A method and system for determining a failure precursor of an air conditioner

Abstract

The present application relates to a kind of air conditioner's failure omen determination method and system, air conditioner's failure omen determination method includes the following steps: obtaining the current operating data of air conditioner;Determine whether there is failure omen based on current operating data and normal operating data;If there is failure omen, then based on current operating data and normal operating data, obtain the first unit deviation vector;Determine the fault type based on the first unit deviation vector, fault type and the first unit deviation vector have preset mapping relationship.The air conditioner's failure omen determination method provided by the present application can identify potential failure mode and corresponding hidden danger components in the air conditioner failure omen stage, so that maintenance personnel can carry out inspection and repair work on air conditioner components, avoid invalid investigation link, thereby improving maintenance efficiency;And avoid the serious consequences caused by air conditioner downtime for a long time due to sudden failure, cause business continuity interruption and potential property loss.

Description

Technical Field

[0001] This invention relates to the field of air conditioning technology, and more specifically, to a method and system for determining the early signs of air conditioning malfunctions. Background Technology

[0002] Commercial air conditioning systems play a crucial role in environmental regulation in commercial environments such as shops and offices. Their continuous and stable operation is essential to ensuring user comfort and business continuity. However, when a substantial malfunction occurs in the air conditioning system, leading to an abnormal shutdown, the system will completely lose its cooling or heating capacity until repairs are completed.

[0003] With the development of intelligent operation and maintenance technology, commercial air conditioners have begun to adopt fault precursor detection technology, which can detect potential fault risks in advance. However, the existing fault precursor detection technology can only determine whether there is an anomaly, and cannot accurately predict the fault mode type in the fault precursor stage. It also has the following shortcomings: the air conditioner can only give a warning of the presence of an anomaly, but cannot identify the specific fault type and the corresponding hidden parts in advance. Operation and maintenance personnel can only check possible fault points one by one, and cannot carry out targeted inspection, repair or replacement of corresponding parts, resulting in low maintenance efficiency. Summary of the Invention

[0004] This invention provides a method for predicting air conditioner malfunctions, which addresses the shortcomings of existing technologies. These technologies cannot determine the type of malfunction during the malfunction prediction stage, nor can they identify the specific malfunction type and corresponding potential components in advance. As a result, maintenance personnel can only check for possible malfunctions one by one, leading to low maintenance efficiency.

[0005] This invention provides a method for determining the early signs of an air conditioner malfunction, the method comprising the following steps: Obtain the current operating data of the air conditioner; Based on the current operating data and the normal operating data, determine whether the air conditioner shows any signs of malfunction; If there are signs of a fault, a first unit deviation vector is obtained based on the current operating data and the normal operating data; Based on the first unit deviation vector, the fault type is determined, wherein the fault type and the first unit deviation vector have a preset mapping relationship.

[0006] Optionally, the preset mapping relationship is obtained based on the training of a fault prediction model, and the training method of the fault prediction model includes: Obtain the historical normal operation dataset and the historical fault operation dataset of the air conditioner. The historical normal operation dataset includes data of multiple operating parameters at different times under normal operating conditions, and the historical fault operation dataset includes data of multiple operating parameters corresponding to multiple fault types at different times. The historical normal operation dataset is subjected to dimensionality reduction processing to obtain the dimensionality-reduced historical normal state coordinates; The historical fault operation dataset is subjected to dimensionality reduction processing to obtain the dimensionality-reduced historical fault state coordinates; Based on the historical fault state coordinates and the historical normal state coordinates, a second unit deviation vector is obtained; Using a machine learning algorithm, based on the second unit deviation vector and the fault type, the preset mapping relationship is fitted to obtain the trained fault prediction model.

[0007] Optionally, the step of performing dimensionality reduction processing on the historical normal operation dataset to obtain the dimensionality-reduced historical normal state coordinates includes: Based on the historical normal operation dataset, a normal mean vector and covariance matrix composed of the data mean values ​​of each operating parameter are obtained; satisfying the formula: ; in, Z For historically normal operation datasets, This is a normal mean vector. This represents the average of the operating parameters during historical normal operation. n The number of running parameters, k This represents the number of learning parameters for each operating parameter during historical normal operation. Let covariance matrix be the variance matrix. For the first i Normal operating data for each operating parameter. For the first j Normal operating data for each operating parameter. For the first i The average data of each operating parameter i =1,2,3,… n ; For the first j The average data of each operating parameter j =1,2,3,… n ; The covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues ​​and corresponding eigenvectors; satisfying the formula: ; in, For eigenvalues, ; For feature vectors, , t The number of eigenvalues; The feature values ​​are sorted from largest to smallest, and the feature values ​​at the top of the sort are selected as the principal feature values. Based on the normal mean vector and the principal eigenvector corresponding to the principal eigenvalue, the dimensionality-reduced historical normal state coordinates are obtained, satisfying the formula: ; in, m The number of principal eigenvalues This is a normal mean vector. The principal eigenvectors corresponding to the principal eigenvalues; These are the coordinates of the historical normal state after dimensionality reduction.

[0008] Optionally, the step of performing dimensionality reduction processing on the historical fault operation dataset to obtain the dimensionality-reduced historical fault state coordinates includes: Based on the historical fault operation dataset, a fault mean vector is obtained, consisting of the average values ​​of each operating parameter; satisfying the formula: ; Based on the fault mean vector and the principal feature vector, the dimensionality-reduced historical fault state coordinates are obtained, satisfying the formula: ; in, D For historical fault running datasets, The fault mean vector, This represents the average of various operating parameters during historical fault scenarios. n The number of running parameters, r The number of times each operating parameter was learned during historical fault testing; m The number of principal eigenvalues The principal eigenvectors corresponding to the principal eigenvalues; These are the coordinates of the historical fault states after dimensionality reduction.

[0009] Optionally, the step of obtaining the second unit deviation vector based on the historical fault state coordinates and the historical normal state coordinates satisfies the formula: ; in, The second unit deviation vector, This is a vector connecting the coordinates of historical normal states and historical fault states. The magnitude of the vector connecting the historical normal state coordinates and the historical fault state coordinates.

[0010] Optionally, the normal operation data is the historical normal operation dataset, and the step of obtaining the first unit deviation vector based on the current operation data and the normal operation data includes: The current running data is subjected to dimensionality reduction processing to obtain the dimensionality-reduced current running state coordinates; Based on the current operating state coordinates and the historical normal state coordinates, the first unit deviation vector is obtained.

[0011] Optionally, the step of performing dimensionality reduction processing on the current running data to obtain the dimensionality-reduced current running state coordinates is as follows: Based on the current running data and the main feature vector, the dimensionality-reduced current running state coordinates are obtained, satisfying the formula: ; in, X For current running data, P m The coordinates represent the current running state. The step of obtaining the first unit deviation vector based on the current operating state coordinates and the historical normal state coordinates includes: Based on the current operating state coordinates and the historical normal state coordinates, a first deviation vector is obtained; satisfying the formula: ; Based on the first deviation vector and its magnitude, the first unit deviation vector is obtained; satisfying the formula: ; in, The deviation direction vector, The first unit deviation vector, This represents the magnitude of the deviation vector.

[0012] Optionally, the step of determining whether the air conditioner has signs of a malfunction based on the current operating data and the normal operating data includes: obtaining an anomaly degree based on the current operating data and the normal operating data; and determining whether the air conditioner has signs of a malfunction based on the anomaly degree. And / or, the step of determining the fault type based on the first unit deviation vector includes: Using the preset mapping relationship, the first unit deviation vector is mapped and the fault confidence level corresponding to each fault type is output. The fault type is determined based on the fault confidence level and the preset fault threshold.

[0013] Optionally, the operating parameters include at least two of the following: outer ring temperature, indoor temperature, high pressure, low pressure, suction superheat, exhaust superheat, subcooling, compressor frequency, and compressor current. And / or, the fault types include at least two of the following: refrigerant leakage, compressor deterioration, external fan motor deterioration, internal fan motor deterioration, blockage of outdoor heat exchanger, blockage of indoor heat exchanger, malfunction of external expansion valve, malfunction of internal expansion valve, and oil return solenoid valve failure; And / or, the machine learning algorithm includes either a neural network or a random forest.

[0014] The air conditioner fault prediction method provided by this invention has at least the following beneficial technical effects: By acquiring the current operating data of the air conditioner and comparing it with the normal operating data, it is possible to determine whether there are any signs of a fault. If there are signs of a fault, the first unit deviation vector is further calculated, and the specific fault type is determined based on the first unit deviation vector and the preset mapping relationship. This allows potential fault modes and corresponding hidden components to be identified in the air conditioner fault warning stage, enabling maintenance personnel to carry out targeted inspections and repairs on the air conditioner components, avoiding ineffective troubleshooting steps and thus improving maintenance efficiency. Furthermore, it avoids serious consequences such as prolonged air conditioner shutdowns due to sudden faults, resulting in business continuity interruptions and potential property damage.

[0015] This invention also provides a fault prediction system for air conditioners, which applies the above-described fault prediction method for air conditioners. The fault prediction system for air conditioners includes: The acquisition module is used to acquire the current operating data of the air conditioner; The determination module is used to determine whether the air conditioner shows signs of malfunction based on the current operating data and the normal operating data; The calculation module is configured to, if fault precursors exist, obtain a first unit deviation vector based on the current fault operation data and the historical normal operation dataset; and, The fault type determination module is used to determine the fault type based on the first unit deviation vector, wherein the fault type and the first unit deviation vector have a preset mapping relationship.

[0016] The air conditioner fault prediction system provided by this invention is applied to the above-mentioned air conditioner fault prediction method and has all the advantages of the method, which will not be elaborated here. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a method for determining early signs of air conditioning malfunctions according to an embodiment of the present invention; Figure 2 A flowchart illustrating the training method of the fault estimation model in an air conditioner fault prediction method provided in an embodiment of the present invention. Figure 3A flowchart illustrating the process of obtaining historical normal state coordinates in an air conditioner fault prediction method provided in an embodiment of the present invention. Figure 4 A flowchart illustrating the process of obtaining historical fault state coordinates in an air conditioner fault prediction method provided in an embodiment of the present invention. Figure 5 A flowchart illustrating the process of obtaining the first unit deviation vector in an air conditioner fault prediction method provided in this embodiment of the invention. Figure 1 ; Figure 6 A flowchart illustrating the process of obtaining the first unit deviation vector in an air conditioner fault prediction method provided in this embodiment of the invention. Figure 2 ; Figure 7 A schematic diagram of the fault prediction process in an air conditioner fault prediction method provided in an embodiment of the present invention; Figure 8 A flowchart illustrating the process of determining the fault type in an air conditioner fault prediction method provided in an embodiment of the present invention; Figure 9 This is a structural block diagram of an air conditioner fault prediction system provided in an embodiment of the present invention.

[0018] Explanation of reference numerals in the attached figures: 10. Acquisition Module; 20. Judgment Module; 30. Calculation Module; 40. Fault Type Determination Module. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the following description is provided in conjunction with the accompanying drawings. Figures 1-9 Specific embodiments of the present invention will be described in detail below.

[0020] This invention provides a method for determining the early signs of an air conditioner malfunction; see attached document. Figure 1 The method for determining the early signs of malfunction in this air conditioner includes the following steps: S100: Obtain the current operating data of the air conditioner; collect operating data of the air conditioner during operation based on sensors; S120, Based on the current operating data and normal operating data, determine whether the air conditioner has any signs of failure. In this step, normal operating data refers to the operating data of the air conditioner in a fault-free state. Signs of failure refer to the trend of the air conditioner's operating parameters deviating from the normal range even though the air conditioner has not yet experienced a functional failure, indicating that the air conditioner may have potential component deterioration or performance degradation. S140, if there are signs of a fault, a first unit deviation vector is obtained based on the current operating data and the normal operating data; in this step, the first unit deviation vector can quantify the direction of the current operating state from the normal operating state, and provide a directional basis for determining the subsequent fault type; S160, Based on the first unit deviation vector, determine the fault type, where the fault type and the first unit deviation vector have a preset mapping relationship. In this step, the fault type refers to the specific fault mode of the air conditioner, such as refrigerant leakage, compressor deterioration, etc.; the preset mapping relationship refers to the correspondence established between the first unit deviation vector and the fault type. This preset mapping relationship can be obtained through model training, or it can be obtained through table definition or function method.

[0021] This invention provides a method for determining air conditioner fault precursors. It acquires the current operating data of the air conditioner and compares it with normal operating data to determine if fault precursors exist. If fault precursors are found, a first unit deviation vector is further calculated, and the specific fault type is determined based on the first unit deviation vector and a preset mapping relationship. This method can identify potential fault modes and corresponding hazardous components during the air conditioner fault precursor stage, enabling maintenance personnel to conduct targeted inspections and repairs on the air conditioner components, avoiding ineffective troubleshooting and thus improving maintenance efficiency. Furthermore, it avoids serious consequences such as prolonged air conditioner shutdown due to sudden faults, leading to business continuity interruptions and potential property damage.

[0022] In this embodiment of the invention, see appendix. Figure 2 In step S160, the preset mapping relationship is obtained based on the training of the fault estimation model. The training method of the fault estimation model includes: S170, acquire the historical normal operation dataset and historical fault operation dataset of the air conditioner. The historical normal operation dataset includes data on multiple operating parameters at different times under normal operating conditions, while the historical fault operation dataset includes data on multiple operating parameters corresponding to various fault types at different times. In this step, the historical normal operation dataset and historical fault operation dataset provide rich learning samples for subsequent model training. Operating parameters include at least two of the following: external ambient temperature, indoor temperature, high pressure, low pressure, suction superheat, exhaust superheat, subcooling, compressor frequency, and compressor current. Specifically, external ambient temperature and indoor temperature directly reflect the environmental load of the air conditioner; high pressure and low pressure are core pressure parameters in the refrigerant cycle, directly related to the operating conditions of the compressor, condenser, and evaporator; suction superheat, ... Discharge superheat and subcooling refine the thermodynamic state of the refrigerant during circulation, which is crucial for determining refrigerant charge, expansion valve opening, and heat exchange efficiency. Compressor frequency and compressor current characterize the compressor's workload and energy consumption. Fault types include at least two of the following: refrigerant leakage, compressor deterioration, external fan motor deterioration, internal fan motor deterioration, blockage of the outdoor heat exchanger, blockage of the indoor heat exchanger, malfunction of the external expansion valve, malfunction of the internal expansion valve, and oil return solenoid valve failure. Specifically, for example, refrigerant leakage may cause abnormal high or low pressure; compressor deterioration may manifest as abnormal compressor current; fan motor deterioration may affect heat exchange efficiency; blockage of the heat exchanger will hinder heat exchange; malfunction of the expansion valve may affect refrigerant flow control; and oil return valve failure may affect compressor lubrication and lifespan.

[0023] S172, Dimensionality reduction is performed on the historical normal operation dataset to obtain the dimensionality-reduced historical normal state coordinates; the purpose of dimensionality reduction in this step is to extract core features from the high-dimensional historical normal operation data and remove noise from the data, thereby mapping the high-dimensional historical normal operation data to a low-dimensional space to obtain the dimensionality-reduced historical normal state coordinates. S174, Dimensionality reduction is performed on the historical fault operation dataset to obtain the dimensionality-reduced historical fault state coordinates; Similarly, the purpose of dimensionality reduction on the historical fault operation dataset in this step is to extract core features from the high-dimensional historical fault operation data and remove noise from the data, thereby mapping the high-dimensional historical fault operation data to a low-dimensional space to obtain the dimensionality-reduced historical fault state coordinates. S176, Based on the historical fault state coordinates and historical normal state coordinates, the second unit deviation vector is obtained; In this step, the second unit deviation vector quantifies the offset direction between the historical fault state coordinates and the historical normal state coordinates in the low-dimensional space, and serves as the input for the subsequent model. S178, using a machine learning algorithm, a pre-defined mapping relationship is fitted based on the second unit deviation vector and the fault type to obtain a trained fault prediction model. It should be noted that in this step, the machine learning algorithm can include either a neural network or a random forest. The fault type has a label value, FR, which includes 0 and 1, where 0 represents a normal state and 1 represents a fault state. For example, if the number of fault types is 9, they are defined as follows: FR1: Refrigerant leak; FR2: Compressor deterioration; FR3: Outdoor fan motor deterioration; FR4: Indoor fan motor deterioration; FR5: Outdoor heat exchanger blockage; FR6: Indoor heat exchanger blockage; FR7: External expansion valve malfunction; FR8: Internal expansion valve malfunction; FR9: Oil return solenoid valve malfunction. For example, when only the refrigerant leak fault mode occurs, the values ​​of each flag are as follows: (FR1,FR2,FR3,FR4,FR5,FR6,FR7,FR8,FR9) = (1,0,0,0,0,0,0,0,0).

[0024] The air conditioner fault prediction method provided in this invention collects a large amount of historical normal operation data and historical fault operation data, and performs dimensionality reduction processing on both to effectively extract the core features of the air conditioner's operating status. This allows the normal and fault states to be represented in a low-dimensional space using historical normal state coordinates and historical fault state coordinates. Based on these historical fault state coordinates and historical normal state coordinates, a second unit deviation vector is calculated to quantify the offset direction of different fault modes relative to the normal state. Then, a machine learning algorithm is used to fit and train the second unit deviation vector with the corresponding fault type to obtain a preset mapping relationship. This method overcomes the limitations of traditional experience-based judgment, improves the accuracy and reliability of fault type determination, and enables the air conditioning system to identify potential fault types earlier and more accurately. This allows maintenance personnel to conduct targeted inspections and repairs, thereby improving maintenance efficiency and avoiding serious consequences such as prolonged air conditioner downtime due to sudden faults, which could lead to business continuity interruptions and potential property damage.

[0025] In this embodiment of the invention, see appendix. Figure 3 Step S172, the step of performing dimensionality reduction processing on the historical normal operation dataset to obtain the dimensionality-reduced historical normal state coordinates, includes: S1720, based on the historical normal operation dataset, yields a normal mean vector and covariance matrix composed of the data means of each operating parameter; satisfying the formula: ; in, Z For historically normal operation datasets, This is a normal mean vector. This represents the average of the operating parameters during historical normal operation. n The number of running parameters, k The number of learning parameters for each operating parameter during historical normal operation, for example, k It is 200; Let covariance matrix be the variance matrix. For the first i Normal operating data for each operating parameter. For the first j Normal operating data for each operating parameter. For the first i The average data of each operating parameter i =1,2,3, n ,For example, n It is 9; For the first j The average data of each operating parameter j =1,2,3, n ,For example, n The value is 9; this step uses statistical methods to express the data distribution characteristics of the air conditioner under normal operating conditions, where the normal mean vector represents the average level of each operating parameter under normal operating conditions; the covariance matrix represents the coordinated changes between different operating parameters; S1722, Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and their corresponding eigenvectors; satisfying the formula: ; in, For eigenvalues, ; For feature vectors, , t The number of eigenvalues; this step, eigenvalue decomposition, is the core step of principal component analysis, used to identify the most significant direction of variation in normal operating data. After eigenvalue decomposition, a set of eigenvalues ​​and corresponding eigenvectors can be obtained, where each eigenvector represents a principal component, indicating the direction of the data with the largest variance, and the corresponding eigenvalue is used to quantify the variance of the principal component. S1724, Sort the eigenvalues ​​in descending order of value and select the top-ranked eigenvalues ​​as principal eigenvalues; This step, by sorting the eigenvalues ​​in descending order of value, can identify the principal components that contribute the most to the data variance; Selecting the top-ranked eigenvalues ​​as principal eigenvalues ​​means discarding principal components that contribute less and may contain noise, while retaining principal components that contribute more, thereby achieving data dimensionality reduction; S1726, based on the normal mean vector and the principal eigenvectors corresponding to the principal eigenvalues, obtains the dimensionality-reduced historical normal state coordinates, satisfying the formula: ; in, m The number of principal eigenvalues, for example, m It is 3. This is a normal mean vector. The principal eigenvectors corresponding to the principal eigenvalues; These are the coordinates of the historical normal state after dimensionality reduction. This step projects the high-dimensional historical normal operation data into a low-dimensional space composed of principal feature vectors.

[0026] The fault prediction method for air conditioners provided in this invention can understand the interrelationship of various operating parameters by calculating the normal mean vector and covariance matrix. Then, by performing eigenvalue decomposition on the covariance matrix and selecting the principal eigenvalues ​​and their corresponding principal eigenvectors, the most significant variation direction in the data can be identified, retaining the principal components that contribute significantly, achieving data dimensionality reduction, and obtaining the dimensionality-reduced historical normal state coordinates. This provides a stable benchmark for subsequent fault prediction, thereby improving the accuracy and robustness of fault prediction.

[0027] In this embodiment of the invention, see appendix. Figure 4 Step S174, the step of performing dimensionality reduction processing on the historical fault operation dataset to obtain the dimensionality-reduced historical fault state coordinates, includes: S1740, based on the historical fault operation dataset, obtains a fault mean vector composed of the average values ​​of each operating parameter; satisfying the formula: ; S1742, based on the fault mean vector and the principal feature vector, obtains the dimensionality-reduced historical fault state coordinates, satisfying the formula: ; in, D For historical fault running datasets, The fault mean vector, This represents the average of various operating parameters during historical fault scenarios. n The number of running parameters, r The number of learning parameters for each running parameter during historical faults, for example, r It is 200; m The number of principal eigenvalues The principal eigenvectors corresponding to the principal eigenvalues; This step projects the high-dimensional historical fault state coordinates onto a low-dimensional space defined by the historical normal operation data, thus obtaining the dimensionality-reduced historical state coordinates.

[0028] The air conditioner fault prediction method provided in this embodiment of the invention first obtains the fault mean vector based on the historical fault operation dataset when performing dimensionality reduction processing on the historical fault operation dataset. Then, it uses the fault mean vector and the main feature vector extracted from the historical normal operation dataset to obtain the dimensionality-reduced historical fault state coordinates, so that the historical fault state coordinates and the historical normal state coordinates are in the same feature representation. The same dimensionality reduction method improves the accuracy and robustness of the fault prediction model training and avoids the difficulty of fault feature identification caused by feature space mismatch.

[0029] In this embodiment of the invention, step S176, the step of obtaining the second unit deviation vector based on historical fault state coordinates and historical normal state coordinates, satisfies the formula: ; in, The second unit deviation vector, This is a vector connecting the coordinates of historical normal states and historical fault states. This refers to the magnitude of the vector connecting the historical normal state coordinates and the historical fault state coordinates. In this step, the second unit deviation vector is a vector with a specific direction and a magnitude of 1, reflecting the direction information of the deviation.

[0030] The air conditioner fault prediction method provided in this invention quantifies the deviation between historical fault state coordinates and historical normal state coordinates into a second unit deviation vector, thereby achieving a standardized expression of the fault direction. This unitization eliminates the difference in the magnitude of the offset between different fault samples of the same fault type, allowing the machine learning algorithm to focus on the directional features of the fault when training the fault prediction model, thus improving the accuracy and robustness of the model in identifying different fault types. In addition, the features input to the model are unit vectors, which simplifies the model training process and improves the model's learning efficiency.

[0031] In this embodiment of the invention, the normal operation data is a historical normal operation dataset; see appendix. Figure 5 Step S140, the step of obtaining the first unit deviation vector based on the current running data and the normal running data, includes: S142, perform dimensionality reduction on the current running data to obtain the dimensionality-reduced coordinates of the current running state; S144: Based on the current operating state coordinates and historical normal state coordinates, the first unit deviation vector is obtained. This setting, by reducing the dimensionality of the current operating state coordinates so that they are in the same reduced-dimensional space as the historical normal state coordinates, ensures that the calculated first unit deviation vector has high consistency with the second unit deviation vector during training. This improves the accuracy and reliability of fault precursor judgment, enabling the fault estimation model to more accurately identify potential fault types.

[0032] In this embodiment of the invention, see appendix. Figure 6 Step S142, the step of performing dimensionality reduction processing on the current running data to obtain the dimensionality-reduced current running state coordinates, is as follows: S1422, based on the current running data and the main feature vector, obtain the current running state coordinates after dimensionality reduction, satisfying the formula: ; in, X For current running data, P m The current running state coordinates are used to transform the high-dimensional current running data into a low-dimensional space, forming the dimensionality-reduced current running state coordinates. By using the principal feature vectors determined during the training process, the current running data is projected into the principal component space defined by the principal feature vectors, thereby obtaining the key features of the current running state and ensuring that the current running state coordinates are comparable to the historical normal state coordinates in the same low-dimensional space.

[0033] Step S144, the step of obtaining the first unit deviation vector based on the current operating state coordinates and the historical normal state coordinates, includes: S1442, based on the current operating state coordinates and the historical normal state coordinates, obtains the first deviation vector; satisfying the formula: ; S1444, based on the first deviation vector and its magnitude, obtain the first unit deviation vector; satisfying the formula: ; in, The deviation direction vector, The first unit deviation vector, The first unit deviation vector is the magnitude of the deviation vector. In this step, the first unit deviation vector is a vector with a specific direction and a magnitude of 1, reflecting the directional information of the deviation from the current operating state. This eliminates the influence of the deviation magnitude on the fault type determination, enabling faults with different degrees of deviation but the same direction to be identified, thereby improving the robustness of the fault estimation model to fault direction features.

[0034] The air conditioner fault prediction method provided in this embodiment of the invention projects the current operating data into a low-dimensional space trained from historical normal operating data by multiplying the current operating data with the main feature vector. This results in the dimensionality-reduced coordinates of the current operating state, ensuring that the current operating state and the historical normal state are compared in the same low-dimensional space, thus avoiding errors caused by spatial inconsistency. Furthermore, by calculating and normalizing the first deviation vector to obtain the first unit deviation vector, the interference of the deviation magnitude on the fault type determination is eliminated. This allows the fault type determined based on the preset mapping relationship to focus more on the directional features of the deviation, thereby improving the accuracy and robustness of fault prediction.

[0035] In this embodiment of the invention, see appendix. Figure 7 Step S120, the step of determining whether the air conditioner has signs of a malfunction based on the current operating data and the normal operating data, includes: S122, based on the current running data and the normal running data, the anomaly degree is obtained; where the anomaly degree is used to quantify the degree of deviation between the current running state and the normal running state. This anomaly degree can be obtained in a variety of ways, such as calculating the distance between the coordinates of the current running state and the coordinates of the historical normal state in the low-dimensional space after dimensionality reduction (such as Euclidean distance or Mahalanobis distance). The larger the distance, the higher the anomaly degree.

[0036] S124, Based on the anomaly level, determine whether the air conditioner shows signs of a malfunction. This step can use various methods for judgment. For example, if the anomaly level is greater than or equal to a preset anomaly level value, then the air conditioner is determined to have signs of a malfunction. Another example is to obtain the anomaly level change rate based on the anomaly level; if the anomaly level change rate is greater than or equal to a preset change rate value, then the air conditioner is determined to have signs of a malfunction. This setting, by introducing anomaly levels, quantifies the degree to which the current operating data deviates from the normal state, improving the sensitivity and accuracy of malfunction prediction and helping to detect potential malfunctions in their early stages.

[0037] In this embodiment of the invention, see appendix. Figure 8 Step S160, the step of determining the fault type based on the first unit deviation vector, includes: S162, using a preset mapping relationship, perform mapping operations on the first unit deviation vector and output the fault confidence level corresponding to each fault type; where the fault confidence level represents the possibility or probability of a specific fault type existing given the first unit deviation vector. S164, Determine the fault type based on the fault confidence level and a preset fault threshold. Specifically, if the fault confidence level exceeds the preset fault threshold, the existence of that fault type is confirmed. The preset fault threshold can be set based on experience or actual needs. For example, if the number of fault types is 9, it is defined as follows: FR1: Refrigerant leak; FR2: Compressor deterioration; FR3: Outdoor fan motor deterioration; FR4: Indoor fan motor deterioration; FR5: Outdoor heat exchanger blockage; FR6: Indoor heat exchanger blockage; FR7: External expansion valve malfunction; FR8: Internal expansion valve malfunction; FR9: Oil return solenoid valve malfunction. The preset fault threshold is 0.8. The fault confidence scores for each fault type are output as follows: (FR1,FR2,FR3,FR4,FR5,FR6,FR7,FR8,FR9) = (0.1, 0.9, 0.2, 0.1, 0.1, 0.2,0.3, 0.1, 0.1); then the fault type is compressor performance degradation.

[0038] The fault prediction method for air conditioners provided in this embodiment of the invention outputs the fault confidence level corresponding to each fault type and compares and judges it with a preset fault threshold, thereby quantitatively determining the fault type, improving the reliability of fault judgment, and providing a basis for subsequent maintenance and repair.

[0039] This invention also provides a fault prediction system for air conditioners, applying the aforementioned fault prediction method for air conditioners (see attached diagram). Figure 9 The air conditioner malfunction prediction system includes: Module 10 is used to acquire the current operating data of the air conditioner; The judgment module 20 is used to determine whether the air conditioner has any signs of malfunction based on the current operating data and the normal operating data; Calculation module 30 is used to obtain a first unit deviation vector based on the current fault operation data and the historical normal operation dataset if fault precursors exist; and, The fault type determination module 40 is used to determine the fault type based on the first unit deviation vector, wherein the fault type and the first unit deviation vector have a preset mapping relationship.

[0040] The air conditioner fault prediction system provided in this embodiment of the invention applies the above-mentioned air conditioner fault prediction method and has all the advantages of the method, which will not be repeated here.

[0041] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for determining the early signs of an air conditioner malfunction, characterized in that, The method for determining the fault signs of the air conditioner includes the following steps: Obtain the current operating data of the air conditioner; Based on the current operating data and the normal operating data, determine whether the air conditioner shows any signs of malfunction; If there are signs of a fault, a first unit deviation vector is obtained based on the current operating data and the normal operating data; Based on the first unit deviation vector, the fault type is determined, wherein the fault type and the first unit deviation vector have a preset mapping relationship.

2. The method for determining the fault signs of an air conditioner according to claim 1, characterized in that, The preset mapping relationship is obtained based on the training of a fault prediction model, and the training method of the fault prediction model includes: Obtain the historical normal operation dataset and the historical fault operation dataset of the air conditioner. The historical normal operation dataset includes data of multiple operating parameters at different times under normal operating conditions, and the historical fault operation dataset includes data of multiple operating parameters corresponding to multiple fault types at different times. The historical normal operation dataset is subjected to dimensionality reduction processing to obtain the dimensionality-reduced historical normal state coordinates; The historical fault operation dataset is subjected to dimensionality reduction processing to obtain the dimensionality-reduced historical fault state coordinates; Based on the historical fault state coordinates and the historical normal state coordinates, a second unit deviation vector is obtained; Using a machine learning algorithm, based on the second unit deviation vector and the fault type, the preset mapping relationship is fitted to obtain the trained fault prediction model.

3. The method for determining the fault signs of an air conditioner according to claim 2, characterized in that, The step of performing dimensionality reduction processing on the historical normal operation dataset to obtain the dimensionality-reduced historical normal state coordinates includes: Based on the historical normal operation dataset, a normal mean vector and covariance matrix composed of the data mean values ​​of each operating parameter are obtained; satisfying the formula: ； in, Z For historically normal operating datasets, This is a normal mean vector. This represents the average of the operating parameters during historical normal operation. n The number of running parameters, k This represents the number of learning parameters for each operating parameter during historical normal operation. Let covariance matrix be the variance matrix. For the first i Normal operating data for each operating parameter. For the first j Normal operating data for each operating parameter. For the first i The average data of each operating parameter i =1,2,3,…, n ; For the first j The average data of each operating parameter j =1,2,3,…, n ; The covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues ​​and corresponding eigenvectors; satisfying the formula: ； in, For eigenvalues, ; For feature vectors, , t The number of eigenvalues; The feature values ​​are sorted from largest to smallest, and the feature values ​​at the top of the sort are selected as the principal feature values. Based on the normal mean vector and the principal eigenvector corresponding to the principal eigenvalue, the dimensionality-reduced historical normal state coordinates are obtained, satisfying the formula: ； in, m The number of principal eigenvalues This is a normal mean vector. The principal eigenvectors corresponding to the principal eigenvalues; These are the coordinates of the historical normal state after dimensionality reduction.

4. The method for determining the fault signs of an air conditioner according to claim 3, characterized in that, The step of performing dimensionality reduction processing on the historical fault operation dataset to obtain the dimensionality-reduced historical fault state coordinates includes: Based on the historical fault operation dataset, a fault mean vector is obtained, consisting of the average values ​​of each operating parameter; satisfying the formula: ； Based on the fault mean vector and the principal feature vector, the dimensionality-reduced historical fault state coordinates are obtained, satisfying the formula: ； in, D For historical fault running datasets, The fault mean vector, This represents the average of various operating parameters during historical fault scenarios. n The number of running parameters, r The number of times each operating parameter was learned during historical fault testing; m The number of principal eigenvalues The principal eigenvectors corresponding to the principal eigenvalues; These are the coordinates of the historical fault states after dimensionality reduction.

5. The method for determining the fault signs of an air conditioner according to claim 4, characterized in that, The step of obtaining the second unit deviation vector based on the historical fault state coordinates and the historical normal state coordinates satisfies the formula: ； in, The second unit deviation vector, This is a vector connecting the coordinates of historical normal states and historical fault states. The magnitude of the vector connecting the historical normal state coordinates and the historical fault state coordinates.

6. The method for determining the fault signs of an air conditioner according to any one of claims 3-5, characterized in that, The normal operation data is the historical normal operation dataset, and the step of obtaining the first unit deviation vector based on the current operation data and the normal operation data includes: The current running data is subjected to dimensionality reduction processing to obtain the dimensionality-reduced current running state coordinates; Based on the current operating state coordinates and the historical normal state coordinates, the first unit deviation vector is obtained.

7. The method for determining the fault signs of an air conditioner according to claim 6, characterized in that, The step of performing dimensionality reduction processing on the current running data to obtain the dimensionality-reduced current running state coordinates is as follows: Based on the current running data and the main feature vector, the dimensionality-reduced current running state coordinates are obtained, satisfying the formula: ； in, X For current running data, P m The coordinates represent the current running state. The step of obtaining the first unit deviation vector based on the current operating state coordinates and the historical normal state coordinates includes: Based on the current operating state coordinates and the historical normal state coordinates, a first deviation vector is obtained; satisfying the formula: ； Based on the first deviation vector and its magnitude, the first unit deviation vector is obtained; satisfying the formula: ; in, The deviation direction vector, The first unit deviation vector, This represents the magnitude of the deviation vector.

8. The method for determining the fault signs of an air conditioner according to any one of claims 1-5, characterized in that, The step of determining whether the air conditioner has signs of malfunction based on the current operating data and the normal operating data includes: obtaining an anomaly degree based on the current operating data and the normal operating data; and determining whether the air conditioner has signs of malfunction based on the anomaly degree. And / or, the step of determining the fault type based on the first unit deviation vector includes: Using the preset mapping relationship, the first unit deviation vector is mapped and the fault confidence level corresponding to each fault type is output. The fault type is determined based on the fault confidence level and the preset fault threshold.

9. The method for determining the fault signs of an air conditioner according to any one of claims 2-5, characterized in that, The operating parameters include at least two of the following: outer ring temperature, indoor temperature, high pressure, low pressure, suction superheat, exhaust superheat, subcooling, compressor frequency, and compressor current. And / or, the fault types include at least two of the following: refrigerant leakage, compressor deterioration, external fan motor deterioration, internal fan motor deterioration, blockage of outdoor heat exchanger, blockage of indoor heat exchanger, malfunction of external expansion valve, malfunction of internal expansion valve, and oil return solenoid valve failure; And / or, the machine learning algorithm includes either a neural network or a random forest.

10. A fault prediction system for an air conditioner, characterized in that, The fault prediction method for an air conditioner according to any one of claims 1-9, wherein the fault prediction system for the air conditioner comprises: The acquisition module is used to acquire the current operating data of the air conditioner; The determination module is used to determine whether the air conditioner shows signs of malfunction based on the current operating data and the normal operating data; The calculation module is configured to, if fault precursors exist, obtain a first unit deviation vector based on the current fault operation data and the historical normal operation dataset; and, The fault type determination module is used to determine the fault type based on the first unit deviation vector, wherein the fault type and the first unit deviation vector have a preset mapping relationship.

Images