Power generation equipment reliability dynamic evaluation method and system based on Bayesian network

Abstract

The invention provides a power generation equipment reliability dynamic evaluation method based on a Bayesian network, and belongs to the technical field of power equipment state monitoring. Through joint analysis of mutual information and Granger causality test, a coupling relationship between parameters and a health state evolution rule are accurately captured, and the reliability of power generation equipment evaluation is improved; dynamic self-adjustment of model parameters and structures is realized based on variational Bayesian inference of a sliding window and KL divergence drift detection, and the method can adapt to non-stationary environments such as equipment aging and working condition change. Besides, a forward-backward algorithm is combined with an expert knowledge base, a probabilistic health state and a quantitative reliability index are output, and closed-loop support from component-level fault early warning to system-level maintenance decision making is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment status monitoring, and particularly relates to a method and system for dynamic reliability evaluation of power generation equipment based on a Bayesian network. Background Art

[0002] In today's era, the wheels of social and economic development are rolling forward at an unprecedented pace, with all sectors experiencing vigorous growth. Industrial production continues to expand, emerging industries are mushrooming, and people's living standards are improving, leading to a surge in demand for electricity. Against this backdrop, my country's power industry has ushered in an unprecedented opportunity for development, successfully entering an era of large-scale grids, large capacities, and large generating units.

[0003] As a core component of the power system, power generation equipment acts as the "heart" of the entire system, continuously supplying energy to the system. Its reliability is directly related to the safe and stable operation of the power system. A failure in power generation equipment is like a problem with the heart, severely impacting the operation of the entire power system. At the very least, this can cause a localized power outage, bringing numerous inconveniences to residents, such as lighting interruptions and the inability to operate electrical appliances. At worst, it can trigger a large-scale power outage, leading to a series of serious consequences, including industrial production stagnation, traffic paralysis, and communication interruptions, resulting in enormous social and economic losses.

[0004] However, power generation equipment faces numerous complex factors during operation. Equipment aging is a significant issue. Over time, various components of power generation equipment gradually wear out, fatigue, and their performance declines. Just as a car's engine, transmission, and other key components lose performance and become prone to various failures after years of operation, power generation equipment is also prone to various safety hazards after long-term operation. Environmental changes are also a significant factor. Different climate conditions, such as temperature and humidity, can affect the operation of power generation equipment. For example, high temperatures increase the difficulty of heat dissipation in power generation equipment, which can easily lead to overheating and affect normal operation. Humid environments can degrade the insulation performance of equipment, increasing the likelihood of faults such as short circuits. Furthermore, improper operation is a common problem. If operators fail to strictly follow operating procedures, equipment damage or malfunctions can occur. For example, failing to follow the prescribed procedures when starting or stopping equipment can cause impacts to the mechanical components and shorten its service life.

[0005] Traditional methods for assessing the reliability of power generation equipment primarily rely on statistical data and empirical models. For a period of time, these methods provided a useful basis for assessing the reliability of power generation equipment. However, with the development of the power industry and the continuous advancement of power generation equipment technology, the limitations of these methods have gradually become apparent. They struggle to accurately reflect the dynamic changes in equipment during actual operation. During actual operation, the status of power generation equipment constantly changes with a variety of factors, including operating time, load conditions, and environmental factors. Traditional assessment methods often rely solely on static analysis based on historical data and are unable to capture these dynamic changes in a timely manner.

[0006] Furthermore, these methods often overlook the interdependencies between devices and the correlations between faults. In a power system, individual generating equipment does not exist in isolation; instead, they are complexly interconnected and interact with each other. A failure in one device can trigger a chain reaction in other devices, leading to the expansion and spread of the fault. Traditional assessment methods, which fail to account for these factors, result in inaccurate and incomplete assessment results, failing to provide reliable assurance for the safe and stable operation of the power system.

[0007] Therefore, in today's era of large power grids, large capacities, and large units, in order to improve the safety and stability of the power system, there is an urgent need for a more efficient and accurate dynamic evaluation method for the reliability of power generation equipment. Summary of the Invention

[0008] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a method and system for dynamic evaluation of power generation equipment reliability based on Bayesian networks.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for dynamic reliability evaluation of power generation equipment based on a Bayesian network, comprising the following steps: Collect time-series monitoring data of power generation equipment in real time and pre-process the collected data to generate a standardized data set; Mutual information is used to quantify the statistical correlation between monitoring parameters in the standardized data set, and Granger causality test is used to determine the temporal dependency between monitoring parameters. Based on the statistical correlation and temporal dependency, an initial dynamic Bayesian network topology is constructed; a hidden Markov model is introduced to define state nodes of the health status of the power generation equipment in the dynamic Bayesian network, and the state transition probability is initialized; Using a sliding time window mechanism, the data in the standardized data set is divided into a plurality of continuous data subsequences; using a variational Bayesian inference algorithm, the KL divergence is calculated based on the data subsequences in the current sliding window, and the conditional probability table parameters of the dynamic Bayesian network are updated; The real-time collected time-series monitoring data of power generation equipment is input into the updated dynamic Bayesian network, and the posterior probability distribution of the state node at the current moment is calculated through the forward-backward algorithm. The posterior probability distribution is integrated with the rules in the predefined expert knowledge base to output a quantitative equipment reliability assessment index.

[0010] In the step of using mutual information to quantify the statistical correlation between the monitoring parameters of the standardized data set and combining it with the Granger causality test to determine the temporal dependency between the monitoring parameters, the mutual information value between the monitoring parameters is calculated as follows:

[0011] in, represents the mutual information value of X and Y, Represents the set of cause variable parameters, such as the bearing temperature at time t-1 ; Represents the result variable parameter set, such as the vibration value at time t , represents the joint probability distribution of x and y.

[0012] For variable pairs with mutual information dependence, the lag order is The Granger test formula is as follows:

[0013] in, represents the intercept term, which is the inherent deviation of the system; is the coefficient of the Y lag term; is the coefficient of the X lag term, is the random error term, which refers to the noise or sudden disturbance not captured by the model.

[0014] Based on the statistical correlation and temporal dependency, an initial dynamic Bayesian network topology is constructed; a hidden Markov model is introduced, and state nodes of the health status of the power generation equipment are defined in the dynamic Bayesian network, and in the step of initializing the state transition probability, the state nodes have three discrete states: normal, warning, and fault.

[0015] The method for constructing the initial dynamic Bayesian network topology is as follows: Combine the mutual information value and Granger causality results to construct the directed acyclic graph of the initial dynamic Bayesian network: Node: sensor parameters and implicit status node ; Edges: Directional edges and state transfer edges based on causality testing.

[0016] The steps of using a sliding time window mechanism to divide the data in the standardized data set into a plurality of continuous data subsequences; using a variational Bayesian inference algorithm to calculate the KL divergence based on the data subsequences within the current sliding window, and updating the conditional probability table parameters of the dynamic Bayesian network are as follows: Set the time window length to split historical data into continuous subsequences ; The variational Bayesian inference algorithm is used to update the conditional probability table parameters of the Bayesian network; Compare current window Distribution window with historical benchmark , calculate the KL divergence: Based on the calculated KL divergence, if the KL divergence exceeds the preset threshold, the dynamic Bayesian network is updated; otherwise, the dynamic Bayesian network topology is reconstructed; When the KL divergence no longer changes, an updated dynamic Bayesian network is obtained.

[0017] The formula for updating the conditional probability table parameters of the Bayesian network using the variational Bayesian inference algorithm is as follows:

[0018] in, represents the variational lower bound, X is the observation data in the window, is the hidden state sequence, Represents the observed data and the hidden state at a given parameter The generative relationship below represents the variational distribution.

[0019] Compare current window Distribution window with historical benchmark , the method for calculating KL divergence is as follows:

[0020] like ,renew , when the KL divergence no longer changes, the updated dynamic Bayesian network is obtained; when multiple consecutive windows , the dynamic Bayesian network topology is reconstructed.

[0021] In the step of inputting the real-time collected time series monitoring data of the power generation equipment into the updated dynamic Bayesian network, calculating the posterior probability distribution of the state node at the current moment through the forward-backward algorithm; fusing the posterior probability distribution with the rules in the predefined expert knowledge base to output the quantitative equipment reliability evaluation index, the specific method is as follows: Forward process:

[0022] in, represents the forward probability, represents the observation probability under a given hidden state s, represents the state transition probability, represents the forward probability at time t-1; Backward process:

[0023] in, represents the backward probability, Indicates the possible state of the device at the next moment, represents the state transition probability of the device, represents the backward probability at time t+1; Posterior probability distribution:

[0024] Output quantitative equipment reliability evaluation indicators: .

[0025] In a second aspect, the present invention provides a power generation equipment reliability dynamic evaluation system based on a Bayesian network, comprising: Data acquisition and processing module, used to collect time-series monitoring data of power generation equipment in real time and pre-process the collected data to generate a standardized data set; The temporal dependency relationship establishment module is used to quantify the statistical correlation between the monitoring parameters of the standardized data set using mutual information, and to determine the temporal dependency relationship between the monitoring parameters in combination with the Granger causality test; A model building module is used to build an initial dynamic Bayesian network topology based on the statistical correlation and temporal dependency; introduce a hidden Markov model, define state nodes of the power generation equipment operating health status in the dynamic Bayesian network, and initialize state transition probabilities; A model training module is configured to segment the data in the standardized dataset into a plurality of continuous data subsequences using a sliding time window mechanism; calculate the KL divergence based on the data subsequences within the current sliding window using a variational Bayesian inference algorithm, and update the conditional probability table parameters of the dynamic Bayesian network; The reliability assessment index output module is used to input the time series monitoring data of the power generation equipment collected in real time into the updated dynamic Bayesian network, calculate the posterior probability distribution of the state node at the current moment through the forward-backward algorithm; and fuse the posterior probability distribution with the rules in the predefined expert knowledge base to output a quantitative equipment reliability assessment index.

[0026] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a dynamic evaluation method for the reliability of power generation equipment based on a Bayesian network, comprising the following steps: real-time collection of time-series monitoring data of power generation equipment and preprocessing the collected data to generate a standardized data set; using mutual information to quantify the statistical correlation between monitoring parameters of the standardized data set, and combining Granger causality test to determine the time-series dependency between the monitoring parameters; constructing an initial dynamic Bayesian network topology structure based on the statistical correlation and time-series dependency; introducing a hidden Markov model to define state nodes of the power generation equipment's operating health status in the dynamic Bayesian network and initializing state transition probabilities; using a sliding time window mechanism to segment the data in the standardized data set into a plurality of continuous data subsequences; using a variational Bayesian inference algorithm to calculate the KL divergence based on the data subsequences within the current sliding window and update the conditional probability table parameters of the dynamic Bayesian network; inputting the time-series monitoring data of the power generation equipment collected in real time into the updated dynamic Bayesian network, and calculating the posterior probability distribution of the state node at the current moment through a forward-backward algorithm; fusing the posterior probability distribution with rules in a predefined expert knowledge base to output a quantitative equipment reliability evaluation index. Through the combined analysis of mutual information and Granger causality tests, the coupling relationship between parameters and the evolution of health status are accurately captured, improving the reliability of power generation equipment assessment. Variational Bayesian inference based on sliding windows and KL divergence drift detection enable dynamic self-adjustment of model parameters and structure, adapting to non-stationary environments such as equipment aging and changing operating conditions. Furthermore, through a forward-backward algorithm combined with an expert knowledge base, probabilistic health status and quantitative reliability indicators are output, achieving closed-loop support from component-level fault warning to system-level maintenance decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0028] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0029] Example 1 like Figure 1 As shown, a dynamic evaluation method for power generation equipment reliability based on Bayesian network includes the following steps: S1: Collect the time series monitoring data of power generation equipment in real time and pre-process the collected data to generate a standardized data set; S2: Mutual information is used to quantify the statistical correlation between monitoring parameters in the standardized dataset, and Granger causality test is used to determine the temporal dependency between monitoring parameters. S3: Based on the statistical correlation and temporal dependency, construct an initial dynamic Bayesian network topology structure; introduce a hidden Markov model, define state nodes of the power generation equipment's operating health status in the dynamic Bayesian network, and initialize state transition probabilities; S4: Using a sliding time window mechanism, the data in the standardized data set is divided into a plurality of continuous data subsequences; using a variational Bayesian inference algorithm, the KL divergence is calculated based on the data subsequences in the current sliding window, and the conditional probability table parameters of the dynamic Bayesian network are updated; S5: Input the real-time collected time series monitoring data of the power generation equipment into the updated dynamic Bayesian network, calculate the posterior probability distribution of the state node at the current moment through the forward-backward algorithm; integrate the posterior probability distribution with the rules in the predefined expert knowledge base to output a quantitative equipment reliability evaluation index.

[0030] Specifically, in S1, the sensor network deployed at the key parts of the power generation equipment collects time series monitoring data in real time. The time series monitoring data includes vibration signals, temperature parameters, pressure parameters, current parameters and operating condition parameters; the collected time series monitoring data is processed by wavelet packet denoising, basis missing value filling and Min-Max normalization in turn to generate a standardized data set ; Specifically, in S2, based on the standardized data set, the statistical correlation between the monitoring parameters is quantified using mutual information, and the temporal dependency between the monitoring parameters is determined in combination with the Granger causality test. The specific method is as follows: The method of using mutual information to quantify the statistical correlation between detection parameters is as follows: calculate the mutual information value between each monitoring parameter, retain the variable pairs with mutual information values ​​higher than the threshold, and generate an undirected dependency graph.

[0031] The calculation formula of the mutual information value between each monitoring parameter is as follows:

[0032] in, represents the mutual information value of X and Y, Represents the set of cause variable parameters, such as the bearing temperature at time t-1 ; Represents the result variable parameter set, such as the vibration value at time t , represents the joint probability distribution of x and y.

[0033] For variable pairs with mutual information dependence, the lag order is Granger test for :

[0034] in, represents the intercept term, which is the inherent deviation of the system; is the coefficient of the Y lag term; is the coefficient of the X lag term, is the random error term, which refers to the noise or sudden disturbance not captured by the model.

[0035] Specifically, in S3, an initial dynamic Bayesian network topology is constructed based on the statistical correlation and temporal dependency. A hidden Markov model is introduced, and a state node representing the health status of the power generation equipment is defined in the dynamic Bayesian network. The state node has three discrete states: normal, warning, and fault, and the state transition probability is initialized. The specific method is as follows: 1) Combine mutual information values ​​and Granger causality results to construct the directed acyclic graph of the initial dynamic Bayesian network: Nodes: sensor parameters (such as temperature, vibration) and implicit status nodes ; Edge: Directional edge based on causal test (such as ) and state transition edges (such as ).

[0036] 2) Hidden Markov Model (HMM) fusion: Define hidden state nodes ; Add transition edges across time steps and observation edge Sensor node t.

[0037] 3) Parameter initialization: State transition probability : Statistical status transition frequency based on historical equipment maintenance records (for example: normal → warning probability = 0.02; warning → failure probability = 0.1).

[0038] Observation probability : For continuous sensor data, assume that it obeys Gaussian distribution and initialize the parameters: For example: The mean value of vibration in fault state is ,variance .

[0039] Specifically, in S4, a sliding time window mechanism is used to divide the data in the standardized dataset into a number of continuous data subsequences; a variational Bayesian inference algorithm is used to calculate the KL divergence based on the data subsequences in the current sliding window, and the conditional probability table parameters of the dynamic Bayesian network are updated. The specific method is as follows: S41: Set the time window length, such as W=24 hours, to split the historical data into continuous subsequences ; If the posterior probability of the state node suddenly changes (such as ), then reduce W To improve sensitivity.

[0040] S42: Update the conditional probability table parameters of the Bayesian network using the variational Bayesian inference algorithm: Objective: Maximize the variational lower bound:

[0041] in, represents the variational lower bound, X is the observation data in the window, is the hidden state sequence, Represents the observed data and the hidden state at a given parameter The generative relationship below represents the variational distribution.

[0042] E-step update: fixed parameters , (mean ,variance , transfer matrix), solve the hidden state posterior ; M-step update: update parameters .

[0043] S43: Compare current window Distribution with historical benchmark (such as data when the device is running stably), calculate the KL divergence:

[0044] Relearning trigger condition: If ,renew , when the KL divergence no longer changes, the updated dynamic Bayesian network is obtained; only when multiple consecutive windows , execute S2 again.

[0045] Specifically, in S5, the time series monitoring data of the power generation equipment collected in real time is input into the updated dynamic Bayesian network, and the posterior probability distribution of the state node at the current moment is calculated through the forward-backward algorithm. The posterior probability distribution is integrated with the rules in the predefined expert knowledge base to output a quantitative equipment reliability evaluation index. The specific method is as follows: Forward process:

[0046] in, represents the forward probability, represents the observation probability under a given hidden state s, represents the state transition probability, represents the forward probability at time t-1; Backward process:

[0047] in, represents the backward probability, Indicates the possible state of the device at the next moment, represents the state transition probability of the device, represents the backward probability at time t+1; Posterior probability distribution:

[0048] Output quantitative equipment reliability evaluation indicators: .

[0049] Example 2 A dynamic evaluation system for reliability of power generation equipment based on Bayesian networks, comprising: Data acquisition and processing module, used to collect time-series monitoring data of power generation equipment in real time and pre-process the collected data to generate a standardized data set; The temporal dependency relationship establishment module is used to quantify the statistical correlation between the monitoring parameters of the standardized data set using mutual information, and to determine the temporal dependency relationship between the monitoring parameters in combination with the Granger causality test; A model building module is used to build an initial dynamic Bayesian network topology based on the statistical correlation and temporal dependency; introduce a hidden Markov model, define state nodes of the power generation equipment operating health status in the dynamic Bayesian network, and initialize state transition probabilities; A model training module is configured to segment the data in the standardized dataset into a plurality of continuous data subsequences using a sliding time window mechanism; calculate the KL divergence based on the data subsequences within the current sliding window using a variational Bayesian inference algorithm, and update the conditional probability table parameters of the dynamic Bayesian network; The reliability assessment index output module is used to input the time series monitoring data of the power generation equipment collected in real time into the updated dynamic Bayesian network, calculate the posterior probability distribution of the state node at the current moment through the forward-backward algorithm; and fuse the posterior probability distribution with the rules in the predefined expert knowledge base to output a quantitative equipment reliability assessment index.

[0050] Finally, it should be noted that the above embodiments merely illustrate the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the present invention is defined by the appended claims, not the foregoing description. It is intended that all variations within the meaning and range of equivalents of the claims be encompassed within the present invention. Any reference signs in the claims should not be construed as limiting the claim to which they relate.

[0051] In addition, it should be understood that although this specification describes the embodiments, not every embodiment contains only one independent technical solution. This description is for clarity only. Those skilled in the art should consider the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for the purpose of illustrating the technical concept of the present invention and cannot be used to limit the scope of protection of the present invention. Any changes made based on the technical solution in accordance with the technical concept proposed by the present invention fall within the scope of protection of the claims of the present invention.

Claims

1. A dynamic evaluation method for power generation equipment reliability based on Bayesian network, characterized in that: The steps include: Collect time-series monitoring data of power generation equipment in real time and pre-process the collected data to generate a standardized data set; Mutual information is used to quantify the statistical correlation between monitoring parameters in the standardized data set, and Granger causality test is used to determine the temporal dependency between monitoring parameters. Based on the statistical correlation and temporal dependency, an initial dynamic Bayesian network topology is constructed; a hidden Markov model is introduced to define state nodes of the health status of the power generation equipment in the dynamic Bayesian network, and the state transition probability is initialized; Using a sliding time window mechanism, the data in the standardized data set is divided into a plurality of continuous data subsequences; using a variational Bayesian inference algorithm, the KL divergence is calculated based on the data subsequences in the current sliding window, and the conditional probability table parameters of the dynamic Bayesian network are updated; The real-time collected time-series monitoring data of power generation equipment is input into the updated dynamic Bayesian network, and the posterior probability distribution of the state node at the current moment is calculated through the forward-backward algorithm. The posterior probability distribution is integrated with the rules in the predefined expert knowledge base to output a quantitative equipment reliability assessment index.

2. A Bayesian network-based dynamic evaluation method for power generation equipment reliability according to claim 1, characterized in that: In the step of using mutual information to quantify the statistical correlation between the monitoring parameters of the standardized data set and combining it with the Granger causality test to determine the temporal dependency between the monitoring parameters, the mutual information value between the monitoring parameters is calculated as follows: in, represents the mutual information value of X and Y, Represents the set of cause variable parameters, such as the bearing temperature at time t-1 ; Represents the result variable parameter set, such as the vibration value at time t , represents the joint probability distribution of x and y.

3. The method for dynamic reliability evaluation of power generation equipment based on Bayesian network according to claim 2, characterized in that: For variable pairs with mutual information dependence, the lag order is The Granger test formula is as follows: in, represents the intercept term, which is the inherent deviation of the system; is the coefficient of the Y lag term; is the coefficient of the X lag term, is the random error term, which refers to the noise or sudden disturbance not captured by the model.

4. The method for dynamic reliability evaluation of power generation equipment based on Bayesian network according to claim 1, characterized in that: Based on the statistical correlation and temporal dependency, an initial dynamic Bayesian network topology is constructed; a hidden Markov model is introduced, and state nodes of the health status of the power generation equipment are defined in the dynamic Bayesian network, and in the step of initializing the state transition probability, the state nodes have three discrete states: normal, warning, and fault.

5. The method for dynamic reliability evaluation of power generation equipment based on Bayesian network according to claim 4, characterized in that: The method for constructing the initial dynamic Bayesian network topology is as follows: Combine the mutual information value and Granger causality results to construct the directed acyclic graph of the initial dynamic Bayesian network: Node: sensor parameters and implicit status node ; Edges: Directional edges and state transfer edges based on causality testing.

6. The method for dynamic reliability evaluation of power generation equipment based on Bayesian network according to claim 1, characterized in that: The steps of using a sliding time window mechanism to divide the data in the standardized data set into a plurality of continuous data subsequences; using a variational Bayesian inference algorithm to calculate the KL divergence based on the data subsequences within the current sliding window, and updating the conditional probability table parameters of the dynamic Bayesian network are as follows: Set the time window length to split historical data into continuous subsequences ; The variational Bayesian inference algorithm is used to update the conditional probability table parameters of the Bayesian network; Compare current window Distribution window with historical benchmark , calculate the KL divergence: Based on the calculated KL divergence, if the KL divergence exceeds the preset threshold, the dynamic Bayesian network is updated; otherwise, the dynamic Bayesian network topology is reconstructed; When the KL divergence no longer changes, an updated dynamic Bayesian network is obtained.

7. A Bayesian network-based dynamic evaluation method for power generation equipment reliability according to claim 6, characterized in that: The formula for updating the conditional probability table parameters of the Bayesian network using the variational Bayesian inference algorithm is as follows: in, represents the variational lower bound, X is the observation data in the window, is the hidden state sequence, Represents the observed data and the hidden state at a given parameter The generative relationship below represents the variational distribution.

8. The method for dynamic evaluation of power generation equipment reliability based on Bayesian network according to claim 6, characterized in that: Compare current window Distribution window with historical benchmark , the method for calculating KL divergence is as follows: like ,renew , when the KL divergence no longer changes, the updated dynamic Bayesian network is obtained; when multiple consecutive windows , the dynamic Bayesian network topology is reconstructed.

9. The method for dynamic reliability evaluation of power generation equipment based on Bayesian network according to claim 1, characterized in that: In the step of inputting the real-time collected time series monitoring data of the power generation equipment into the updated dynamic Bayesian network, calculating the posterior probability distribution of the state node at the current moment through the forward-backward algorithm; fusing the posterior probability distribution with the rules in the predefined expert knowledge base to output the quantitative equipment reliability evaluation index, the specific method is as follows: Forward process: in, represents the forward probability, represents the observation probability under a given hidden state s, represents the state transition probability, represents the forward probability at time t-1; Backward process: in, represents the backward probability, Indicates the possible state of the device at the next moment, represents the state transition probability of the device, represents the backward probability at time t+1; Posterior probability distribution: Output quantitative equipment reliability evaluation indicators: .

10. A power generation equipment reliability dynamic evaluation system based on a Bayesian network according to any one of claims 1 to 9, characterized in that: include: Data acquisition and processing module, used to collect time-series monitoring data of power generation equipment in real time and pre-process the collected data to generate a standardized data set; The temporal dependency relationship establishment module is used to quantify the statistical correlation between the monitoring parameters of the standardized data set using mutual information, and to determine the temporal dependency relationship between the monitoring parameters in combination with the Granger causality test; A model building module is used to build an initial dynamic Bayesian network topology based on the statistical correlation and temporal dependency; introduce a hidden Markov model, define state nodes of the power generation equipment operating health status in the dynamic Bayesian network, and initialize state transition probabilities; A model training module is configured to segment the data in the standardized dataset into a plurality of continuous data subsequences using a sliding time window mechanism; calculate the KL divergence based on the data subsequences within the current sliding window using a variational Bayesian inference algorithm, and update the conditional probability table parameters of the dynamic Bayesian network; The reliability assessment index output module is used to input the time series monitoring data of the power generation equipment collected in real time into the updated dynamic Bayesian network, calculate the posterior probability distribution of the state node at the current moment through the forward-backward algorithm; and fuse the posterior probability distribution with the rules in the predefined expert knowledge base to output a quantitative equipment reliability assessment index.

Images