Electrical equipment risk assessment method and system based on Bayesian network

Abstract

The invention belongs to the field of electrical equipment risk assessment, and discloses an electrical equipment risk assessment method and system based on a Bayesian network, and the method comprises the steps: constructing a hierarchical Bayesian network model of electrical equipment, then multi-source heterogeneous data including equipment state monitoring data, operation condition data, environment data and historical operation and maintenance data are collected in real time; feature calculation is carried out on multi-source heterogeneous data based on a sliding time window, a dynamic feature vector is generated, and the vector comprises key information such as a data mean value, a standard deviation, a physical change rate and an abnormal event cumulative count in the time window. And inputting the dynamic feature vector into a hierarchical Bayesian network model, and calculating the probability that the equipment is in a high-risk state. And when the probability is greater than a dynamic probability threshold set based on the equipment criticality, risk early warning is triggered, so that the limitation of a traditional regular maintenance method is overcome, real-time risk assessment and accurate early warning of the electrical equipment are realized, the operation and maintenance cost is reduced, and the equipment reliability is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electrical equipment risk assessment, and relates to an electrical equipment risk assessment method and system based on a Bayesian network. Background Art

[0002] In power systems, electrical equipment is a core element in maintaining safe and efficient system operation. From power generation and transmission to distribution, electrical equipment in every link works closely together. The stable and reliable operation of electrical equipment is crucial to the normal operation of society and life. Aging is an inevitable natural process during the operation of electrical equipment. With increasing service life, key indicators such as insulation performance and mechanical strength of electrical equipment gradually decline, increasing the probability of failure.

[0003] Many power companies rely primarily on regular equipment inspection and maintenance plans to ensure the safe operation of electrical equipment. These plans are often based on empirical judgment and simple statistical analysis, such as conducting equipment inspections, tests, and repairs at fixed intervals. However, this approach has significant limitations. On the one hand, regular inspections fail to capture the actual operating status and potential risks of equipment in real time. Problems are often only discovered after obvious signs of failure have already appeared, making it impossible to prevent them before they occur. On the other hand, due to the lack of in-depth analysis and mining of equipment operating data, traditional methods cannot accurately assess the risk level of equipment, which can lead to over- or under-inspection, increasing operation and maintenance costs while failing to effectively improve equipment reliability. Summary of the Invention

[0004] In order to solve the above problems existing in the prior art, the present invention provides an electrical equipment risk assessment method and system based on a Bayesian network.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: The present invention provides an electrical equipment risk assessment method based on a Bayesian network, comprising the following steps: constructing a hierarchical Bayesian network model of the electrical equipment; collecting multi-source heterogeneous data in real time, wherein the multi-source heterogeneous data includes equipment status monitoring data, operating condition data, environmental data, and historical operation and maintenance data; performing feature calculation on the multi-source heterogeneous data based on a sliding time window to generate a dynamic feature vector; inputting the dynamic feature vector into a hierarchical Bayesian network model to calculate the probability of a high risk level; and triggering a risk warning when the probability of the high risk level is greater than a set dynamic probability threshold, wherein the dynamic probability threshold is set based on the criticality of the equipment.

[0006] Furthermore, the dynamic feature vector is :

[0007] in, is the mean value of the data in the time window, is the standard deviation of the data in the time window, is the rate of physical change, Cumulative count of abnormal events within the time window.

[0008] Furthermore, the high risk level probability is :

[0009] in, The observed eigenvector Under these conditions, the equipment is in a high-risk state. The probability of A high-risk state, is an implicit state, In the implicit state The probability of occurrence is high, and the device is in a high-risk state The probability of The observed eigenvector Under the condition of Probability of occurrence.

[0010] Furthermore, the dynamic probability threshold is:

[0011]

[0012] in, is the dynamic probability threshold, The criticality level of the equipment.

[0013] Furthermore, the equipment criticality level for:

[0014] in, is the per unit value of the voltage level, is the load importance factor value, The historical failure frequency.

[0015] Furthermore, the physical change rate for:

[0016]

[0017]

[0018] in, is the effective time interval, is the time interval, is the time node, is the weight coefficient, For the time point The physical quantity of The value of For the time point The physical quantity of The value of .

[0019] Furthermore, the collection of multi-source heterogeneous data also includes data preprocessing: aligning the timestamps of equipment status monitoring data and environmental data; performing one-hot encoding on discrete state quantities in operating condition data; and performing time-attenuation weighted processing on fault records in historical operation and maintenance data.

[0020] Furthermore, the construction of a hierarchical Bayesian network model for electrical equipment includes: determining the hierarchical structure of the electrical equipment, the hierarchical structure including the overall equipment layer, the component layer and the sub-component layer; defining the conditional probability distribution between nodes at each level based on the operating principle, failure mode and logical relationship between the levels of the electrical equipment; and performing parameter learning on the conditional probability distribution based on historical data and expert experience to obtain the parameters of the hierarchical Bayesian network model.

[0021] Furthermore, after triggering the risk warning, the system also includes: recording the dynamic feature vector, high risk level probability and corresponding time information when the risk warning is triggered; generating a risk warning report based on the recorded information, the risk warning report including the equipment identification and warning time; and sending the risk warning report to the terminal device of the relevant operation and maintenance personnel.

[0022] The present invention also provides an electrical equipment life assessment system based on Weibull distribution, comprising: a construction module: used to construct a hierarchical Bayesian network model of electrical equipment; an acquisition module: used to acquire multi-source heterogeneous data in real time, the multi-source heterogeneous data including equipment status monitoring data, operating condition data, environmental data and historical operation and maintenance data; a first calculation module: used to perform feature calculation on the multi-source heterogeneous data based on a sliding time window to generate a dynamic feature vector; a second calculation module: used to input the dynamic feature vector into a hierarchical Bayesian network model to calculate the probability of a high risk level; an early warning module: used to trigger a risk warning when the probability of a high risk level is greater than a set dynamic probability threshold, and the dynamic probability threshold is set based on the criticality of the equipment.

[0023] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention provides an electrical equipment risk assessment method based on a Bayesian network. The method collects multi-source heterogeneous data in real time and performs feature calculations to generate dynamic feature vectors. The method can capture changes in the equipment's operating status in real time and promptly identify potential risks. By constructing a hierarchical Bayesian network model, the method can more accurately reflect the logical relationships and fault propagation mechanisms between various levels of electrical equipment, thereby improving the accuracy of risk assessment. Dynamic probability thresholds are set based on the criticality of the equipment, making risk warnings more targeted and flexible, avoiding excessive or insufficient maintenance, reducing operation and maintenance costs, and improving equipment reliability.

[0024] The present invention provides an electrical equipment risk assessment method based on a Bayesian network. When a risk warning is triggered, it records the dynamic feature vector, high risk level probability and corresponding time information at the triggering moment, and generates a risk warning report containing detailed information such as equipment identification and warning time, providing comprehensive risk information to operation and maintenance personnel.

[0025] The present invention provides an electrical equipment risk assessment method based on a Bayesian network. Through real-time risk assessment and accurate risk prediction, it helps operation and maintenance personnel to promptly discover and handle potential faults, avoid production interruptions and increased maintenance costs caused by equipment failures, and reduce the occurrence of over-maintenance or under-maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of a method for risk assessment of electrical equipment based on Bayesian networks according to the present invention; Figure 2 This is a structural diagram of an electrical equipment risk assessment system based on a Bayesian network according to the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0028] Example 1 The present invention provides a method for risk assessment of electrical equipment based on Bayesian network, such as Figure 1As shown, the method includes the following steps: constructing a hierarchical Bayesian network model of electrical equipment; collecting multi-source heterogeneous data in real time, wherein the multi-source heterogeneous data includes equipment status monitoring data, operating condition data, environmental data and historical operation and maintenance data; performing feature calculation on the multi-source heterogeneous data based on a sliding time window to generate a dynamic feature vector; inputting the dynamic feature vector into the hierarchical Bayesian network model to calculate the probability of a high risk level; the high risk level probability is greater than a set dynamic probability threshold, triggering a risk warning, wherein the dynamic probability threshold is set based on the criticality of the equipment.

[0029] Constructing a hierarchical Bayesian network model for electrical equipment includes determining the hierarchical structure of the electrical equipment, where each hierarchical structure includes an overall equipment layer, a component layer, and a sub-component layer. In this embodiment, the layers are structured using a transformer as an example. The overall equipment layer represents the entire transformer device; the component layer includes major components such as the core, windings, and oil tank; and the sub-component layer, such as windings, can be divided into high-voltage windings and low-voltage windings. The logical relationships between the layers are clarified. For example, the overall state of the equipment is affected by the state of its components, which in turn depends on the state of its sub-components. This hierarchical structure helps to more clearly describe the propagation path and influencing factors of equipment failures.

[0030] A core failure can cause the winding temperature to rise, affecting the overall transformer operation. Based on historical data and expert experience, a conditional probability distribution is determined between nodes at each level. For example, the probability of a winding temperature rise given a known core failure can be determined based on historical failure records and expert assessments. Assuming that, in historical data, core failures leading to winding temperature rise account for 70% of all core failure cases, this probability can be set as a parameter in the conditional probability distribution.

[0031] A large amount of historical transformer operating data, including equipment status monitoring data and fault records, is collected. Using this data, Bayesian parameter learning methods are used to learn the parameters of the conditional probability distribution of the hierarchical Bayesian network model. For example, using maximum likelihood estimation or Bayesian estimation methods, the parameters of the conditional probability distribution are adjusted based on the frequency and interrelationships of each node state in the historical data, enabling the model to better fit the actual data. By continuously optimizing the parameters, the accuracy and reliability of the model are improved.

[0032] Real-time collection of multi-source heterogeneous data and equipment status monitoring data collection. Various sensors, such as temperature sensors, vibration sensors, and partial discharge sensors, are installed on transformers to collect real-time data on the transformer's operating status. Temperature sensors can be installed in key locations such as the windings and core to monitor temperature changes in real time. Vibration sensors can detect transformer vibration to determine potential mechanical failures. Partial discharge sensors can detect partial discharge within the transformer to identify insulation faults in advance. Collected data is transmitted to the data processing center via a data acquisition system to ensure real-time and accurate data.

[0033] Operation condition data collection: obtains transformer operation condition data such as load current, voltage, power factor, etc. from the power system monitoring system, reflecting the actual operation load of the transformer.

[0034] Environmental data collection: Install environmental monitoring equipment around the transformer to collect environmental data such as temperature, humidity, and wind speed. Environmental factors have a significant impact on the operating status and lifespan of the transformer. For example, high temperature and high humidity can accelerate the aging of the transformer's insulation. Correlating environmental data with equipment status monitoring data and operating condition data provides more comprehensive information for risk assessment.

[0035] Historical operation and maintenance data collection: Obtain historical operation and maintenance data of transformers from the power company's operation and maintenance management system, including fault records, maintenance records, test reports, etc. This records the transformer's past fault conditions, maintenance measures, and test results.

[0036] After collecting multi-source heterogeneous data, data preprocessing is performed. The preprocessing includes aligning the timestamps of equipment status monitoring data and environmental data, performing one-hot encoding on discrete state quantities in operating condition data, and performing time-attenuation weighted processing on fault records in historical operation and maintenance data.

[0037] Timestamp alignment occurs because the collection frequencies of device condition monitoring data and environmental data may differ, resulting in inconsistent timestamps. To facilitate subsequent feature calculation and analysis, these data need to be timestamp aligned. For example, interpolation can be used to align data of varying frequencies to the same time interval, making the data comparable across the time dimension. For example, if device condition monitoring data is collected every minute and environmental data every five minutes, linear interpolation can be used to interpolate the environmental data to every one minute.

[0038] One-hot encoding: Operating condition data may contain discrete state variables, such as the transformer's operating status (normal operation, shutdown, maintenance, etc.). To facilitate Bayesian network model processing, these discrete state variables need to be one-hot encoded. For example, "normal operation" is encoded as [1, 0, 0], "shutdown" is encoded as [0, 1, 0], and "maintenance" is encoded as [0, 0, 1]. One-hot encoding converts discrete state variables into numeric vectors, facilitating probability calculations in the model.

[0039] The impact of fault records in historical O&M data on current equipment risk may gradually diminish over time. To more effectively reflect the impact of fault records on current risk assessments, time-decay weighting is required. For example, an exponential decay function can be used to assign different weights to fault records based on the difference between the fault occurrence time and the current time. This time-decay weighting allows more recent fault records to have a greater impact on the current risk assessment.

[0040] Determine the sliding window size and sliding step size based on the transformer's operating characteristics and risk assessment requirements. For example, choose a sliding window size of 1 hour and a sliding step size of 10 minutes. The sliding window size should be large enough to reflect the changing operating status of the equipment over a period of time, while the sliding step size should be small enough to ensure real-time feature calculation.

[0041] Calculate the mean of equipment status monitoring data, operating condition data, and environmental data within a sliding time window. For example, calculate the average value of winding temperature and load current within the time window. Calculate the standard deviation of data within the time window. , the standard deviation of the data within the time window Reflects the fluctuation of the data. The larger the standard deviation, the more severe the data fluctuation is and the more unstable the equipment operation status may be.

[0042]

[0043]

[0044]

[0045] in, is the effective time interval, is the time interval, is the time node, is the weight coefficient, For the time point The physical quantity of The value of For the time point The physical quantity of The value of .

[0046] Physical change rate It reflects the rate of change of the equipment operating parameters within the time window, which is helpful for judging whether there are abnormal changes in the equipment.

[0047] Count the number of abnormal events that occurred within the time window, such as the number of partial discharges, the number of temperature exceeding the limit, etc. Accumulated count of abnormal events The more there are, the higher the risk of equipment failure may be.

[0048] Based on the mean of data within the time window , standard deviation of data within the time window , rate of physical change , cumulative count of abnormal events within the time window Combined into dynamic feature vector , dynamic feature vector It can fully reflect the operating status and risk characteristics of the equipment within the sliding event window.

[0049] The dynamic feature vector is input into the hierarchical Bayesian network model to calculate the probability of high risk level. First, according to the structure and conditional probability distribution of the hierarchical Bayesian network model, the Bayesian formula is used to calculate the probability of high risk level when the feature vector is observed. Under the condition of Probability of occurrence .

[0050] For example, assuming the implicit state For "winding insulation aging", according to the Bayesian formula = ,in In the implicit state Under the conditions that occur, the characteristic vector is observed The probability of can be obtained through the conditional probability distribution of the model; Implicit state The prior probability of can be determined based on historical data and expert experience; is the eigenvector The marginal probability of can be calculated using the total probability formula.

[0051]

[0052] according to Calculate the probability of high risk level, where The observed eigenvector Under these conditions, the equipment is in a high-risk state. The probability of A high-risk state, is an implicit state, In the implicit state The probability of occurrence is high, and the device is in a high-risk state The probability of The observed eigenvector Under the condition of By calculating the probability of high risk level, the risk level of the equipment in its current operating state can be assessed.

[0053] Calculating equipment criticality level

[0054]

[0055] in, is the per unit value of the voltage level, is the load importance factor value, The historical failure frequency.

[0056] For example, assuming the transformer voltage level is 220kV, the voltage level per unit value V can be calculated based on the ratio of the voltage level to the reference voltage; the load importance coefficient value L can be evaluated based on the importance of the load carried by the transformer, such as 1 for important loads and 0.5 for general loads; the historical fault frequency F can be obtained from historical operation and maintenance data.

[0057] Dynamic probability threshold

[0058]

[0059]

[0060] in, is the dynamic probability threshold, The criticality level of the equipment.

[0061] Calculating dynamic probability thresholds ,The higher the critical level of the equipment, the larger the dynamic probability threshold, that is, the risk warning requirements for critical equipment,are more stringent.

[0062] When the high risk level probability Greater than the set dynamic probability threshold , triggering a risk warning. After the risk warning is triggered, the dynamic feature vector, high risk level probability, and corresponding time information at the time of the risk warning are recorded; a risk warning report is generated based on the recorded information, including the device identification and warning time; and the risk warning report is sent to the terminal device of the relevant operation and maintenance personnel.

[0063] For example: setting a dynamic probability threshold is 0.8. At a certain moment, the dynamic feature vector collected by the data acquisition module is The vibration frequency is 50Hz, the motor current is 10A, and the ambient temperature is 30℃. The input is into the risk prediction model to calculate the probability of high risk level. The value of 0.85 is 0.85. Since 0.85 is greater than 0.8, a risk warning is triggered. The dynamic eigenvector at the time of triggering is recorded: vibration frequency 50 Hz, motor current 10 A, ambient temperature 30°C, high risk level probability 0.85, and warning time. A risk warning report is generated and emailed to the terminal device of the engineer responsible for equipment maintenance. After receiving the warning report, the engineer promptly inspected and maintained the equipment, avoiding a possible production accident.

[0064] Example 2 The present invention provides an electrical equipment risk assessment system based on Bayesian network, comprising a construction module, a collection module, a first calculation module, a second calculation module, and an early warning module. Figure 2 shown.

[0065] Construction module: used to construct a hierarchical Bayesian network model of electrical equipment; acquisition module: used to collect multi-source heterogeneous data in real time, and the multi-source heterogeneous data includes equipment status monitoring data, operating condition data, environmental data and historical operation and maintenance data; first calculation module: used to perform feature calculation on the multi-source heterogeneous data based on a sliding time window to generate a dynamic feature vector; second calculation module: used to input the dynamic feature vector into the hierarchical Bayesian network model to calculate the probability of a high risk level; early warning module: used to trigger a risk warning when the probability of a high risk level is greater than a set dynamic probability threshold, and the dynamic probability threshold is set based on the criticality of the equipment.

[0066] The electrical equipment risk assessment system based on Bayesian network provided by the present invention can implement the same method steps as the above method, so it will not be described in detail.

[0067] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

Claims

1. A method for risk assessment of electrical equipment based on Bayesian networks, characterized in that: The following steps are involved: Construct a hierarchical Bayesian network model of electrical equipment; Real-time collection of multi-source heterogeneous data, including equipment status monitoring data, operating condition data, environmental data, and historical operation and maintenance data; Performing feature calculation on the multi-source heterogeneous data based on a sliding time window to generate a dynamic feature vector; Inputting the dynamic feature vector into a hierarchical Bayesian network model to calculate the probability of a high risk level; The high risk level probability is greater than a set dynamic probability threshold, triggering a risk warning, and the dynamic probability threshold is set based on the criticality of the equipment.

2. The electrical equipment risk assessment method based on Bayesian network according to claim 1, characterized in that: The dynamic feature vector is : in, is the mean value of the data in the time window, is the standard deviation of the data in the time window, is the rate of physical change, Cumulative count of abnormal events within the time window.

3. The electrical equipment risk assessment method based on Bayesian network according to claim 2, characterized in that: The high risk level probability is : in, The observed eigenvector Under these conditions, the equipment is in a high-risk state. The probability of A high-risk state, is an implicit state, In the implicit state The probability of occurrence is high, and the device is in a high-risk state The probability of The observed eigenvector Under the condition of Probability of occurrence.

4. The electrical equipment risk assessment method based on Bayesian network according to claim 1, characterized in that: The dynamic probability threshold is: in, is the dynamic probability threshold, The criticality level of the equipment.

5. The electrical equipment risk assessment method based on Bayesian network according to claim 4 is characterized in that: The criticality level of the equipment for: in, is the per unit value of the voltage level, is the load importance factor value, The historical failure frequency.

6. The electrical equipment risk assessment method based on Bayesian network according to claim 2, characterized in that: The physical change rate for: in, is the effective time interval, is the time interval, is the time node, is the weight coefficient, For the time point The physical quantity of The value of For the time point The physical quantity of The value of .

7. The electrical equipment risk assessment method based on Bayesian network according to claim 1, characterized in that: The acquisition of multi-source heterogeneous data also includes data preprocessing: Align the timestamps of equipment status monitoring data and environmental data; Perform one-hot encoding on discrete state quantities in operating condition data; Perform time-attenuation weighted processing on fault records in historical operation and maintenance data.

8. The electrical equipment risk assessment method based on Bayesian network according to claim 1, characterized in that: The construction of the hierarchical Bayesian network model of the electrical equipment includes: Determining hierarchical structures of the electrical equipment, wherein the hierarchical structures include an overall equipment layer, a component layer, and a sub-component layer; Define the conditional probability distribution between nodes at each level based on the operating principle, failure mode and logical relationship between each level of electrical equipment; Based on historical data and expert experience, parameter learning is performed on the conditional probability distribution to obtain parameters of the hierarchical Bayesian network model.

9. The electrical equipment risk assessment method based on Bayesian network according to claim 1, characterized in that: The triggering of the risk warning also includes: Record the dynamic feature vector, high risk level probability and corresponding time information when the risk warning is triggered; Generate a risk warning report based on the recorded information, the risk warning report including equipment identification and warning time; The risk warning report is sent to the terminal devices of relevant operation and maintenance personnel.

10. An electrical equipment risk assessment system based on Bayesian network, characterized in that: include: Building module: used to build a hierarchical Bayesian network model of electrical equipment; Acquisition module: used to collect multi-source heterogeneous data in real time, including equipment status monitoring data, operating condition data, environmental data and historical operation and maintenance data; A first calculation module is used to perform feature calculation on the multi-source heterogeneous data based on a sliding time window to generate a dynamic feature vector; A second calculation module is used to input the dynamic feature vector into a hierarchical Bayesian network model to calculate the probability of a high risk level; Early warning module: used to trigger risk early warning when the probability of the high risk level is greater than the set dynamic probability threshold, and the dynamic probability threshold is set based on the criticality of the equipment.

Images