A fault detection method, device and equipment of a power device

Abstract

This application provides a method, apparatus, and device for fault detection of power equipment. The method includes: acquiring an initial temperature value corresponding to the power equipment; for each target dimension, determining a target temperature value based on the initial temperature value, and determining a feature value of the target dimension based on the target temperature value and a target weight value of the target dimension; constructing a feature vector to be detected based on the feature values ​​of multiple target dimensions; and inputting the feature vector to be detected into a trained target machine learning model, so that the target machine learning model outputs a fault level corresponding to the power equipment. Through the technical solution of this application, machine learning technology can be used to detect the fault level corresponding to power equipment without relying on the accuracy of temperature thresholds, thereby improving the accuracy of fault detection results for power equipment.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method, apparatus and equipment for fault detection of power equipment. Background Technology

[0002] A substation is a location in a power system that transforms voltage and current, receives electrical energy, and distributes it. The electrical equipment within a substation is divided into primary power equipment and secondary power equipment. Primary power equipment refers to equipment that directly participates in the production, transformation, transmission, distribution, and use of electrical energy. Examples include generators, motors, transformers, switching devices (such as circuit breakers, disconnectors, contactors, and fuses), current-carrying conductors and gas-insulated equipment (such as busbars, power cables, insulators, and wall bushings), overcurrent and overvoltage limiting equipment (such as current-limiting reactors and surge arresters), and instrument transformers (such as voltage transformers and current transformers). Secondary power equipment refers to equipment that measures, controls, protects, and regulates primary power equipment. Examples include measuring meters, relay protection devices, measurement and control devices, metering devices, and DC power supply equipment.

[0003] For the numerous electrical devices within a substation, fault detection is necessary to promptly identify any malfunctions. When a fault occurs, the equipment must be repaired to prevent continued operation and avoid potential safety hazards.

[0004] In order to detect faults in power equipment, it is necessary to collect the temperature value of the power equipment. If the temperature value is greater than the temperature threshold, it is determined that the power equipment has failed. If the temperature value is not greater than the temperature threshold, it is determined that the power equipment has not failed. That is, the analysis of whether the power equipment has failed is based on the temperature threshold.

[0005] However, the above methods rely on the accuracy of the temperature threshold, which is configured by managers based on experience. Inaccurate temperature thresholds can lead to errors in fault detection results for electrical equipment. For example, if the temperature threshold is too high, the equipment may not be faulty even when its temperature is below the threshold. Conversely, if the temperature threshold is too low, the equipment may be faulty even when its temperature is above the threshold. Summary of the Invention

[0006] This application provides a fault detection method for power equipment, wherein the power equipment is associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the power equipment. The method includes:

[0007] Obtain the initial temperature value corresponding to the power equipment;

[0008] For each target dimension, a target temperature value is determined based on the initial temperature value, and a feature value of the target dimension is determined based on the target temperature value and the target weight value of the target dimension.

[0009] Construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions;

[0010] The feature vector to be detected is input into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment.

[0011] This application provides a fault detection device for power equipment. The power equipment is associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the power equipment. The device includes:

[0012] The acquisition module is used to acquire the initial temperature value corresponding to the power equipment;

[0013] The determination module is configured to, for each target dimension associated with the power equipment, determine a target temperature value for the target dimension based on the initial temperature value, and determine a feature value for the target dimension based on the target temperature value and the target weight value of the target dimension.

[0014] The generation module is used to construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions;

[0015] The processing module is used to input the feature vector to be detected into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment.

[0016] This application provides an electronic device, in which a power device is associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the power device. The electronic device includes a processor and a machine-readable storage medium, the machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to achieve the following steps:

[0017] Obtain the initial temperature value corresponding to the power equipment;

[0018] For each target dimension, a target temperature value is determined based on the initial temperature value, and a feature value of the target dimension is determined based on the target temperature value and the target weight value of the target dimension.

[0019] Construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions;

[0020] The feature vector to be detected is input into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment.

[0021] As can be seen from the above technical solutions, in this embodiment, target temperature values ​​of multiple target dimensions associated with power equipment can be collected. A detection feature vector is constructed based on the target temperature values ​​and target weight values ​​of multiple target dimensions. This detection feature vector is then input into a target machine learning model, which outputs the fault level corresponding to the power equipment. In other words, machine learning technology is used to detect the fault level of the power equipment without relying on the accuracy of temperature thresholds. This improves the accuracy of fault detection results (i.e., fault levels) and allows for timely detection of whether a power equipment fault has occurred. Furthermore, when a power equipment fault occurs, it can be repaired promptly, preventing the faulty equipment from continuing to operate and avoiding potential safety hazards. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments of this application or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings of the embodiments of this application.

[0023] Figure 1 This is a flowchart illustrating a fault detection method for power equipment according to one embodiment of this application.

[0024] Figure 2 This is a flowchart illustrating a fault detection method for power equipment according to one embodiment of this application.

[0025] Figure 3 This is a schematic diagram of an offline environment process in one embodiment of this application;

[0026] Figure 4 This is a schematic diagram of the training feature vector in one embodiment of this application;

[0027] Figure 5 This is a schematic diagram of the structure of a fault detection device for power equipment according to one embodiment of this application;

[0028] Figure 6 This is a hardware structure diagram of an electronic device according to one embodiment of this application. Detailed Implementation

[0029] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “the,” and “the” as used in this application and claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to any and all possible combinations comprising one or more of the associated listed items.

[0030] It should be understood that although the terms first, second, third, etc., may be used to describe various information in embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" may also be interpreted as "when," "when," or "in response to a determination."

[0031] This application proposes a fault detection method for power equipment. This method detects the corresponding fault level of the power equipment. The fault level can be two-level, namely a normal level and an abnormal level. Alternatively, it can be multi-level; for example, a four-level fault level could be normal, moderate, severe, and emergency. Of course, other numbers of fault levels can also be used, without limitation.

[0032] In this embodiment, the power equipment can be associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the power equipment. For example, if a certain type of temperature value can reflect the fault level of the power equipment, then this type is used as the target dimension. In one possible implementation, the multiple target dimensions associated with the power equipment include, but are not limited to, at least one of the following: ambient temperature dimension; maximum temperature dimension; minimum temperature dimension; temperature difference dimension; phase-to-phase temperature dimension; temperature rise dimension; and relative temperature difference dimension.

[0033] For example, when the ambient temperature value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the ambient temperature dimension. When the highest temperature value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the highest temperature dimension. When the lowest temperature value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the lowest temperature dimension. When the temperature difference value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the temperature difference dimension. When the phase-to-phase temperature value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the phase-to-phase temperature dimension. When the temperature rise value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the temperature rise dimension. When the relative temperature difference value of electrical equipment can reflect the fault level of the equipment, the target dimension includes the relative temperature difference dimension.

[0034] See Figure 1 The diagram shows a flowchart of a fault detection method for power equipment, which may include:

[0035] Step 101: Obtain the initial temperature value corresponding to the power equipment.

[0036] For example, the initial temperature value may include the initial ambient temperature value corresponding to the power equipment, multiple measured temperature values ​​corresponding to the power equipment, and the average temperature value corresponding to the multiple measured temperature values.

[0037] In one possible implementation, in each acquisition cycle, an initial temperature value corresponding to the power equipment can be acquired. This initial temperature value may include the initial ambient temperature value of the acquisition cycle, multiple measured temperature values ​​of the acquisition cycle, and the average temperature value corresponding to the multiple measured temperature values.

[0038] Step 102: For each target dimension, determine the target temperature value of the target dimension based on the initial temperature value, and determine the feature value of the target dimension based on the target temperature value and the target weight value of the target dimension.

[0039] For example, if the target dimension is the ambient temperature dimension, the target ambient temperature value of the target dimension can be determined based on the initial ambient temperature value (i.e., the target temperature value includes the target ambient temperature value).

[0040] If the target dimension is the highest temperature dimension, the target maximum temperature value of the target dimension can be determined based on the maximum measured temperature value among multiple measured temperature values ​​(i.e., the target temperature value includes the target maximum temperature value).

[0041] If the target dimension is the lowest temperature dimension, the target minimum temperature value of the target dimension can be determined based on the minimum measured temperature value among multiple measured temperature values ​​(i.e., the target temperature value includes the target minimum temperature value).

[0042] If the target dimension is a temperature difference dimension, the target temperature difference value of the target dimension can be determined based on the target maximum temperature value and the target minimum temperature value (i.e., the target temperature value includes the target temperature difference value).

[0043] If the target dimension is the interphase temperature dimension, the target interphase temperature value of the target dimension can be determined based on the average temperature value corresponding to multiple measured temperature values ​​(i.e., the target temperature value includes the target interphase temperature value).

[0044] If the target dimension is the temperature rise dimension, the target temperature rise value of the target dimension in the current acquisition period can be determined based on the average temperature value of the current acquisition period and the average temperature value of the previous acquisition period (i.e., the acquisition period before the current acquisition period).

[0045] If the target dimension is a relative temperature difference dimension, the target relative temperature difference value of the target dimension in the current acquisition period can be determined based on the target highest temperature value of the current acquisition period and the target highest temperature value of the target acquisition period (i.e., the target temperature value includes the target relative temperature difference value). The target highest temperature value of the current acquisition period is determined based on multiple measured temperature values ​​of the current acquisition period (i.e., the maximum value among multiple measured temperature values), and the target highest temperature value of the target acquisition period is determined based on multiple measured temperature values ​​of the target acquisition period. The fault level corresponding to the target acquisition period can be a normal level. For example, if the fault level corresponding to the previous acquisition period is a normal level, then the target acquisition period is the previous acquisition period. If the fault level corresponding to the previous acquisition period is not a normal level, and the fault level corresponding to the second acquisition period preceding the current acquisition period is a normal level, then the target acquisition period is the second acquisition period preceding the current acquisition period, and so on.

[0046] In one possible implementation, prior to step 102, a target weight value for the target dimension can also be determined. For example, the target weight value for the target dimension can be determined as follows: obtain the configured initial weight value of the target dimension (the initial weight value is configured based on experience). If the target dimension is a first-type target dimension, then the initial weight value is determined as the target weight value for the target dimension; or, if the target dimension is a second-type target dimension, then the initial weight value is reduced, and the reduced initial weight value is determined as the target weight value for the target dimension; or, if the target dimension is a third-type target dimension, then the initial weight value is increased, and the increased initial weight value is determined as the target weight value for the target dimension.

[0047] For example, the first type of target dimension may include, but is not limited to, at least one of the following: maximum temperature dimension, minimum temperature dimension, temperature difference dimension, and phase temperature dimension; the second type of target dimension may include, but is not limited to, the ambient temperature dimension; and the third type of target dimension may include, but is not limited to, the temperature rise dimension and / or the relative temperature difference dimension.

[0048] Step 103: Construct the feature vector to be detected based on the feature values ​​of multiple target dimensions. For example, the feature values ​​of multiple target dimensions can be combined into a matrix, and this matrix can be used as the feature vector to be detected.

[0049] Step 104: Input the feature vector to be detected into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment. For example, the target machine learning model is used to detect the fault level corresponding to the power equipment. Therefore, after inputting the feature vector to be detected into the target machine learning model, the target machine learning model can process the feature vector to be detected. There are no restrictions on this processing method, thereby obtaining the fault level corresponding to the power equipment. For example, the fault level can be a normal level, a general level, a severe level, or an emergency level.

[0050] In one possible implementation, the target machine learning model includes, but is not limited to: a machine learning model based on the GBDT (Gradient Boosting Decision Tree) algorithm, a machine learning model based on the KNN (K-Nearest Neighbor) algorithm, and a machine learning model based on the SVM (Support Vector Machines) algorithm. Of course, the above are just a few examples of target machine learning models, and this embodiment does not limit the type of target machine learning model.

[0051] In one possible implementation, the target machine learning model can be trained before step 101, see [link to relevant documentation]. Figure 2 As shown, the following steps can be taken to train the target machine learning model:

[0052] Step 201: Obtain the training temperature value and training fault level corresponding to the power equipment.

[0053] For example, the training temperature value may include the initial ambient temperature value corresponding to the power equipment, multiple measured temperature values ​​corresponding to the power equipment, and the average temperature value corresponding to the multiple measured temperature values.

[0054] In one possible implementation, in each acquisition cycle, a training temperature value corresponding to the power equipment can be acquired. The training temperature value may include the initial ambient temperature value of the acquisition cycle, multiple measured temperature values ​​of the acquisition cycle, and the average temperature value corresponding to the multiple measured temperature values.

[0055] Step 202: For each target dimension, determine the target temperature value for that target dimension based on the training temperature value, and determine the feature value for that target dimension based on the target temperature value and the target weight value for that target dimension.

[0056] For example, if the target dimension is the ambient temperature dimension, the target ambient temperature value of the target dimension can be determined based on the initial ambient temperature value (i.e., the target temperature value includes the target ambient temperature value).

[0057] If the target dimension is the highest temperature dimension, the target maximum temperature value of the target dimension can be determined based on the maximum measured temperature value among multiple measured temperature values ​​(i.e., the target temperature value includes the target maximum temperature value).

[0058] If the target dimension is the lowest temperature dimension, the target minimum temperature value of the target dimension can be determined based on the minimum measured temperature value among multiple measured temperature values ​​(i.e., the target temperature value includes the target minimum temperature value).

[0059] If the target dimension is a temperature difference dimension, the target temperature difference value of the target dimension can be determined based on the target maximum temperature value and the target minimum temperature value (i.e., the target temperature value includes the target temperature difference value).

[0060] If the target dimension is the interphase temperature dimension, the target interphase temperature value of the target dimension can be determined based on the average temperature value corresponding to multiple measured temperature values ​​(i.e., the target temperature value includes the target interphase temperature value).

[0061] If the target dimension is the temperature rise dimension, the target temperature rise value of the target dimension in the current acquisition period can be determined based on the average temperature value of the current acquisition period and the average temperature value of the previous acquisition period (i.e., the acquisition period before the current acquisition period).

[0062] If the target dimension is a relative temperature difference dimension, the target relative temperature difference value of the target dimension in the current acquisition period can be determined based on the target highest temperature value in the current acquisition period and the target highest temperature value in the target acquisition period (i.e., the target temperature value includes the target relative temperature difference value). The target highest temperature value in the current acquisition period is determined based on multiple measured temperature values ​​in the current acquisition period (i.e., the maximum value among multiple measured temperature values), and the target highest temperature value in the target acquisition period is determined based on multiple measured temperature values ​​in the target acquisition period. Furthermore, the training fault level corresponding to the target acquisition period can be a normal level.

[0063] In one possible implementation, prior to step 202, the target weight value of the target dimension can be determined as follows: The configured initial weight value of the target dimension is obtained. If the target dimension is a first-type target dimension, the initial weight value is determined as the target weight value of the target dimension; or, if the target dimension is a second-type target dimension, the initial weight value is decreased, and the decreased initial weight value is determined as the target weight value of the target dimension; or, if the target dimension is a third-type target dimension, the initial weight value is increased, and the increased initial weight value is determined as the target weight value of the target dimension. For example, the first-type target dimensions include, but are not limited to, at least one of the following: maximum temperature dimension, minimum temperature dimension, temperature difference dimension, and phase temperature dimension; the second-type target dimensions include, but are not limited to, ambient temperature dimension; and the third-type target dimensions include, but are not limited to, temperature rise dimension and / or relative temperature difference dimension.

[0064] Step 203: Construct a training feature vector based on the feature values ​​of multiple target dimensions. For example, the feature values ​​of multiple target dimensions can be combined into a matrix, and this matrix can be used as the training feature vector.

[0065] Step 204: Input the training feature vector and the training fault level into the configured initial machine learning model to train the initial machine learning model and obtain the trained target machine learning model.

[0066] For example, the training fault level can be used as the label value of the training feature vector. The initial machine learning model can be trained based on the training feature vector and the training fault level. There are no restrictions on this training process. The machine learning model that has been trained can be called the target machine learning model.

[0067] In one possible implementation, the initial machine learning model includes, but is not limited to: a machine learning model based on the GBDT algorithm, a machine learning model based on the KNN algorithm, and a machine learning model based on the SVM algorithm. Of course, the above are just a few examples of initial machine learning models and are not intended to be limiting.

[0068] As can be seen from the above technical solutions, in this embodiment, target temperature values ​​of multiple target dimensions associated with power equipment can be collected. A detection feature vector is constructed based on the target temperature values ​​and target weight values ​​of multiple target dimensions. This detection feature vector is then input into a target machine learning model, which outputs the fault level corresponding to the power equipment. In other words, machine learning technology is used to detect the fault level of the power equipment without relying on the accuracy of temperature thresholds. This improves the accuracy of fault detection results (i.e., fault levels) and allows for timely detection of whether a power equipment fault has occurred. Furthermore, when a power equipment fault occurs, it can be repaired promptly, preventing the faulty equipment from continuing to operate and avoiding potential safety hazards.

[0069] The technical solutions of the embodiments of this application will be described below in conjunction with specific application scenarios.

[0070] This application proposes a fault detection method for power equipment. This method is used to detect the corresponding fault level of power equipment, also known as electrical equipment. It can be power equipment within a substation or power equipment in other scenarios; there are no restrictions. The following example focuses on power equipment within a substation. In this embodiment, fault detection can be performed on primary or secondary power equipment within a substation. For example, fault detection can be performed on generators, motors, transformers, switchgear, current-carrying conductors and gas-insulated equipment, overcurrent and overvoltage limiting equipment, instrument transformers, measuring meters, relay protection devices, measurement and control devices, metering devices, DC power supply equipment, and other power equipment. There are no restrictions on the type of power equipment requiring fault detection. For instance, the fault detection method can be used to detect faults in transformers, such as detecting faults in the bushing terminals of a transformer.

[0071] The fault levels corresponding to power equipment can be two-level fault levels, three-level fault levels, four-level fault levels, or more fault levels, without any restrictions. Taking the four-level fault level as an example, the four-level fault level can be normal level, general level, severe level, or emergency level.

[0072] In this embodiment, there may be offline and production environment processes. In the offline process, a target machine learning model needs to be trained. In the production environment process, the fault level of the power equipment needs to be detected based on the trained target machine learning model. The execution entity of the offline process can be called a training device, and the execution entity of the production environment process can be called a management device. The training device and the management device can be the same device or different devices; there is no limitation on this.

[0073] See Figure 3 The diagram shown illustrates the offline environment process, which may include data acquisition, data governance, feature engineering, model training, and model deployment.

[0074] First, the data acquisition process. During data acquisition, it is necessary to obtain the training temperature value and training fault level corresponding to the power equipment (e.g., the bushing head of a transformer). For example, in each acquisition cycle, the training temperature value and training fault level corresponding to the power equipment can be obtained. Taking a 1-second acquisition cycle as an example, in the 1st second, the training temperature value a1 and training fault level b1 corresponding to the power equipment are obtained; in the 2nd second, the training temperature value a2 and training fault level b2 are obtained; in the 3rd second, the training temperature value a3 and training fault level b3 are obtained, and so on.

[0075] For example, the training temperature values ​​may include the initial ambient temperature value corresponding to the electrical equipment, multiple measured temperature values ​​corresponding to the electrical equipment, and the average temperature value corresponding to the multiple measured temperature values. For instance, training temperature value a1 may include the initial ambient temperature value a1-1, measured temperature value a1-2-1, measured temperature value a1-2-2, measured temperature value a1-2-3, measured temperature value a1-2-4, and average temperature value a1-3. Training temperature value a2 may include the initial ambient temperature value a2-1, measured temperature value a2-2-1, measured temperature value a2-2-2, measured temperature value a2-2-3, measured temperature value a2-2-4, and average temperature value a2-3, and so on.

[0076] For example, regarding the method of obtaining the initial ambient temperature value, temperature measuring devices (such as thermal imaging devices, infrared temperature measuring devices, etc.) can be deployed on the power equipment. In each acquisition cycle, the temperature value of the environment in which the power equipment is located is collected by the temperature measuring device. This temperature value is called the initial ambient temperature value. The initial ambient temperature value of each acquisition cycle is sent to the training device, which collects the initial ambient temperature value.

[0077] Of course, the above is just an example of obtaining the initial ambient temperature value, and there are no limitations on this.

[0078] For example, regarding the method of obtaining the measured temperature value, in each acquisition cycle, the temperature value of multiple physical locations (i.e., locations in the actual physical space) of the power equipment is collected by the temperature measuring device. The temperature value of each physical location is called the measured temperature value. There is no restriction on the method of collecting this measured temperature value. The measured temperature value of each acquisition cycle is sent to the training device, which collects the measured temperature value.

[0079] Of course, the above is just an example of obtaining measured temperature values, and there are no limitations on this.

[0080] For example, regarding the method of obtaining the average temperature value, in each acquisition cycle, the average value of all measured temperature values ​​in that acquisition cycle can be calculated, and this average value is the average temperature value of that acquisition cycle.

[0081] In one possible implementation, all physical locations of the electrical equipment can be assigned to at least one temperature measurement point (a temperature measurement point can also be called a temperature measurement frame or temperature measurement set), and there is no limitation on this assignment method. For example, a transformer bushing head has three faces. All physical locations of the first face can be assigned to temperature measurement point c1, all physical locations of the second face to temperature measurement point c2, and all physical locations of the third face to temperature measurement point c3. Of course, the above assignment method is just an example.

[0082] Based on this, the initial ambient temperature values ​​include the initial ambient temperature values ​​at measurement points c1, c2, and c3. These three initial ambient temperature values ​​may be the same or different; assuming they are the same, we will take this as an example. The measured temperature values ​​include the measured temperature values ​​at the physical locations corresponding to measurement points c1 (which can be multiple), c2 (which can be multiple), and c3 (which can be multiple). The average temperature values ​​include the average temperature value corresponding to measurement point c1 (i.e., the average of all measured temperature values ​​corresponding to measurement point c1), the average temperature value corresponding to measurement point c2, and the average temperature value corresponding to measurement point c3.

[0083] For example, regarding the method of obtaining training fault levels, in each collection period, the training fault level of that collection period can be obtained. For instance, the user knows the training fault level of the power equipment in that collection period and provides the training fault level, such as normal level, general level, severe level, or emergency level. Of course, the above is just an example of obtaining training fault levels and is not limited thereto.

[0084] In summary, the training temperature value and training fault level can be obtained, as shown in Table 1.

[0085] Table 1

[0086]

[0087] Second, the data governance process (i.e., preprocessing). During data governance, training temperature values ​​and training fault levels can be analyzed to identify and remove invalid and redundant data. For example, assuming the maximum measured temperature value is 100 degrees Celsius, if a measured temperature value is greater than 100 degrees Celsius, it is considered invalid data and needs to be removed. Similarly, assuming the minimum measured temperature value is 20 degrees Celsius, if a measured temperature value is less than 20 degrees Celsius, it is considered invalid data and needs to be removed. Furthermore, assuming the initial ambient temperature range is [5, 50], if an initial ambient temperature value does not fall within this range, it is considered invalid data and needs to be removed. Of course, these are just a few examples, and the data governance process is not limited to these examples.

[0088] Third, the feature engineering process. During feature engineering, for each target dimension associated with the power equipment, a target temperature value for that target dimension can be determined based on the training temperature value. This target dimension's temperature value reflects the fault level of the power equipment; that is, the fault level of the power equipment can be determined through the temperature value of the target dimension. In one possible implementation, the multiple target dimensions associated with the power equipment may include, but are not limited to, at least one of the following: ambient temperature dimension; maximum temperature dimension; minimum temperature dimension; temperature difference dimension; phase-to-phase temperature dimension; temperature rise dimension; and relative temperature difference dimension. Of course, in practical applications, other types of target dimensions may also exist; this is not a limitation. The above seven target dimensions are used as an example for illustration.

[0089] 1. Ambient temperature dimension: The target temperature value is the target ambient temperature value. In this case, the target ambient temperature value can be determined based on the initial ambient temperature value. For example, if the initial ambient temperature value is a1-1 in sampling period 1, then the target ambient temperature value for sampling period 1 is a1-1. If the initial ambient temperature value is a2-1 in sampling period 2, then the target ambient temperature value for sampling period 2 is a2-1, and so on.

[0090] For example, an initial weight value can be configured for each target dimension. The initial weight values ​​for different target dimensions can be the same or different. There is no restriction on the initial weight value. In the following example, we will take the initial weight value of all target dimensions as the same, and denote it as w, that is, the initial weight value of the ambient temperature dimension is w.

[0091] Since the ambient temperature is relatively stable and there is no large diurnal temperature difference that would affect the temperature measurement results, the initial weight value w of the ambient temperature dimension can be reduced to obtain the reduced weight value w1. The weight value w1 is then used as the target weight value of the ambient temperature dimension, meaning that the target weight value w1 is less than the initial weight value w.

[0092] 2. Maximum Temperature Dimension: This refers to the target temperature value being the highest possible temperature. In this case, the target maximum temperature value can be determined based on the maximum measured temperature value among multiple measured temperature values. For example, in acquisition cycle 1, the maximum value is selected from all measured temperature values ​​corresponding to temperature measurement point c1 (such as a1-2-1, a1-2-2, a1-2-3) as the target maximum temperature value for temperature measurement point c1; the maximum value is selected from all measured temperature values ​​corresponding to temperature measurement point c2 as the target maximum temperature value for temperature measurement point c2; the maximum value is selected from all measured temperature values ​​corresponding to temperature measurement point c3 as the target maximum temperature value for temperature measurement point c3, and so on.

[0093] For example, if the highest temperature dimension has no strong correlation with the final detection result and is only used as a reference parameter in the training process, the initial weight value w of the highest temperature dimension can be maintained to obtain the weight value w2. The weight value w2 can then be used as the target weight value of the highest temperature dimension, that is, the target weight value w2 is equal to the initial weight value w.

[0094] 3. Lowest temperature dimension: This refers to the target temperature value being the lowest possible temperature. In this case, the target lowest temperature value can be determined based on the minimum measured temperature value among multiple measured temperature values. For example, in acquisition cycle 1, the minimum value is selected from all measured temperature values ​​corresponding to temperature measurement point c1 (such as a1-2-1, a1-2-2, a1-2-3) as the target lowest temperature value for temperature measurement point c1; the minimum value is selected from all measured temperature values ​​corresponding to temperature measurement point c2 as the target lowest temperature value for temperature measurement point c2; the minimum value is selected from all measured temperature values ​​corresponding to temperature measurement point c3 as the target lowest temperature value for temperature measurement point c3, and so on.

[0095] For example, if the lowest temperature dimension has no strong correlation with the final detection result and is only used as a reference parameter in the training process, the initial weight value w of the lowest temperature dimension can be maintained to obtain the weight value w3. The weight value w3 can then be used as the target weight value of the lowest temperature dimension, that is, the target weight value w3 is equal to the initial weight value w.

[0096] 4. Temperature difference dimension: In this case, the target temperature value is the target temperature difference value. The target temperature difference value can be determined based on the target maximum and minimum temperature values. For example, in acquisition cycle 1, the difference between the target maximum and minimum temperature values ​​at measurement point c1 is taken as the target temperature difference value for measurement point c1; the difference between the target maximum and minimum temperature values ​​at measurement point c2 is taken as the target temperature difference value for measurement point c2; the difference between the target maximum and minimum temperature values ​​at measurement point c3 is taken as the target temperature difference value for measurement point c3, and so on.

[0097] For example, if the temperature difference dimension is not strongly correlated with the final detection result and is only used as a reference parameter in the training process, the initial weight value w of the temperature difference dimension can be maintained to obtain the weight value w4. The weight value w4 can then be used as the target weight value of the temperature difference dimension, that is, the target weight value w4 is equal to the initial weight value w.

[0098] 5. Interphase temperature dimension: In this case, the target temperature value is the interphase temperature value, which can be determined based on the average temperature values ​​of multiple measurement points. For example, in acquisition cycle 1, the target interphase temperature value can be calculated based on the average temperature values ​​a1-3-1 of measurement point c1, a1-3-2 of measurement point c2, and a1-3-3 of measurement point c3. For instance, the absolute values ​​of the differences between a1-3-1 and a1-3-2, a1-3-3, and a1-3-3 can be calculated, and the maximum value among these three absolute values ​​is taken as the target interphase temperature value.

[0099] For example, if the interphase temperature dimension has no strong correlation with the final detection result and is only used as a reference parameter in the training process, the initial weight value w of the interphase temperature dimension can be maintained to obtain the weight value w5. The weight value w5 can then be used as the target weight value of the interphase temperature dimension, that is, the target weight value w5 is equal to the initial weight value w.

[0100] 6. Temperature Rise Dimension: In this case, the target temperature value is the target temperature rise value. The target temperature rise value for the current acquisition cycle can be determined based on the average temperature value of the current acquisition cycle and the average temperature value of the previous acquisition cycle. For example, in acquisition cycle 1, the absolute value of the difference between the average temperature value a1-3-1 of measurement point c1 and the default temperature value (if there is no previous acquisition cycle, the default temperature value is used as the average temperature value of the previous acquisition cycle, such as 0) is taken as the target temperature rise value for measurement point c1. Similarly, the target temperature rise values ​​for measurement points c2 and c3 can be obtained. In acquisition cycle 2, the absolute value of the difference between the average temperature value a2-3-1 of measurement point c1 (i.e., the average temperature value of acquisition cycle 2) and the average temperature value a1-3-1 of measurement point c1 (i.e., the average temperature value of acquisition cycle 1) is taken as the target temperature rise value for measurement point c1. Similarly, the target temperature rise values ​​for measurement points c2 and c3 can be obtained, and so on.

[0101] For example, if the temperature rise dimension (the temperature rise value corresponding to the same temperature measurement point) is positively correlated with the temperature measurement result, and the temperature rise dimension is an important indicator of temperature anomaly or temperature trend towards anomaly, then the initial weight value w of the temperature rise dimension can be increased to obtain the increased weight value w6. The weight value w6 is then used as the target weight value of the temperature rise dimension, that is, the target weight value w6 is greater than the initial weight value w.

[0102] 7. Relative temperature difference dimension, where the target temperature value is the target relative temperature difference value. In this case, the target relative temperature difference value for the current acquisition period can be determined based on the target maximum temperature value of the current acquisition period and the target maximum temperature value of the target acquisition period. For example, it can be determined based on the target maximum temperature value of the current acquisition period, the target maximum temperature value of the target acquisition period, and the target ambient temperature value. For instance, the target relative temperature difference value can be calculated using the following formula: r = (T1 - T2) / (T1 - T0) * 100%. In the above formula, r represents the target relative temperature difference value, T1 represents the target maximum temperature value of the current acquisition period, T2 represents the target maximum temperature value of the target acquisition period, and T0 represents the target ambient temperature value.

[0103] Clearly, for each acquisition cycle, the target relative temperature difference of temperature point c1 in the current acquisition cycle can be calculated based on the target maximum temperature value of temperature point c1 in the target acquisition cycle, the target maximum temperature value of temperature point c1 in the target acquisition cycle, and the target ambient temperature value. Similarly, the target relative temperature difference of temperature point c2 in the current acquisition cycle can be calculated based on the target maximum temperature value of temperature point c2 in the target acquisition cycle, the target maximum temperature value of temperature point c3 in the target acquisition cycle, and the target ambient temperature value.

[0104] Assuming the current acquisition period is acquisition period 6, the training fault level corresponding to acquisition period 5 (i.e., the acquisition period preceding acquisition period 6) is determined. If the training fault level is normal, acquisition period 5 is taken as the target acquisition period. If the training fault level is not normal, the training fault level corresponding to acquisition period 4 is determined. If the training fault level is normal, acquisition period 4 is taken as the target acquisition period. If the training fault level is not normal, the training fault level corresponding to acquisition period 3 is determined, and so on, until an acquisition period with a normal level is found as the target acquisition period.

[0105] For example, if the relative temperature difference dimension is an important indicator for assessing the temperature anomaly of power equipment, has accuracy for normal environmental scenarios, and helps improve the generalization ability of the model, then the initial weight value w of the relative temperature difference dimension can be increased to obtain the increased weight value w7. The weight value w7 is then used as the target weight value of the relative temperature difference dimension, that is, the target weight value w7 is greater than the initial weight value w.

[0106] In summary, during feature engineering, based on initial ambient temperature, measured temperature, and average temperature, feature engineering can be performed through feature derivation and feature selection to obtain factors and indicators that describe power equipment anomalies, namely the target temperature values ​​for the seven target dimensions mentioned above: target ambient temperature, target maximum temperature, target minimum temperature, target temperature difference, target phase-to-phase temperature, target temperature rise, and target relative temperature difference. Furthermore, the target weight value for each target dimension can be obtained.

[0107] For example, in each data acquisition cycle, for the ambient temperature dimension, the target ambient temperature value and target weight value w1 are obtained. For the highest temperature dimension, the target highest temperature values ​​for measurement points c1, c2, and c3, and the target weight value w2 are obtained. For the lowest temperature dimension, the target lowest temperature values ​​for measurement points c1, c2, and c3, and the target weight value w3 are obtained. For the temperature difference dimension, the target temperature difference values ​​for measurement points c1, c2, and c3, and the target weight value w4 are obtained. For the phase-to-phase temperature dimension, the target phase-to-phase temperature value and the target weight value w5 are obtained. For the temperature rise dimension, the target temperature rise values ​​for measurement points c1, c2, and c3, and the target weight value w6 are obtained. For the relative temperature difference dimension, the target relative temperature difference values ​​of temperature measurement point c1, c2, and c3, and the target weight value w7 are obtained.

[0108] Fourth, the model training process. During the model training process, an initial machine learning model can be obtained and trained to obtain the trained target machine learning model.

[0109] For example, an initial machine learning model (such as a classification model) can be pre-configured. There are no restrictions on the structure and function of this initial machine learning model, as long as the initial machine learning model can perform classification, that is, it can output classification results such as normal level, general level, severe level, and urgent level.

[0110] For example, the initial machine learning model could be a machine learning model based on the GBDT algorithm, or a machine learning model based on the KNN algorithm, or a machine learning model based on the SVM algorithm. For instance, if the performance of the machine learning model based on the GBDT algorithm is better than that of the machine learning model based on the KNN algorithm, and the performance of the machine learning model based on the GBDT algorithm is better than that of the machine learning model based on the SVM algorithm, then the machine learning model based on the GBDT algorithm can be selected as the initial machine learning model.

[0111] For example, to train an initial machine learning model, the following method can be used:

[0112] For each target dimension, feature values ​​can be determined based on the target temperature value and target weight value for that dimension. Then, the feature values ​​from multiple target dimensions are combined into a matrix, which serves as the training feature vector. (See [link to documentation]). Figure 4 The image shows an example of training feature vectors.

[0113] exist Figure 4 In this diagram, s11 represents the target ambient temperature value, w1 represents the target weight value for the ambient temperature dimension, and s11*w1 represents the feature value for the ambient temperature dimension. s21 represents the target maximum temperature value for measurement point c1, s22 represents the target maximum temperature value for measurement point c2, s23 represents the target maximum temperature value for measurement point c3, w2 represents the target weight value for the highest temperature dimension, s21*w2 represents the feature value for measurement point c1 in the highest temperature dimension, s22*w2 represents the feature value for measurement point c2 in the highest temperature dimension, and s23*w2 represents the feature value for measurement point c3 in the highest temperature dimension. s31*w3 represents the feature value for measurement point c1 in the lowest temperature dimension, s32*w3 represents the feature value for measurement point c2 in the lowest temperature dimension, and s33*w3 represents the feature value for measurement point c3 in the lowest temperature dimension. s41*w4 represents the characteristic value of temperature measurement point c1 in the temperature difference dimension, s42*w4 represents the characteristic value of temperature measurement point c2 in the temperature difference dimension, and s43*w4 represents the characteristic value of temperature measurement point c3 in the temperature difference dimension. s51*w5 represents the characteristic value in the phase-to-phase temperature dimension. s61*w6 represents the characteristic value of temperature measurement point c1 in the temperature rise dimension, s62*w6 represents the characteristic value of temperature measurement point c2 in the temperature rise dimension, and s63*w6 represents the characteristic value of temperature measurement point c3 in the temperature rise dimension. s71*w7 represents the characteristic value of temperature measurement point c1 in the relative temperature difference dimension, s72*w7 represents the characteristic value of temperature measurement point c2 in the relative temperature difference dimension, and s73*w7 represents the characteristic value of temperature measurement point c3 in the relative temperature difference dimension.

[0114] In summary, a 3*7 dimensional training feature vector can be obtained, and the label value corresponding to this training feature vector is the training fault level. For example, when the training fault level is normal, the label value of the training feature vector is 0; when the training fault level is moderate, the label value is 1; when the training fault level is severe, the label value is 2; and when the training fault level is critical, the label value is 3. Based on this, the training feature vector and the label value can be input into the initial machine learning model to train it. In practical applications, a large number of training feature vectors and their label values ​​can be input into the initial machine learning model together for training. The training process of this initial machine learning model is not restricted. The machine learning model that has been trained can be called the target machine learning model. Thus, the trained target machine learning model can be obtained.

[0115] In one possible implementation, if the initial machine learning model is based on the GBDT algorithm, then the target machine learning model is also based on the GBDT algorithm. Alternatively, if the initial machine learning model is based on the KNN algorithm, then the target machine learning model is also based on the KNN algorithm. Or, if the initial machine learning model is based on the SVM algorithm, then the target machine learning model is also based on the SVM algorithm.

[0116] V. Model Deployment Process (i.e., Online Deployment Process). During the model deployment process, the trained target machine learning model can be saved and deployed to the management device. This allows the management device to detect the corresponding fault level of the power equipment based on the target machine learning model, enabling real-time monitoring of the power equipment's operating status and determining whether a fault has occurred, which can then be used for subsequent diagnostic applications.

[0117] For example, temperature measurement data from various power scenarios can be continuously collected, and the target machine learning model can be automatically iterated. That is, the target machine learning model is used as the initial machine learning model, and the initial machine learning model is trained based on the newly collected temperature measurement data to obtain a new target machine learning model. This process is repeated to continuously optimize the analysis effect and accuracy of the target machine learning model.

[0118] For example, after deploying the target machine learning model to the management device, the management device can detect the corresponding fault level of the power equipment based on the target machine learning model, which may include the following steps:

[0119] Step P11, Data Acquisition Step. For example, in each acquisition cycle, the initial temperature value corresponding to the power equipment can be obtained. This initial temperature value may include the initial ambient temperature value corresponding to the power equipment, multiple measured temperature values ​​corresponding to the power equipment, and the average temperature value corresponding to the multiple measured temperature values.

[0120] All physical locations of the electrical equipment can be divided into multiple temperature measurement points, such as temperature measurement point c1, temperature measurement point c2, and temperature measurement point c3. Based on this, the initial ambient temperature values ​​include the initial ambient temperature values ​​at temperature measurement points c1, c2, and c3. The measured temperature values ​​include the measured temperature values ​​at the physical locations corresponding to temperature measurement points c1, c2, and c3. The average temperature values ​​include the average temperature values ​​corresponding to temperature measurement points c1, c2, and c3.

[0121] Step P11 is similar to the data acquisition process described above, and will not be repeated here.

[0122] Step P12, Feature Engineering Steps. For example, in each acquisition cycle, for each target dimension associated with the power equipment, the target temperature value of that target dimension can be determined based on the initial temperature value. Multiple target dimensions may include, but are not limited to, at least one of the following: ambient temperature dimension; maximum temperature dimension; minimum temperature dimension; temperature difference dimension; phase-to-phase temperature dimension; temperature rise dimension; relative temperature difference dimension. The target temperature value may include, but is not limited to, at least one of the following: target ambient temperature value, target maximum temperature value, target minimum temperature value, target temperature difference value, target phase-to-phase temperature value, target temperature rise value, and target relative temperature difference value.

[0123] For example, regarding the ambient temperature dimension, the target ambient temperature value can be determined based on the initial ambient temperature value. Regarding the maximum temperature dimension, the target maximum temperature value can be determined based on the maximum measured temperature value among multiple measured temperature values. Regarding the minimum temperature dimension, the target minimum temperature value can be determined based on the minimum measured temperature value among multiple measured temperature values. Regarding the temperature difference dimension, the target temperature difference value can be determined based on the target maximum temperature value and the target minimum temperature value. Regarding the interphase temperature dimension, the target interphase temperature value can be determined based on the average temperature value corresponding to multiple measured temperature values. Regarding the temperature rise dimension, the target temperature rise value for the current acquisition cycle can be determined based on the average temperature value of the current acquisition cycle and the average temperature value of the previous acquisition cycle. Regarding the relative temperature difference dimension, the target relative temperature difference value for the current acquisition cycle can be determined based on the target maximum temperature value of the current acquisition cycle and the target maximum temperature value of the target acquisition cycle.

[0124] For example, the target weight value corresponding to each target dimension can be determined, such as the target weight value w1 for the ambient temperature dimension, the target weight value w2 for the highest temperature dimension, the target weight value w3 for the lowest temperature dimension, the target weight value w4 for the temperature difference dimension, the target weight value w5 for the phase temperature dimension, the target weight value w6 for the temperature rise dimension, and the target weight value w7 for the relative temperature difference dimension.

[0125] Step P12 is similar to the feature engineering process described above, and will not be repeated here.

[0126] Step P13, Fault Level Detection Steps. For example, for each target dimension, determine the feature value of that target dimension based on the target temperature value and the target weight value of that target dimension. Then, combine the feature values ​​of multiple target dimensions into a matrix, and use this matrix as the feature vector to be detected. This feature vector to be detected is then compared with... Figure 4 The training feature vectors shown are similar, all being 3*7 dimensional feature vectors. Then, this feature vector to be detected is input into the target machine learning model, which processes it without restrictions to obtain the fault level corresponding to the power equipment and outputs the corresponding fault level. For example, if the label value of the output feature vector is 0, the fault level is determined to be normal; if the label value is 1, the fault level is determined to be moderate; if the label value is 2, the fault level is determined to be severe; and if the label value is 3, the fault level is determined to be emergency. In summary, the fault level corresponding to the power equipment can be obtained, meaning that in each data acquisition cycle, the fault level of the power equipment in that cycle can be obtained.

[0127] As can be seen from the above technical solutions, in this embodiment, machine learning technology can be used to detect the fault level of power equipment without relying on the accuracy of temperature thresholds, thereby improving the accuracy of fault detection results (i.e., fault level) and enabling timely detection of whether power equipment has malfunctioned. When power equipment malfunctions, it can be repaired promptly, preventing the malfunctioning equipment from continuing to operate and avoiding potential safety hazards. When relying on temperature thresholds to detect the fault level of power equipment, the accuracy rate is 80%, while the accuracy rate using the technical solution of this embodiment is 92%. Clearly, the detection accuracy of the technical solution of this embodiment is significantly better than the detection method relying on temperature thresholds. In this embodiment, through feature engineering, the training weights of different target dimensions are controlled, and feature learning from a large number of samples is performed, enabling the target machine learning model to have strong generalization ability, thereby improving the accuracy of the detection results. After deploying the target machine learning model, temperature measurement data from various power scenarios can be continuously collected, and automatic iteration can be performed on the existing model to optimize the model analysis effect and accuracy, greatly reducing analysis and optimization costs and improving analysis accuracy.

[0128] Based on the same concept as the above method, this application proposes a fault detection device for power equipment. The power equipment is associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the power equipment. (See [link]). Figure 5 The diagram shown is a structural diagram of the device, which includes:

[0129] The acquisition module 51 is used to acquire the initial temperature value corresponding to the power equipment;

[0130] The determination module 52 is used to determine the target temperature value of the target dimension based on the initial temperature value for each target dimension associated with the power equipment, and to determine the feature value of the target dimension based on the target temperature value of the target dimension and the target weight value of the target dimension.

[0131] Generation module 53 is used to construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions;

[0132] The processing module 54 is used to input the feature vector to be detected into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment.

[0133] In one possible implementation, the multiple target dimensions associated with the power equipment include at least one of the following: ambient temperature dimension; maximum temperature dimension; minimum temperature dimension; temperature difference dimension; phase-to-phase temperature dimension; temperature rise dimension; and relative temperature difference dimension. Of course, the above are just a few examples of target dimensions.

[0134] For example, the initial temperature value includes the initial ambient temperature value corresponding to the power equipment, multiple measured temperature values ​​corresponding to the power equipment, and the average temperature value corresponding to the multiple measured temperature values; when the determining module 52 determines the target temperature value of the target dimension based on the initial temperature value, it is specifically used to: if the target dimension is the ambient temperature dimension, then determine the target ambient temperature value of the target dimension based on the initial ambient temperature value; if the target dimension is the highest temperature dimension, then determine the target highest temperature value of the target dimension based on the highest measured temperature value among the multiple measured temperature values; if the target dimension is the lowest temperature dimension, then determine the target lowest temperature value of the target dimension based on the lowest measured temperature value among the multiple measured temperature values; if the target dimension is the temperature difference dimension, then determine the target temperature difference value of the target dimension based on the target highest temperature value and the target lowest temperature value; if the target dimension is the phase-to-phase temperature dimension, then determine the target phase-to-phase temperature value of the target dimension based on the average temperature value corresponding to the multiple measured temperature values.

[0135] For example, when the acquisition module 51 acquires the initial temperature value corresponding to the power equipment, it is specifically used to: acquire the initial temperature value corresponding to the power equipment in each acquisition cycle, wherein the initial temperature value includes multiple measured temperature values ​​in the acquisition cycle and the average temperature value corresponding to the multiple measured temperature values;

[0136] When determining the target temperature value of the target dimension based on the initial temperature value, the determining module 52 is specifically used for: if the target dimension is the temperature rise dimension, then determining the target temperature rise value of the target dimension in the current acquisition cycle based on the average temperature value of the current acquisition cycle and the average temperature value of the previous acquisition cycle; if the target dimension is the relative temperature difference dimension, then determining the target relative temperature difference value of the target dimension in the current acquisition cycle based on the target maximum temperature value of the current acquisition cycle and the target maximum temperature value of the target acquisition cycle; wherein, the target maximum temperature value of the current acquisition cycle is determined based on multiple measured temperature values ​​of the current acquisition cycle, the target maximum temperature value of the target acquisition cycle is determined based on multiple measured temperature values ​​of the target acquisition cycle, and the fault level corresponding to the target acquisition cycle is the normal level.

[0137] In one possible implementation, the acquisition module 51 is further configured to: acquire the configured initial weight value of the target dimension; if the target dimension is a first type of target dimension, determine the initial weight value as the target weight value of the target dimension; or, if the target dimension is a second type of target dimension, decrease the initial weight value and determine the decreased initial weight value as the target weight value of the target dimension; or, if the target dimension is a third type of target dimension, increase the initial weight value and determine the increased initial weight value as the target weight value of the target dimension.

[0138] For example, the first type of target dimension includes at least one of the following: maximum temperature dimension, minimum temperature dimension, temperature difference dimension, and phase temperature dimension; the second type of target dimension includes the ambient temperature dimension; and the third type of target dimension includes the temperature rise dimension and / or the relative temperature difference dimension.

[0139] In one possible implementation, the fault detection device for the power equipment further includes: a training module (in... Figure 5 (Not shown in the image) is used to obtain the training temperature value and training fault level corresponding to the power equipment; for each target dimension, the target temperature value of the target dimension is determined based on the training temperature value, and the feature value of the target dimension is determined based on the target temperature value and the target weight value of the target dimension; a training feature vector is constructed based on the feature values ​​of the multiple target dimensions; the training feature vector and the training fault level are input to the configured initial machine learning model to train the initial machine learning model and obtain the trained target machine learning model.

[0140] Based on the same concept as the above method, this application proposes an electronic device in which the power equipment is associated with multiple target dimensions. The temperature value of each target dimension is used to reflect the fault level of the power equipment. See [link to relevant documentation]. Figure 6 As shown, the electronic device may include: a processor 61 and a machine-readable storage medium 62, the machine-readable storage medium 62 storing machine-executable instructions executable by the processor 61; the processor 61 is used to execute the machine-executable instructions to perform the following steps:

[0141] Obtain the initial temperature value corresponding to the power equipment;

[0142] For each target dimension, a target temperature value is determined based on the initial temperature value, and a feature value of the target dimension is determined based on the target temperature value and the target weight value of the target dimension.

[0143] Construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions;

[0144] The feature vector to be detected is input into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment.

[0145] Based on the same concept as the above method, this application also provides a machine-readable storage medium storing a plurality of computer instructions, which, when executed by a processor, can implement the fault detection method for power equipment disclosed in the above examples of this application.

[0146] The aforementioned machine-readable storage medium can be any electronic, magnetic, optical, or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, machine-readable storage media can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0147] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0148] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0149] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0151] Furthermore, these computer program instructions can also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in the process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0152] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0153] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A fault detection method for power equipment, characterized in that, The method involves associating electrical equipment with multiple target dimensions, where the temperature value of each target dimension reflects the fault level of the electrical equipment. Obtain the initial temperature value corresponding to the power equipment; For each target dimension, a target temperature value is determined based on the initial temperature value, and a feature value of the target dimension is determined based on the target temperature value and the target weight value of the target dimension. Construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions; The feature vector to be detected is input into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment. The target temperature values ​​associated with the power equipment across multiple target dimensions include: The target ambient temperature value in the ambient temperature dimension, wherein the initial ambient temperature value in the current acquisition period is taken as the target ambient temperature value; The target maximum temperature value for the highest temperature dimension, wherein, within the current acquisition period, the maximum value is selected from all the measured temperature values ​​corresponding to the temperature measurement point as the target maximum temperature value for that temperature measurement point; The target minimum temperature value for the lowest temperature dimension, wherein, within the current acquisition period, the minimum value is selected from all the measured temperature values ​​corresponding to the temperature measurement point as the target minimum temperature value for that temperature measurement point; The target temperature difference value in the temperature difference dimension is the difference between the target highest temperature value and the target lowest temperature value of the temperature measurement point within the current acquisition period, which is taken as the target temperature difference value of the temperature measurement point. The target interphase temperature value in the interphase temperature dimension is obtained by acquiring the average temperature value of multiple temperature measurement points within the current acquisition period, and calculating the absolute value of the difference between the average temperature values ​​of any two temperature measurement points. The maximum value among the absolute values ​​is taken as the target interphase temperature value. The target temperature rise value in the temperature rise dimension is determined by the absolute value of the difference between the average temperature value of the current acquisition cycle and the average temperature value of the previous acquisition cycle. When there is no previous acquisition cycle, the default temperature value is used as the average temperature value of the previous acquisition cycle. The target relative temperature difference value in the relative temperature difference dimension is determined based on the target maximum temperature value of the current acquisition cycle, the target maximum temperature value of the target acquisition cycle, and the target ambient temperature value. The target maximum temperature value of the current acquisition cycle is the maximum value among multiple measured temperature values ​​in the current acquisition cycle; the target maximum temperature value of the target acquisition cycle is the maximum value among multiple measured temperature values ​​in the target acquisition cycle, and the target acquisition cycle is the first acquisition cycle preceding the current acquisition cycle where the fault level is normal. Before determining the feature value of the target dimension based on the target temperature value and the target weight value of the target dimension, the method further includes: obtaining the configured initial weight value of the target dimension; if the target dimension is a first type of target dimension, then determining the initial weight value as the target weight value of the target dimension; if the target dimension is a second type of target dimension, then decreasing the initial weight value and determining the decreased initial weight value as the target weight value of the target dimension; if the target dimension is a third type of target dimension, then increasing the initial weight value and determining the increased initial weight value as the target weight value of the target dimension. The first type of target dimension includes the highest temperature dimension, the lowest temperature dimension, the temperature difference dimension, and the phase temperature dimension; the second type of target dimension includes the ambient temperature dimension; and the third type of target dimension includes the temperature rise dimension and the relative temperature difference dimension.

2. The method according to claim 1, characterized in that, The target machine learning models include: machine learning models based on the GBDT algorithm, machine learning models based on the KNN algorithm, and machine learning models based on the SVM algorithm.

3. The method according to claim 1, characterized in that, The training process of the target machine learning model includes: Obtain the training temperature value and training fault level corresponding to the power equipment; For each target dimension, a target temperature value for that target dimension is determined based on the training temperature value, and a feature value for that target dimension is determined based on the target temperature value and the target weight value for that target dimension. Construct a training feature vector based on the feature values ​​of the multiple target dimensions; The training feature vector and the training fault level are input into the configured initial machine learning model to train the initial machine learning model and obtain the trained target machine learning model.

4. A fault detection device for power equipment, characterized in that, The electrical equipment is associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the electrical equipment. The device includes: The acquisition module is used to acquire the initial temperature value corresponding to the power equipment; The determination module is configured to, for each target dimension associated with the power equipment, determine a target temperature value for the target dimension based on the initial temperature value, and determine a feature value for the target dimension based on the target temperature value and the target weight value of the target dimension. The generation module is used to construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions; The processing module is used to input the feature vector to be detected into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment; The target temperature values ​​associated with the power equipment across multiple target dimensions include: The target ambient temperature value in the ambient temperature dimension, wherein the initial ambient temperature value in the current acquisition period is taken as the target ambient temperature value; The target maximum temperature value for the highest temperature dimension, wherein, within the current acquisition period, the maximum value is selected from all the measured temperature values ​​corresponding to the temperature measurement point as the target maximum temperature value for that temperature measurement point; The target minimum temperature value for the lowest temperature dimension, wherein, within the current acquisition period, the minimum value is selected from all the measured temperature values ​​corresponding to the temperature measurement point as the target minimum temperature value for that temperature measurement point; The target temperature difference value in the temperature difference dimension is the difference between the target highest temperature value and the target lowest temperature value of the temperature measurement point within the current acquisition period, which is taken as the target temperature difference value of the temperature measurement point. The target interphase temperature value in the interphase temperature dimension is obtained by acquiring the average temperature value of multiple temperature measurement points within the current acquisition period, and calculating the absolute value of the difference between the average temperature values ​​of any two temperature measurement points. The maximum value among the absolute values ​​is taken as the target interphase temperature value. The target temperature rise value in the temperature rise dimension is determined by the absolute value of the difference between the average temperature value of the current acquisition cycle and the average temperature value of the previous acquisition cycle. When there is no previous acquisition cycle, the default temperature value is used as the average temperature value of the previous acquisition cycle. The target relative temperature difference value in the relative temperature difference dimension is determined based on the target maximum temperature value of the current acquisition cycle, the target maximum temperature value of the target acquisition cycle, and the target ambient temperature value. The target maximum temperature value of the current acquisition cycle is the maximum value among multiple measured temperature values ​​in the current acquisition cycle; the target maximum temperature value of the target acquisition cycle is the maximum value among multiple measured temperature values ​​in the target acquisition cycle, and the target acquisition cycle is the first acquisition cycle preceding the current acquisition cycle where the fault level is normal. The acquisition module is further configured to: acquire the configured initial weight value of the target dimension; if the target dimension is a first type of target dimension, determine the initial weight value as the target weight value of the target dimension; if the target dimension is a second type of target dimension, decrease the initial weight value and determine the decreased initial weight value as the target weight value of the target dimension; if the target dimension is a third type of target dimension, increase the initial weight value and determine the increased initial weight value as the target weight value of the target dimension. The first type of target dimension includes the highest temperature dimension, the lowest temperature dimension, the temperature difference dimension, and the phase temperature dimension; the second type of target dimension includes the ambient temperature dimension; and the third type of target dimension includes the temperature rise dimension and the relative temperature difference dimension.

5. An electronic device, characterized in that, The power equipment is associated with multiple target dimensions, and the temperature value of each target dimension is used to reflect the fault level of the power equipment. The electronic device includes a processor and a machine-readable storage medium, the machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to achieve the following steps: Obtain the initial temperature value corresponding to the power equipment; For each target dimension, a target temperature value is determined based on the initial temperature value, and a feature value for the target dimension is determined based on the target temperature value and the target weight value of the target dimension. Construct a feature vector to be detected based on the feature values ​​of the multiple target dimensions; The feature vector to be detected is input into the trained target machine learning model so that the target machine learning model outputs the fault level corresponding to the power equipment. The target temperature values ​​associated with the power equipment across multiple target dimensions include: The target ambient temperature value in the ambient temperature dimension, wherein the initial ambient temperature value in the current acquisition period is taken as the target ambient temperature value; The target maximum temperature value for the highest temperature dimension, wherein, within the current acquisition period, the maximum value is selected from all the measured temperature values ​​corresponding to the temperature measurement point as the target maximum temperature value for that temperature measurement point; The target minimum temperature value for the lowest temperature dimension, wherein, within the current acquisition period, the minimum value is selected from all the measured temperature values ​​corresponding to the temperature measurement point as the target minimum temperature value for that temperature measurement point; The target temperature difference value in the temperature difference dimension is the difference between the target highest temperature value and the target lowest temperature value of the temperature measurement point within the current acquisition period, which is taken as the target temperature difference value of the temperature measurement point. The target interphase temperature value in the interphase temperature dimension is obtained by acquiring the average temperature value of multiple temperature measurement points within the current acquisition period, and calculating the absolute value of the difference between the average temperature values ​​of any two temperature measurement points. The maximum value among the absolute values ​​is taken as the target interphase temperature value. The target temperature rise value in the temperature rise dimension is determined by the absolute value of the difference between the average temperature value of the current acquisition cycle and the average temperature value of the previous acquisition cycle. When there is no previous acquisition cycle, the default temperature value is used as the average temperature value of the previous acquisition cycle. The target relative temperature difference value in the relative temperature difference dimension is determined based on the target maximum temperature value of the current acquisition cycle, the target maximum temperature value of the target acquisition cycle, and the target ambient temperature value. The target maximum temperature value of the current acquisition cycle is the maximum value among multiple measured temperature values ​​in the current acquisition cycle; the target maximum temperature value of the target acquisition cycle is the maximum value among multiple measured temperature values ​​in the target acquisition cycle, and the target acquisition cycle is the first acquisition cycle preceding the current acquisition cycle where the fault level is normal. Before determining the feature value of the target dimension based on the target temperature value and the target weight value of the target dimension, the method further includes: obtaining the configured initial weight value of the target dimension; if the target dimension is a first type of target dimension, then determining the initial weight value as the target weight value of the target dimension; if the target dimension is a second type of target dimension, then decreasing the initial weight value and determining the decreased initial weight value as the target weight value of the target dimension; if the target dimension is a third type of target dimension, then increasing the initial weight value and determining the increased initial weight value as the target weight value of the target dimension. The first type of target dimension includes the highest temperature dimension, the lowest temperature dimension, the temperature difference dimension, and the phase temperature dimension; the second type of target dimension includes the ambient temperature dimension; and the third type of target dimension includes the temperature rise dimension and the relative temperature difference dimension.

Images