A Fault Detection Method for Pole-Mounted Circuit Breakers Based on Naive Bayes

Abstract

The present invention discloses a pole-mounted circuit breaker fault detection method based on Naive Bayes. The method first collects characteristic data such as phase current, ferromagnetic amplitude and phase voltage according to a preset frequency, generates a characteristic data set, and determines its combination type. Then, the characteristic data is analyzed using a Naive Bayes algorithm and an evaluation result is output. In order to enhance the accuracy of detection, the method further obtains abnormal data, calculates the credibility of the abnormal data, and determines the degree of fuzziness of each feature to the target feature in combination with the evaluation result. Based on this, a fault factor is generated and combined with the characteristic data at the current operating moment to determine the fault state of the circuit breaker. During the processing, the method uses Laplace smoothing to process the missing characteristic values, and selects the most likely fault category by calculating the posterior probability.

Description

Technical Field

[0001] The present invention relates to the technical field of circuit breaker fault detection, and in particular to a pole-mounted circuit breaker fault detection method based on Naive Bayes. Background Art

[0002] In the power system, the pole-mounted circuit breaker is a key component of the distribution network, and its reliable operation is crucial to ensure the stability and safety of the power grid. However, due to various reasons such as equipment aging, environmental factors, overload operation, etc., pole-mounted circuit breakers often fail, such as single-phase grounding faults, phase-to-phase short circuits, etc. These faults not only affect the power supply quality, but may also cause equipment damage and even cause power outages on a larger scale. Therefore, timely and accurate detection of the fault status of the pole-mounted circuit breaker is of great significance to improving the operating efficiency of the power system and reducing fault losses.

[0003] Traditional pole-mounted circuit breaker fault detection methods mostly rely on manual inspections, regular maintenance, and empirical judgment. These methods have disadvantages such as long detection cycles, low efficiency, and difficulty in real-time fault detection. With the development of smart grid and Internet of Things technologies, data-driven fault detection methods have gradually become a research hotspot. Among them, the Naive Bayes algorithm has been widely used in the field of power equipment fault diagnosis because of its simplicity, efficiency, ease of implementation, and ability to process multi-feature data.

[0004] However, existing pole-mounted circuit breaker fault detection methods often only focus on a single feature or a simple feature combination, ignoring the complexity and diversity of power system operation data, resulting in limited accuracy and robustness of fault detection. In addition, there is a lack of effective methods for processing feature data and identifying abnormal data, which further affects the accuracy of fault detection. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] To solve the above problems, the present invention proposes a pole-mounted circuit breaker fault detection method based on Naive Bayes, aiming to solve the problem that the pole-mounted circuit breaker fault detection method in the prior art often only focuses on a single feature or a simple feature combination, ignoring the complexity and diversity of the power system operation data, resulting in limited accuracy and robustness of fault detection.

[0007] (II) Technical solution

[0008] A method for detecting a pole-mounted circuit breaker fault based on Naive Bayes of the present invention comprises:

[0009] Acquire first feature data according to a preset collection frequency, generate a first feature data set based on the first feature data, and determine a combination type of the first feature data set;

[0010] Analyze all first feature data in the first feature data set based on the Naive Bayes algorithm, and output an evaluation result of each first feature data;

[0011] Acquire abnormal data corresponding to the first feature data in the first feature data set, generate abnormal data credibility based on the abnormal data, and obtain the fuzziness of each first feature data for each target feature based on the evaluation result and the abnormal data credibility;

[0012] A fault factor is generated based on the fuzzy degree, characteristic data corresponding to the current operation moment of the power system is obtained, and the fault state of the pole-mounted circuit breaker is determined according to the characteristic data corresponding to the current operation moment and the fault factor.

[0013] In the present invention, characteristic data corresponding to each operating moment of the power system during the current collection time and the historical operating time period are obtained, recorded as first characteristic data, the first characteristic data set is preprocessed and a first characteristic data set is generated;

[0014] The first characteristic data includes one or more combinations of phase current data, ferromagnetic amplitude data and phase voltage data. The first characteristic data set determines the combination type of the first characteristic data based on the fault factor.

[0015] In the present invention, the prior probability of phase current data is recorded as , the prior probability of phase voltage data is , the prior probability of the ferromagnetic amplitude is , the probability of occurrence of the fault category in the phase current data is recorded as , the probability of occurrence of the fault category in the phase voltage data is , the probability of occurrence of the fault category in the ferromagnetic amplitude data is ,in is the number of intervals for each category. Assuming that the features are independent of each other, It can be decomposed into the multiplication of the probabilities of individual features: Calculate the posterior probability of the fault category in each data separately ,in is the probability of the fault category, which serves as a normalization factor constant.

[0016] In the present invention, Laplace smoothing is used to process the situation where the feature value does not appear, and the naive Bayes formula is used to calculate the posterior probability of each category under the given feature, and the category with the highest posterior probability is selected as the evaluation result.

[0017] In the present invention, other first feature data in the first feature data set except the adopted combination type are recorded as auxiliary feature data, and the absolute value of the difference between the auxiliary feature data and each first feature data is calculated and recorded as the feature difference between the first feature data and each first feature data.

[0018] In the present invention, the first feature data is then sorted in order of feature difference from small to large to obtain a first feature data sequence.

[0019] In the present invention, starting from the first auxiliary feature data in the first feature data sequence, N first feature data are randomly selected, and a set constructed by the selected N auxiliary feature data is recorded as a set of neighboring auxiliary feature data corresponding to the auxiliary feature data;

[0020] The number of auxiliary feature data with normal abnormal data and the number of auxiliary feature data with single-phase grounding fault abnormal data in the auxiliary feature data set of the neighboring first feature data corresponding to the auxiliary feature data are counted respectively, and recorded as the first number and the second number respectively.

[0021] In the present invention, the ratio of the first number to the total number of auxiliary feature data in the neighboring auxiliary feature data set corresponding to the auxiliary feature data is recorded as the credibility of the normal abnormal data corresponding to the auxiliary feature data.

[0022] In the present invention, the ratio of the second number to the total number of auxiliary feature data in the neighboring auxiliary feature data set corresponding to the auxiliary feature data is recorded as the credibility of the single-phase grounding fault abnormal data corresponding to the auxiliary feature data.

[0023] (III) Beneficial effects

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention acquires a variety of first characteristic data (including phase current data, ferromagnetic amplitude data and phase voltage data, etc.) through a preset collection frequency, and analyzes these characteristic data based on the naive Bayes algorithm, fully considering the complexity and diversity of the power system operation data, thereby improving the accuracy of fault detection.

[0026] The present invention introduces the concept of abnormal data credibility, calculates the feature difference between the auxiliary feature data and each first feature data, constructs a set of neighboring auxiliary feature data, and counts the number of abnormal data to evaluate the credibility of the abnormal data. This method effectively handles the noise and abnormal values ​​in the feature data and enhances the robustness of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 Schematic diagram of the logical structure of the fault detection method. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Example 1

[0031] like Figure 1 The purpose of this embodiment is to reduce or avoid the malfunction of the pole-mounted circuit breaker. Since the malfunction of the pole-mounted circuit breaker is related to the misjudgment of the single-phase grounding fault in the operation of the power system, this embodiment provides a state division optimization method for the pole-mounted circuit breaker to reduce or avoid the problem of malfunction of the pole-mounted circuit breaker. The method determines the fault factor based on the first characteristic data and the abnormal data of the first characteristic data. Based on the determined fault factor, the single-phase grounding fault in the operation of the power system can be judged more accurately, and then the state of the pole-mounted circuit breaker can be determined more accurately.

[0032] Step S100: acquiring first feature data according to a preset collection frequency, generating a first feature data set based on the first feature data, and determining a combination type of the first feature data set.

[0033] This embodiment first obtains characteristic data corresponding to each operating moment of the power system during the current collection time and the historical operation time period, and records it as first characteristic data, and records a set constructed from all the collected first characteristic data as a first characteristic data set.

[0034] Specifically, the collected first feature data needs to be preprocessed before forming the first feature data set, which specifically includes missing value processing, data filtering and data standardization processing of the first feature data set.

[0035] The types of the first characteristic data include phase current data, ferromagnetic amplitude data and phase voltage data. This embodiment analyzes the characteristic data of one or more types as an example. For example, the first characteristic data in this embodiment refers to the phase current data or the combination of ferromagnetic amplitude data and phase current data or the combination of phase current data, ferromagnetic amplitude data and phase voltage data for judgment. That is, this embodiment judges the single-phase grounding fault in the operation of the power system based on one or three combinations.

[0036] As another embodiment, the first characteristic data may also be set to refer to phase voltage data, and the method for determining the single-phase grounding fault in the operation of the power system based on the phase voltage data and the ferromagnetic amplitude data is the same.

[0037] Specifically, in order to better allocate the combination types of the three types of data, during operation, the three types of first characteristic data are first simulated separately, and then another type of data is added according to the feedback results for simulation until the fault point can be found. That is, the phase current data, ferromagnetic amplitude data and phase voltage data are respectively brought into the following steps, and the fault factors are independently run to output, and the specific fault point is determined according to the fault factors.

[0038] In this embodiment, the implementer needs to set the historical operation time period according to the actual situation. For example, in this embodiment, the historical operation time period can be set to refer to the time period from the start of the power system operation to the current operation time.

[0039] In specific applications, the implementer also needs to set the data collection frequency according to actual conditions. For example, the data collection frequency is set to 60 Hz in this embodiment, and since the method of collecting phase current data and phase voltage data during the operation of the power system is a known technology, that is, the method of obtaining the first characteristic data during the operation of the power system is a known technology, this embodiment will not be described in detail.

[0040] Therefore, the first feature data set and all the first feature data in the first feature data set are obtained through the above process.

[0041] Step S200 , analyzing all first feature data in the first feature data set based on a Naive Bayes algorithm, and outputting evaluation results of each first feature data.

[0042] Next, this embodiment will classify all the first feature data in the acquired first feature data set. First, the data collected from the acquired first feature data will be cleaned, that is, data points with missing values ​​will be filled or deleted, and then outliers will be identified and processed to ensure the reliability of the data. Finally, the processed data will be standardized so that all features have the same scale.

[0043] Convert continuous values ​​into discrete categories, that is, divide the phase current data, ferromagnetic amplitude data, and phase voltage data into different intervals, each interval as a category, and then label the data intervals, that is, according to the preset threshold, record the interval parameters that deviate from the threshold and record them as fault categories. Count the number of samples for each fault category and calculate the prior probability of each category. , for each category and each feature, calculate the probability of the feature appearing in that category , Laplace smoothing is used to handle the situation where the feature value does not appear to avoid the probability of 0. The naive Bayes formula is used to calculate the posterior probability of each category under the given feature, and the category with the highest posterior probability is selected as the evaluation result.

[0044] Specifically, the prior probability of phase current data is , the prior probability of phase voltage data is , the prior probability of the ferromagnetic amplitude is , the probability of occurrence of the fault category in the phase current data is recorded as , the probability of occurrence of the fault category in the phase voltage data is , the probability of occurrence of the fault category in the ferromagnetic amplitude data is ,in is the number of intervals for each category. Assuming that the features are independent of each other, It can be decomposed into the multiplication of the probabilities of individual features: Calculate the posterior probability of the fault category in each data separately ,in is the probability of the fault category, which is used as a normalization factor constant. A fault probability threshold is preset, and the category output greater than the fault probability threshold is selected as the evaluation result.

[0045] Step S300, obtain the abnormal data corresponding to each first feature data in the first feature data set, obtain the abnormal data credibility corresponding to each first feature data based on the abnormal data corresponding to each first feature data, and obtain the fuzziness of each first feature data for each target feature based on the evaluation result and the abnormal data credibility.

[0046] Normally, based on the fault category of each target feature for each first feature data obtained in step S200 and the posterior probability of all first feature data corresponding to each data fault category, the fault factor is further obtained by subsequently analyzing the posterior probability of all first feature data corresponding to each data fault category. The fault factor is the basis for the subsequent verification of the state of the pole-mounted circuit breaker; however, since the change from normal operation of the power system to the occurrence of a single-phase grounding fault is a gradual process, and it is not a sudden change from normal operation to a single-phase grounding fault, there is a transition process from normal operation of the power system to the occurrence of a single-phase grounding fault, that is, the transition process refers to the process from normal operation of the power system to the occurrence of a single-phase grounding fault.

[0047] The difference between the characteristic data in the transition process is generally small. Therefore, it is difficult to accurately divide the first characteristic data in the transition process based on the fault category of each target feature for each first characteristic data. For example, the purpose of calculating the posterior probability is to divide normal data and single-phase grounding fault data, that is, the purpose of calculating the posterior probability is to obtain the normal data fault category and the single-phase grounding fault data fault category. Then, the normal data fault category obtained directly through the fault category of each target feature for each first characteristic data may include single-phase grounding fault data, and the obtained single-phase grounding fault data fault category may include normal data, that is, the accuracy of the obtained single-phase grounding fault data fault category and the normal data fault category is low, then the credibility of the fault factor obtained based on the low-accuracy single-phase grounding fault data fault category and the normal data fault category will also be low.

[0048] The state of the pole-mounted circuit breaker cannot be determined accurately based on fault factors with low credibility, that is, the pole-mounted circuit breaker may still malfunction based on fault factors with low credibility, and normal data refers to the first characteristic data collected when the power system operates normally, and single-phase grounding fault data refers to the first characteristic data collected when a single-phase grounding fault occurs in the power system.

[0049] Therefore, this embodiment will adjust the fault category of each first feature data to each target feature based on the abnormal data marked when collecting the first feature data, so as to obtain the fuzziness degree of each first feature data point to each target feature, and subsequently based on the fuzziness degree of each first feature data point to each target feature, the normal data fault category and the single-phase grounding fault data fault category can be obtained more accurately, so that a more accurate fault factor can be obtained. Therefore, the specific process of obtaining the fuzziness degree of each first feature data point to each target feature is:

[0050] For any first feature data in the first feature data set: record the other first feature data in the first feature data set except the first feature data as auxiliary feature data, calculate the absolute value of the difference between the auxiliary feature data and each first feature data, and record it as the feature difference between the first feature data and each first feature data. Then sort the first feature data in the order of the feature difference from small to large to obtain the first feature data sequence.

[0051] Starting from the first auxiliary feature data in the first feature data sequence, N first feature data are randomly selected, and the set constructed by the selected N auxiliary feature data is recorded as the neighboring auxiliary feature data set corresponding to the auxiliary feature data. Then, the number of auxiliary feature data whose abnormal data is normal and the number of auxiliary feature data whose abnormal data is a single-phase grounding fault in the auxiliary feature data set of the neighboring first feature data corresponding to the auxiliary feature data are counted respectively, and are recorded as the first number and the second number respectively.

[0052] The ratio of the first number to the total number of auxiliary feature data in the neighboring auxiliary feature data set corresponding to the auxiliary feature data is recorded as the credibility of the normal abnormal data corresponding to the auxiliary feature data. The ratio of the second number to the total number of auxiliary feature data in the neighboring auxiliary feature data set corresponding to the auxiliary feature data is recorded as the credibility of the single-phase grounding fault abnormal data corresponding to the auxiliary feature data.

[0053] The single-phase ground fault abnormal data credibility and the normal abnormal data credibility are both the abnormal data credibility corresponding to the auxiliary feature data. And when the single-phase ground fault abnormal data credibility corresponding to the auxiliary feature data is higher, it means that the auxiliary feature data of the neighbors of the auxiliary feature data are more of the single-phase ground fault data type, which also indicates that the auxiliary feature data is more likely to be the single-phase ground fault data type. When the normal abnormal data credibility corresponding to the auxiliary feature data is higher, it means that the auxiliary feature data of the neighbors of the auxiliary feature data are more of the normal data type, which also indicates that the auxiliary feature data is more likely to be the normal data type.

[0054] In the specific application, the value of N needs to be set according to the actual situation. In this embodiment, the value of N is set to 20.

[0055] Therefore, through the above process, the credibility of the single-phase grounding fault abnormal data and the credibility of the normal and abnormal data corresponding to each first characteristic data can be obtained.

[0056] The product of the posterior probability of the first feature data and the corresponding credibility of the normal and abnormal data is recorded as the correlation of the first feature data to the normal target feature, and the product of the posterior probability of the first feature data and the corresponding credibility of the single-phase grounding fault abnormal data is recorded as the correlation of the first feature data to the single-phase grounding fault target feature; then the correlation of the first feature data to the normal target feature and the correlation of the first feature data to the single-phase grounding fault target feature are normalized, and the normalized correlations are recorded as the fuzziness of the first feature data to the normal target feature and the fuzziness of the first feature data to the single-phase grounding fault target feature, respectively; and the sum of the normalized correlation of the first feature data to the normal target feature and the correlation of the first feature data to the single-phase grounding fault target feature is a constant 1, and the larger the value of the fuzziness of the first feature data to the feature, the greater the possibility that the corresponding first feature data belongs to the category corresponding to the corresponding feature.

[0057] Since the credibility of the abnormal data corresponding to the first feature data can indicate the possibility that the corresponding first feature data is a single-phase grounding fault data type and the possibility that it is a normal data type, the posterior probability of each target feature of the corresponding first feature data is adjusted based on the credibility of the abnormal data corresponding to the first feature data, and the data fault category obtained based on the adjusted fuzzy degree is more accurate.

[0058] Therefore, through the above process, the fuzziness degree of each first characteristic data to the normal target feature and the fuzziness degree of each first characteristic data to the single-phase grounding fault target feature are obtained.

[0059] Step S400, according to the degree of fuzziness, obtain the fault category of each data, obtain the fault factor according to each data fault category, obtain the characteristic data corresponding to the current operating moment of the power system, and determine the state of the pole-mounted circuit breaker according to the characteristic data and fault factor corresponding to the current operating moment.

[0060] First, the fuzziness of each first characteristic data obtained in step S300 to the normal target characteristic and the fuzziness of the single-phase grounding fault target characteristic are reviewed. These fuzziness values ​​reflect the possibility of each characteristic data belonging to the respective category.

[0061] In order to classify the data into normal or single-phase grounding fault categories, a fuzziness threshold needs to be set. This threshold can be adjusted according to actual conditions and experience to ensure the accuracy and reliability of classification. For each first feature data, compare its fuzziness for the two categories of normal and single-phase grounding fault. If the fuzziness of a data for single-phase grounding fault is greater than the set threshold and greater than its fuzziness for normal, the data is classified as single-phase grounding fault data; otherwise, it is classified as normal data.

[0062] The number of classified single-phase grounding fault data and normal data, as well as their respective characteristic distributions, are counted. By analyzing the characteristics of single-phase grounding fault data, the key factors or characteristics that cause the fault, namely the fault factors, are identified. These fault factors may include specific current waveform anomalies, voltage fluctuations, ferromagnetic amplitude changes, etc. The identified fault factors are verified using historical data or simulation experiments to ensure their accuracy and reliability.

[0063] Through the monitoring equipment or sensors of the power system, the characteristic data such as phase current, phase voltage, ferromagnetic amplitude, etc. at the current operating moment are collected in real time. The collected data is cleaned, denoised and standardized to ensure the accuracy and consistency of the data.

[0064] Match the characteristic data of the current operating moment with the identified fault factors, and analyze whether the current data contains the characteristics of the fault factors. According to the result of the characteristic matching, evaluate the status of the pole-mounted circuit breaker. If the current data contains the characteristics of the fault factors, and the severity of these characteristics exceeds the preset threshold, it is judged that the circuit breaker may be in an abnormal or faulty state; otherwise, it is judged that the circuit breaker is in a normal state, and further increase the data combination type in the first data set for simulation operation. According to the results of the status evaluation, formulate corresponding maintenance or overhaul plans to ensure the stable operation of the circuit breaker and the safety of the power system.

[0065] In step S200, the naive Bayes algorithm has been used to analyze the first feature data set, and the evaluation results of each first feature data are output. These evaluation results provide a basis for subsequent data classification and fault factor identification.

[0066] In steps S300 and S400, fuzzy logic is used to process the fuzziness and uncertainty of the feature data, and the fuzziness degree is calculated to more accurately classify the data and identify the fault factors.

[0067] Throughout the process, threshold setting and decision making methods are used many times. For example, the fuzziness threshold is set in data classification, and the fault feature severity threshold is set in status assessment. The setting of these thresholds needs to be adjusted according to actual conditions and experience to ensure the accuracy and reliability of the entire method.

[0068] The above embodiments are only for describing the preferred implementation of the present invention, and are not intended to limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various modifications and improvements made by ordinary persons in the art to the technical solution of the present invention should fall within the protection scope of the present invention, and the technical contents for which protection is sought in the present invention have been fully recorded in the claims.

Claims

1. A pole-mounted circuit breaker fault detection method based on naive Bayes, characterized in that: The method comprises: Acquire first feature data according to a preset collection frequency, generate a first feature data set based on the first feature data, and determine a combination type of the first feature data set; Analyze all first feature data in the first feature data set based on the Naive Bayes algorithm, and output an evaluation result of each first feature data; Acquire abnormal data corresponding to the first feature data in the first feature data set, generate abnormal data credibility based on the abnormal data, and obtain the fuzziness of each first feature data for each target feature based on the evaluation result and the abnormal data credibility; A fault factor is generated based on the fuzzy degree, characteristic data corresponding to the current operation moment of the power system is obtained, and the fault state of the pole-mounted circuit breaker is determined according to the characteristic data corresponding to the current operation moment and the fault factor.

2. The pole-mounted circuit breaker fault detection method based on Naive Bayes according to claim 1 is characterized in that: Acquire characteristic data corresponding to each operating moment of the power system during the current collection time and the historical operating time period, record it as first characteristic data, pre-process the first characteristic data set and generate a first characteristic data set; The first characteristic data includes one or more combinations of phase current data, ferromagnetic amplitude data and phase voltage data. The first characteristic data set determines the combination type of the first characteristic data based on the fault factor.

3. The pole mounted circuit breaker fault detection method based on Naive Bayes according to claim 2 is characterized in that: Laplace smoothing is used to handle the situation where the feature value does not appear, and the naive Bayes formula is used to calculate the posterior probability of each category under the given feature, and the category with the highest posterior probability is selected as the evaluation result.

4. The pole mounted circuit breaker fault detection method based on Naive Bayes according to claim 3 is characterized in that: The other first feature data in the first feature data set except the adopted combination type are recorded as auxiliary feature data, and the absolute value of the difference between the auxiliary feature data and each first feature data is calculated and recorded as the feature difference between the first feature data and each first feature data.

5. The pole mounted circuit breaker fault detection method based on Naive Bayes according to claim 4 is characterized in that: Then, the first feature data are sorted in ascending order of feature difference values ​​to obtain a first feature data sequence.

6. The method for detecting faults of a pole mounted circuit breaker based on Naive Bayes according to claim 5, characterized in that: Starting from the first auxiliary feature data in the first feature data sequence, N first feature data are randomly selected, and a set constructed by the selected N auxiliary feature data is recorded as a set of neighboring auxiliary feature data corresponding to the auxiliary feature data; The number of auxiliary feature data with normal abnormal data and the number of auxiliary feature data with single-phase grounding fault abnormal data in the auxiliary feature data set of the neighboring first feature data corresponding to the auxiliary feature data are counted respectively, and recorded as the first number and the second number respectively.

7. The pole mounted circuit breaker fault detection method based on Naive Bayes according to claim 6, characterized in that: The ratio of the first number to the total number of auxiliary feature data in the neighboring auxiliary feature data set corresponding to the auxiliary feature data is recorded as the normal abnormal data credibility corresponding to the auxiliary feature data.

8. The method for detecting pole-mounted circuit breaker faults based on Naive Bayes according to claim 7, characterized in that: The ratio of the second number to the total number of auxiliary feature data in the neighboring auxiliary feature data set corresponding to the auxiliary feature data is recorded as the credibility of the single-phase grounding fault abnormal data corresponding to the auxiliary feature data.

Images