
[0003] In recent years, with the rapid development of artificial intelligence, some scholars have introduced deep learning methods into the field of building air conditioning system fault diagnosis, but most of them rely on high-quality data and labels in the same field (same data set), and the application limitations in cross-domain scenarios are gradually becoming prominent. The contradiction between massive data and little labeled data makes these methods only applicable to certain specific occasions. It is difficult to establish an efficient fault diagnosis model for air conditioning systems that lack labeled fault data.
[0004] At present, in the central air-conditioning fault diagnosis model migration method, the target building fault data is scarce, it is difficult to establish an effective fault diagnosis model, and it is impossible to apply the knowledge or patterns learned from the existing model in a certain building to different but related buildings. Summary of the invention
[0005] The present disclosure proposes a central air conditioning fault diagnosis model migration method to solve the technical problem that the central air conditioning fault diagnosis model migration method in the prior art cannot apply the knowledge or pattern learned from the existing model in a certain building to different but related buildings.
[0006] In order to solve the above technical problems, the present disclosure proposes a central air conditioning fault diagnosis model migration method, including:
[0007] Obtain a target building and perform simulation modeling on the target building. The calculation formula is as follows:
[0008] NMBE = (average value of simulation values - average value of measured values) / average value of measured values
[0009] C v (RMBE) = RMSE / average of measured values × 100%
[0010]
[0011] Where, NMBE standard mean offset error, represents the simulated value of building energy consumption at the i-th time point; y irepresents the actual measured energy consumption of the building at the i-th time point; n is the number of samples in the dataset;
[0012] Acquire a source building, and simulate a fault or anomaly of a central air-conditioning system of the source building, wherein the fault or anomaly of the air-conditioning chiller includes refrigerant leakage, overcharging of refrigerant, scaling of the condenser, scaling of the evaporator, the presence of non-condensable gas in the chiller, too little cooling water flow, and too little freezing water flow;
[0013] The calculation formula for the refrigerant leakage fault of the air-conditioning chiller is as follows:
[0014] COP REF1,f =COP REF1,n ×F REF1
[0015] Among them, COP REF1,f Represents the first coefficient of performance of the chiller under fault conditions, dimensionless; COP REF1,n It represents the first energy efficiency coefficient of the chiller under normal working conditions, dimensionless; F REF1 It represents the first failure coefficient in this scenario, describing the reduction degree of chiller performance caused by the refrigerant leakage of the chiller;
[0016] The calculation formula for the overcharge fault of the chiller is as follows:
[0017] COP REF2,f =COP REF2,n ×F REF2
[0018] Among them, COP REF2,f Represents the second coefficient of performance of the chiller under fault conditions, dimensionless; COP REF2,n It represents the second energy efficiency coefficient of the chiller under normal working conditions, dimensionless; F REF2 It represents the second fault coefficient in this scenario, describing the reduction of chiller performance caused by overcharging of chiller refrigerant;
[0019] The calculation formula for the scaling fault of the chiller condenser is as follows:
[0020] K con,f =K con,n ×F con
[0021] Among them, K con,f = indicates the heat transfer coefficient of the chiller condenser under fault conditions, in W / m 2 ·K; K con,n Indicates the heat transfer coefficient of the chiller condenser under normal operating conditions, in W / m 2 ·K; F conDescribe the extent to which condenser scaling reduces the heat transfer coefficient of the chiller condenser;
[0022] The calculation formula for the scaling fault of the chiller evaporator is as follows:
[0023] K eva,f =K eva,n ×F eva
[0024] Among them, K eva,f = indicates the heat transfer coefficient of the chiller evaporator under fault conditions, in W / m 2 ·K; K eva,n Indicates the heat transfer coefficient of the chiller evaporator under normal operating conditions, in W / m 2 ·K; F eva Describe the extent to which evaporator scaling reduces the heat transfer coefficient of the chiller evaporator;
[0025] The calculation formula for the non-condensable gas operation fault of the chiller is as follows:
[0026] P gas,f =P gas,n ×F gas
[0027] Among them, P gas,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P gas,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F gas It represents the failure coefficient in this scenario, describing the increase in the energy consumption of the chiller caused by the presence of non-condensable gas in the chiller;
[0028] The calculation formula for the failure of the chiller to operate due to insufficient cooling water flow is as follows:
[0029] P cw,f =P cw,n ×F cw
[0030] Among them, P cw,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P cw,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F cw It represents the failure coefficient in this scenario, describing the increase in energy consumption of the chiller due to insufficient cooling water flow of the chiller;
[0031] The calculation formula for the chiller unit's operating fault due to insufficient chilled water flow is as follows:
[0032] P chw,f =P chw,n ×F chw
[0033] Among them, P chw,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P chw,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F chw It represents the failure coefficient in this scenario, describing the increase in energy consumption of the chiller due to insufficient chilled water flow of the chiller;
[0034] According to the simulation modeling and the fault or abnormality simulation, target domain data and source domain data are obtained respectively, and the target domain data and the source domain data are preprocessed to obtain a training data set, a verification data set, a fine-tuning data set and a test data set, wherein the training data set and the verification data set are respectively used to train and verify the performance of the pre-trained one-dimensional convolutional neural network CNN_1D model in the fault diagnosis of the central air-conditioning system of the source building, the fine-tuning data set is used for fine-tuning the migration model, and the test data set is used to test the diagnostic performance of the migration model in the central air-conditioning system of the target building;
[0035] The model is fine-tuned by freezing all convolutional layers of the pre-trained model and deleting the fully connected output layer or freezing some convolutional layers of the pre-trained model. The two transfer learning methods migrate the pre-trained one-dimensional convolutional neural network CNN_1D model to the fault diagnosis of the target building, and establish a fault diagnosis migration model.
[0036] According to the fault diagnosis migration model, the first normal data and the first fault data of the central air-conditioning system collected by the source building are used to train the CNN_1D model to obtain a pre-trained model for transfer learning. The pre-trained model is fine-tuned or feature extracted using 20% of the operating data of the central air-conditioning system of the target building. The migration model is tested using the remaining 80% of the operating data, where the operating data includes the second normal data and the second fault data. Then, the situation in which the fault operating data of the central air-conditioning system of the building is relatively scarce in actual situations is simulated.
[0037] Optionally, the C v The value range of (RMBE) is [0, +∞].
[0038] Optionally, in the acquisition of the source building and the simulation of the failure or abnormality of the central air-conditioning system of the source building, the failure or abnormality of the air handling unit includes the failure of the fresh air valve, the mixed air valve, the surface cooler valve and the heater valve; the failure or abnormality of the terminal equipment includes the failure of the air valve, the flow sensor, the temperature sensor and the controller.
[0039] Optionally, the target building and the source building have similar structures, but different central air-conditioning system configurations.
[0040] Optionally, the target domain data and the source domain data are preprocessed by Z-score standardization, and the target domain data and the source domain data after the preprocessing become a normal distribution with an expected value of μ=0 and a standard deviation of σ=1;
[0041] Among them, the Z-socre standardized calculation formula is as follows:
[0042]
[0043] Where μ is the sample mean and σ is the sample standard deviation.
[0044] Optionally, the method for establishing a fault diagnosis migration model includes:
[0045] Acquire operation data of the central air-conditioning system of the source building, perform first data preprocessing on the operation data to obtain a training data set and a first test data set, and obtain a pre-trained CNN_1D model according to the training data set and the first test data set;
[0046] Acquire operation data of the central air-conditioning system of the target building, perform a second data preprocessing on the operation data to obtain a fine-tuning data set and a second test data set, obtain a feature extractor by freezing all convolutional layers according to the fine-tuning data set, and obtain the feature extractor and a fully connected neural network classifier by using the second test data set;
[0047] Model fine-tuning is obtained based on the pre-trained CNN_1D model and the feature extractor, the fully connected neural network classifier is also obtained based on the model fine-tuning, and the central air-conditioning system fault diagnosis result is obtained based on the fully connected neural network classifier.
[0048] Compared with the prior art, in the central air-conditioning fault diagnosis model migration method disclosed in the present invention, the target building is first simulated and verified; then, the central air-conditioning system is fault simulated to obtain source domain data and target domain data respectively; secondly, the source domain data and the target domain data are pre-processed and divided into a training data set, a verification data set, a fine-tuning data set and a test data set respectively, and finally, the pre-trained CNN_1D model is migrated to the fault diagnosis of the target building through two transfer learning methods, thereby solving the problem of lack of target building fault data in actual application scenarios and difficulty in establishing an effective fault diagnosis model. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] One or more embodiments are exemplarily described by corresponding drawings, which do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and the figures in the drawings do not constitute proportional limitations unless otherwise stated.
[0050] Figure 1 This is a flow chart of a method for migrating a central air-conditioning fault diagnosis model in an embodiment of the present disclosure;
[0051] Figure 2 The present invention is a flowchart of a method for establishing a fault diagnosis migration model in an embodiment of the present invention. DETAILED DESCRIPTION
[0052] In order to facilitate the understanding of the present disclosure, the present disclosure is described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that when an element is described as "fixed to" another element, it can be directly on the other element, or there can be one or more centered elements therebetween. When an element is described as "connected to" another element, it can be directly connected to the other element, or there can be one or more centered elements therebetween. The terms "vertical", "horizontal", "left", "right", "inside", "outside" and similar expressions used in this specification are for illustrative purposes only.
[0053] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art of the present disclosure. The terms used in the specification of the present disclosure are only for the purpose of describing specific embodiments and are not used to limit the present disclosure. The term "and / or" used in this specification includes any and all combinations of one or more related listed items.
[0054] In addition, the terms first, second, third, etc. in the specification and claims are only used for the purpose of describing the same technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated, nor necessarily describing the order or time sequence. The terms are interchangeable where appropriate. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features.
[0055] Similarly, the terms "fixed" and "connected" used in the specification and claims should not be construed as limited to direct connection. Therefore, the expression "device A is connected to device B" should not be limited to a device or system in which device A is directly connected to device B, but means that there is a path between device A and device B, which may be a path including other devices or tools.
[0056] In addition, the technical features involved in different embodiments of the present disclosure described below can be combined with each other as long as they do not conflict with each other.
[0057] Please refer to Figure 1 The embodiment of the present disclosure proposes a central air-conditioning fault diagnosis model migration method, including:
[0058] Obtain a target building and perform simulation modeling on the target building. The calculation formula is as follows:
[0059] NMBE = (average value of simulation values - average value of measured values) / average value of measured values
[0060] C v (RMBE) = RMSE / average of measured values × 100%
[0061]
[0062] Where, NMBE standard mean offset error, represents the simulated value of building energy consumption at the i-th time point; y i represents the actual measured energy consumption of the building at the i-th time point; n is the number of samples in the dataset;
[0063] Acquire a source building, and simulate a fault or anomaly of a central air-conditioning system of the source building, wherein the fault or anomaly of the air-conditioning chiller includes refrigerant leakage, overcharging of refrigerant, scaling of the condenser, scaling of the evaporator, the presence of non-condensable gas in the chiller, too little cooling water flow, and too little freezing water flow;
[0064] The calculation formula for the refrigerant leakage fault of the air-conditioning chiller is as follows:
[0065] COP REF1,f =COP REF1,n ×F REF1
[0066] Among them, COP REF1,f Represents the first coefficient of performance of the chiller under fault conditions, dimensionless; COP REF1,n It represents the first energy efficiency coefficient of the chiller under normal working conditions, dimensionless; F REF1 It represents the first failure coefficient in this scenario, describing the reduction degree of chiller performance caused by the refrigerant leakage of the chiller;
[0067] The calculation formula for the overcharge fault of the chiller is as follows:
[0068] COP REF2,f =COP REF2,n ×F REF2
[0069] Among them, COP REF2,f Represents the second coefficient of performance of the chiller under fault conditions, dimensionless; COP REF2,n It represents the second energy efficiency coefficient of the chiller under normal working conditions, dimensionless; F REF2 It represents the second fault coefficient in this scenario, describing the reduction of chiller performance caused by overcharging of chiller refrigerant;
[0070] The calculation formula for the scaling fault of the chiller condenser is as follows:
[0071] K con,f =K con,n ×F con
[0072] Among them, K con,f = indicates the heat transfer coefficient of the chiller condenser under fault conditions, in W / m 2 ·K; K con,n Indicates the heat transfer coefficient of the chiller condenser under normal operating conditions, in W / m 2 ·K; F con Describe the extent to which condenser scaling reduces the heat transfer coefficient of the chiller condenser;
[0073] The calculation formula for the scaling fault of the chiller evaporator is as follows:
[0074] K eva,f =K eva,n ×F eva
[0075] Among them, K eva,f = indicates the heat transfer coefficient of the chiller evaporator under fault conditions, in W / m 2 ·K; K eva,n Indicates the heat transfer coefficient of the chiller evaporator under normal operating conditions, in W / m 2 ·K; F eva Describe the extent to which evaporator scaling reduces the heat transfer coefficient of the chiller evaporator;
[0076] The calculation formula for the non-condensable gas operation fault of the chiller is as follows:
[0077] P gas,f =P gas,n ×F gas
[0078] Among them, P gas,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P gas,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F gas It represents the failure coefficient in this scenario, describing the increase in the energy consumption of the chiller caused by the presence of non-condensable gas in the chiller;
[0079] The calculation formula for the failure of the chiller to operate due to insufficient cooling water flow is as follows:
[0080] P cw,f =P cw,n ×F cw
[0081] Among them, P cw,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P cw,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F cw It represents the failure coefficient in this scenario, describing the increase in energy consumption of the chiller due to insufficient cooling water flow of the chiller;
[0082] The calculation formula for the chiller unit's operating fault due to insufficient chilled water flow is as follows:
[0083] P chw,f =P chw,n ×F chw
[0084] Among them, P chw,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P chw,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F chw It represents the failure coefficient in this scenario, describing the increase in energy consumption of the chiller due to insufficient chilled water flow of the chiller;
[0085] According to the simulation modeling and the fault or abnormality simulation, target domain data and source domain data are obtained respectively, and the target domain data and the source domain data are preprocessed to obtain a training data set, a verification data set, a fine-tuning data set and a test data set, wherein the training data set and the verification data set are respectively used to train and verify the performance of the pre-trained one-dimensional convolutional neural network CNN_1D model in the fault diagnosis of the central air-conditioning system of the source building, the fine-tuning data set is used for fine-tuning the migration model, and the test data set is used to test the diagnostic performance of the migration model in the central air-conditioning system of the target building;
[0086] The model is fine-tuned by freezing all convolutional layers of the pre-trained model and deleting the fully connected output layer or freezing some convolutional layers of the pre-trained model. The two transfer learning methods migrate the pre-trained one-dimensional convolutional neural network CNN_1D model to the fault diagnosis of the target building, and establish a fault diagnosis migration model.
[0087] According to the fault diagnosis migration model, the first normal data and the first fault data of the central air-conditioning system collected by the source building are used to train the CNN_1D model to obtain a pre-trained model for transfer learning. The pre-trained model is fine-tuned or feature extracted using 20% of the operating data of the central air-conditioning system of the target building. The migration model is tested using the remaining 80% of the operating data, where the operating data includes the second normal data and the second fault data. Then, the situation in which the fault operating data of the central air-conditioning system of the building is relatively scarce in actual situations is simulated.
[0088] Among them, the standard mean offset error NMBE represents the degree of deviation between the simulation result and the actual measurement value. The target building can be a large public building, and EnergyPlus can be used to simulate and model the target building. The central air-conditioning fault diagnosis model migration method of this embodiment first simulates and verifies the target building; then simulates the fault of its central air-conditioning system to obtain source domain data and target domain data respectively; secondly, the source domain data and the target domain data are pre-processed and divided into a training data set, a verification data set, a fine-tuning data set and a test data set, and finally the pre-trained CNN_1D model is migrated to the fault diagnosis of the target building through two transfer learning methods, thereby solving the problem of lack of target building fault data in actual application scenarios and difficulty in establishing an effective fault diagnosis model.
[0089] In one embodiment, the C v The value range of (RMBE) is [0, +∞], C v The smaller the value of (RMBE), the higher the model accuracy.
[0090] In one embodiment, in the acquisition of the source building and the simulation of the fault or abnormality of the central air conditioning system of the source building, the fault or abnormality of the air handling unit includes the fresh air valve fault, the mixed air valve fault, the surface cooler valve fault and the heater valve fault; the fault or abnormality of the terminal equipment includes the air valve fault, the flow sensor fault, the temperature sensor fault and the controller fault. Thus, an effective fault diagnosis model can be established more accurately.
[0091] In one embodiment, the target building and the source building have similar structures, but different central air-conditioning system configurations.
[0092] Since the source domain data and the target domain data have many data types, large quantities, and large differences in the dimensions of various parameters, in one embodiment, the target domain data and the source domain data are preprocessed using Z-score standardization, and the target domain data and the source domain data after preprocessing become a normal distribution with an expected value of μ=0 and a standard deviation of σ=1;
[0093] Among them, the Z-socre standardized calculation formula is as follows:
[0094]
[0095] Where μ is the sample mean and σ is the sample standard deviation.
[0096] Please refer to Figure 2 In one embodiment, the method for establishing a fault diagnosis migration model includes:
[0097] Acquire operation data of the central air-conditioning system of the source building, perform first data preprocessing on the operation data to obtain a training data set and a first test data set, and obtain a pre-trained CNN_1D model according to the training data set and the first test data set;
[0098] Acquire operation data of the central air-conditioning system of the target building, perform a second data preprocessing on the operation data to obtain a fine-tuning data set and a second test data set, obtain a feature extractor by freezing all convolutional layers according to the fine-tuning data set, and obtain the feature extractor and a fully connected neural network classifier by using the second test data set;
[0099] The model is fine-tuned according to the pre-trained CNN_1D model and the feature extractor, the fully connected neural network classifier is also obtained according to the model fine-tuning, and the central air-conditioning system fault diagnosis result is obtained according to the fully connected neural network classifier. The CNN_1D model trained based on the data of the source building has a diagnostic accuracy rate of up to 96.50% for the target building in the above scenario, which to a certain extent shows that the distribution of operating data of different systems in similar buildings is highly similar.
[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Under the idea of the present disclosure, the technical features in the above embodiments or different embodiments may also be combined, the steps may be implemented in any order, and there are many other changes in different aspects of the present disclosure as described above, which are not provided in detail for the sake of simplicity. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in each of the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of each embodiment of the present disclosure.
Claims
1. A central air conditioning fault diagnosis model migration method, characterized in that: include: Obtain a target building and perform simulation modeling on the target building. The calculation formula is as follows: NMBE = (average value of simulation values - average value of measured values) / average value of measured values C v (RMBE) = RMSE / average of measured values × 100% Where, NMBE standard mean offset error, represents the simulated value of building energy consumption at the i-th time point; y i represents the actual measured energy consumption of the building at the i-th time point; n is the number of samples in the dataset; Acquire a source building, and simulate a fault or anomaly of a central air-conditioning system of the source building, wherein the fault or anomaly of the air-conditioning chiller includes refrigerant leakage, overcharging of refrigerant, scaling of the condenser, scaling of the evaporator, the presence of non-condensable gas in the chiller, too little cooling water flow, and too little freezing water flow; The calculation formula for the refrigerant leakage fault of the air-conditioning chiller is as follows: COP REF1,f =COP REF1,n ×F REF1 Among them, COP REF1,f Represents the first coefficient of performance of the chiller under fault conditions, dimensionless; COP REF1,n It represents the first energy efficiency coefficient of the chiller under normal working conditions, dimensionless; F REF1 It represents the first failure coefficient in this scenario, describing the reduction degree of chiller performance caused by the refrigerant leakage of the chiller; The calculation formula for the overcharge fault of the chiller is as follows: COP REF2,f =COP REF2,n ×F REF2 Among them, COP REF2,f Represents the second coefficient of performance of the chiller under fault conditions, dimensionless; COP REF2,n It represents the second energy efficiency coefficient of the chiller under normal working conditions, dimensionless; F REF2 It represents the second fault coefficient in this scenario, describing the reduction of chiller performance caused by overcharging of chiller refrigerant; The calculation formula for the scaling failure of the chiller condenser is as follows: K con,f =K con,n ×F con Among them, K con,f = indicates the heat transfer coefficient of the chiller condenser under fault conditions, in W / m 2 ·K; K con,n Indicates the heat transfer coefficient of the chiller condenser under normal operating conditions, in W / m 2 ·K; F con Describe the extent to which condenser scaling reduces the heat transfer coefficient of the chiller condenser; The calculation formula for the scaling fault of the chiller evaporator is as follows: K eva,f =K eva,n ×F eva Among them, K eva,f = indicates the heat transfer coefficient of the chiller evaporator under fault conditions, in W / m 2 ·K; K eva,n Indicates the heat transfer coefficient of the chiller evaporator under normal operating conditions, in W / m 2 ·K; F eva Describe the extent to which evaporator scaling reduces the heat transfer coefficient of the chiller evaporator; The calculation formula for the non-condensable gas operation fault of the chiller is as follows: P gas,f =P gas,n ×F gas Among them, P gas,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P gas,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F gas It represents the failure coefficient in this scenario, describing the increase in the energy consumption of the chiller caused by the presence of non-condensable gas in the chiller; The calculation formula for the failure of the chiller to operate due to insufficient cooling water flow is as follows: P cw,f =P cw,n ×F cw Among them, P cw,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P cw,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F cw It represents the failure coefficient in this scenario, describing the increase in energy consumption of the chiller due to insufficient cooling water flow of the chiller; The calculation formula for the failure of the chiller unit to operate due to insufficient chilled water flow is as follows: P chw,f =P chw,n ×F chw Among them, P chw,f Indicates the energy consumption of the chiller under fault conditions, in kWh; P chw,n Indicates the energy consumption of the chiller under normal operating conditions, in kWh; F chw It represents the failure coefficient in this scenario, describing the increase in energy consumption of the chiller due to insufficient chilled water flow of the chiller; According to the simulation modeling and the fault or abnormality simulation, target domain data and source domain data are obtained respectively, and the target domain data and the source domain data are preprocessed to obtain a training data set, a verification data set, a fine-tuning data set and a test data set, wherein the training data set and the verification data set are respectively used to train and verify the performance of the pre-trained one-dimensional convolutional neural network CNN_1D model in the fault diagnosis of the central air-conditioning system of the source building, the fine-tuning data set is used for fine-tuning the migration model, and the test data set is used to test the diagnostic performance of the migration model in the central air-conditioning system of the target building; The model is fine-tuned by freezing all convolutional layers of the pre-trained model and deleting the fully connected output layer or freezing some convolutional layers of the pre-trained model. The two transfer learning methods migrate the pre-trained one-dimensional convolutional neural network CNN_1D model to the fault diagnosis of the target building, and establish a fault diagnosis migration model. According to the fault diagnosis migration model, the first normal data and the first fault data of the central air-conditioning system collected by the source building are used to train the CNN_1D model to obtain a pre-trained model for transfer learning. The pre-trained model is fine-tuned or feature extracted using 20% of the operating data of the central air-conditioning system of the target building. The migration model is tested using the remaining 80% of the operating data, where the operating data includes the second normal data and the second fault data. Then, the situation in which the fault operating data of the central air-conditioning system of the building is relatively scarce in actual situations is simulated.
2. The central air conditioning fault diagnosis model migration method according to claim 1 is characterized in that: The C v The value range of (RMBE) is [0, +∞].
3. The central air conditioning fault diagnosis model migration method according to claim 1 is characterized in that: In the acquisition of the source building and the simulation of the failure or abnormality of the central air-conditioning system of the source building, the failure or abnormality of the air handling unit includes the failure of the fresh air valve, the mixed air valve, the surface cooler valve and the heater valve; the failure or abnormality of the terminal equipment includes the failure of the air valve, the flow sensor, the temperature sensor and the controller.
4. The central air conditioning fault diagnosis model migration method according to claim 1 is characterized in that: The target building and the source building have similar structures, but different central air-conditioning system configurations.
5. The central air conditioning fault diagnosis model migration method according to claim 1 is characterized in that: The target domain data and the source domain data are preprocessed by using Z-score standardization, so that the target domain data and the source domain data after the preprocessing become a normal distribution with an expected value of μ=0 and a standard deviation of σ=1; Among them, the Z-socre standardized calculation formula is as follows: Where μ is the sample mean and σ is the sample standard deviation.
6. The central air conditioning fault diagnosis model migration method according to claim 1, characterized in that: The method for establishing a fault diagnosis migration model comprises: Acquire operation data of the central air-conditioning system of the source building, perform first data preprocessing on the operation data to obtain a training data set and a first test data set, and obtain a pre-trained CNN_1D model according to the training data set and the first test data set; Acquire operation data of the central air-conditioning system of the target building, perform a second data preprocessing on the operation data to obtain a fine-tuning data set and a second test data set, obtain a feature extractor by freezing all convolutional layers according to the fine-tuning data set, and obtain the feature extractor and a fully connected neural network classifier by using the second test data set; Model fine-tuning is obtained based on the pre-trained CNN_1D model and the feature extractor, the fully connected neural network classifier is also obtained based on the model fine-tuning, and the central air-conditioning system fault diagnosis result is obtained based on the fully connected neural network classifier.