Scheduling automation master station system risk state assessment method considering human errors

Abstract

The invention discloses a human error-considered scheduling automation master station system risk state assessment method, which comprises the following steps of: 1, identifying a main risk event according to the operation condition of a scheduling automation master station system; 2, constructing a risk characterization parameter system considering human errors, and constructing a risk assessment model in combination with a Bayesian network; 3, determining a node variable state; step 4, determining Bayesian network parameters; 5, calculating a risk assessment result of the risk event according to Bayesian reasoning; 6, verifying the performance of the model, and ensuring the modeling accuracy; 7, sorting the importance of the risk events; and 8, based on an evidence reasoning technology of weight distribution and matrix analysis, fusing risk assessment results of a plurality of risk events. According to the method, the risk assessment result of the dispatching automation master station system can be accurately reflected, various faults generated in operation of the dispatching automation master station system can be effectively avoided, and an effective solution can be provided for a scientific and reasonable dispatching process.

Description

Technical Field

[0001] The invention belongs to the technical field of dispatching automation master stations, and in particular relates to a method for assessing risk status of a dispatching automation master station system taking human errors into account. Background Art

[0002] With the rapid development of social economy and the continuous innovation and progress of science and technology, the scale of power grid is constantly expanding, and the management mode of power grid is also tending to be automated and intelligent. In order to promote the efficiency of power supply of power grid, the power grid dispatching automation master station system has gradually been widely popularized. The emergence of emerging automated dispatching technology has gradually replaced manual dispatching technology, which has greatly improved the performance and efficiency of traditional power transmission.

[0003] Although dispatching automation technology has gradually replaced manual dispatching technology and can effectively reduce the adverse effects of human error on the power system, the impact of human error on the power grid dispatching automation master station system cannot be ignored. In the actual power grid operation process, the dispatching automation master station system is often affected by various factors, such as natural disasters, equipment damage, human error, etc., which will cause failures and cause losses of varying degrees. At the same time, it will also greatly reduce the accuracy of power dispatching and cause unscientific resource allocation.

[0004] In addition, it is not uncommon for the dispatching automation master station to fail due to human factors. There are still many problems in how to effectively reduce the impact of human factors on the dispatching system. With the increasing maturity of big data technology, a large amount of data information will be retained during the operation of the dispatching automation master station system, and then this data information can be used to establish an effective control model. However, due to the complexity of the operation process and various functions of the dispatching automation master station system, and the susceptibility to various risk factors, the expert knowledge accumulated by the dispatchers, operation and maintenance personnel, and management personnel during the operation is usually ignored, resulting in poor interpretability. Therefore, it is necessary to integrate expert knowledge and data information, and propose a method for assessing the functional failure risk of the dispatching automation master station system taking into account human errors, so as to effectively avoid various types of failures in the operation of the dispatching automation master station system and provide an effective solution for a scientific and reasonable dispatching process. Summary of the invention

[0005] In view of the problems existing in the above-mentioned prior art, the present invention provides a method for assessing the risk status of a dispatching automation master station system taking human errors into account. The method has simple steps and strong practicability. It can accurately reflect the risk assessment results of the dispatching automation master station system, and can effectively avoid various types of faults generated by the dispatching automation master station system during operation, and can provide an effective solution for a scientific and reasonable power grid dispatching process.

[0006] In order to achieve the above object, the present invention provides a method for assessing the risk status of a dispatching automation master station system taking into account human errors, comprising the following steps:

[0007] Step 1: Identify major risk events based on the operation status of the dispatching automation master station system;

[0008] The sensor group collects the operation data of the dispatching automation master station system, analyzes the operation of the dispatching automation master station system, and then identifies the risk components of the dispatching automation master station system based on the analysis results. The risk components specifically include communication failure M1, data collection failure M2, historical data failure M3 and monitoring control failure M4;

[0009] Step 2: Construct a risk characterization parameter system that takes into account human errors and build a risk assessment model using a Bayesian network;

[0010] According to the operation status of the dispatching automation master station system, the risk characterization parameters of human error, time delay and additional loss are introduced, and a Bayesian network risk assessment model based on FMECA is constructed to measure and judge the severity of the consequences of risk occurrence;

[0011] Step 3: Determine the node variable status;

[0012] According to the constructed FMECA-based Bayesian network risk assessment model, the number of Bayesian nodes is determined, and five states are set for each node, which are recorded as state 1, state 2, state 3, state 4 and state 5, representing the low probability, lower probability, medium probability, higher probability and high probability of the event occurring at the node, respectively;

[0013] Step 4: Determine the Bayesian network parameters;

[0014] First, the trapezoidal fuzzy function and expert knowledge are combined to obtain the corresponding trapezoidal function for each state of the node according to the fuzzy rating, and the fuzzy rating is combined with the expert knowledge trapezoidal function to obtain the prior probability of the five states of the node; then the expert knowledge evaluation matrix is ​​used to obtain the conditional probability distribution of the node;

[0015] Step 5: Calculate the risk assessment results of risk events based on Bayesian reasoning;

[0016] The obtained node prior probability and conditional probability are input into the established FMECA-based Bayesian network risk assessment model for reasoning. Specifically, the Bayesian network node X = {X1, X2, ... X n}, and obtain the probability distribution of risk events according to the joint probability distribution P(X);

[0017]

[0018] Where Pa(X i ) is X i The parent node of p(X i |Pa(X i )) represents the probability of the child node occurring under the condition that the parent node event occurs;

[0019] Step 6: Verify model performance and ensure modeling accuracy;

[0020] The correctness of the modeling is verified through sensitivity analysis; if a slight increase / decrease in the prior subjective probability of each input node leads to a relative increase / decrease in the posterior probability value of the output node, and if the total impact of the combined probability change from the x attribute on the risk priority value is always greater than the impact from the xy attribute set, where y∈x, then the correctness of the Bayesian network risk assessment model based on FMECA meets the requirements;

[0021] Step 7: Rank the importance of risk events;

[0022] According to the probability distribution of the risk event nodes obtained in step 5, the state probability distribution of each risk event is converted into a corresponding importance representation value using the utility function, and the importance of the risk event is judged by comparing the importance representation values. Specifically, the importance representation value CI is obtained according to formula (2), and then the importance of the risk events is ranked according to the size of the importance representation value CI.

[0023] CI=P(C)×U m (2);

[0024] Where P(C) represents the state probability distribution of each risk event; U m represents the utility function, U m =(1, 25, 50, 75, 100) T ;

[0025] Step 8: Based on the evidence reasoning technology of weight allocation and matrix analysis, integrate the risk assessment results of multiple risk events;

[0026] All risk event results are integrated through evidence reasoning based on weight allocation and matrix analysis to achieve risk status assessment from an overall perspective. The specific process is as follows:

[0027] S81: Let the risk events be A and B respectively, let the matrix A = (a1, a2, a3, a4, a5), B = (b1, b2, b3, b4, b5);

[0028] S82: Calculate the weight of each risk event using the fuzzy comprehensive evaluation method, and multiply the corresponding weight by the corresponding matrix;

[0029] S83: Use formula (3) to convert the transposed matrix A after weight distribution T Multiply it with matrix B to get the fusion matrix M1;

[0030]

[0031] S84: Understand the probability distribution of the operation risk of the dispatching automation master station system according to the fusion matrix M1, and judge the risk status of the dispatching automation master station system according to the distribution results. If the risk of the dispatching automation master station system is too high, correct the risk events according to the importance ranking of the risk events until the risk of the dispatching automation master station is lower than the set risk threshold.

[0032] In the present invention, firstly, the main risk events are identified according to the operation status of the dispatching automation master station system, which can effectively reduce the amount of calculation in the subsequent evaluation process, help save transportation capacity, and help to quickly obtain the evaluation results. Then, human errors are incorporated into the risk status assessment process, a new risk characterization parameter system taking human errors into account is constructed, and a Bayesian network risk assessment model based on FMECA is constructed in combination with the Bayesian network, which can facilitate the use of the assessment module to more accurately and quickly reflect the system operation status, which is conducive to accurately measuring and distinguishing the severity of the consequences of risk occurrence. Next, determine the node variable state, and use the fuzzy trapezoidal function combined with expert knowledge to solve the node's prior probability distribution, use the expert judgment matrix to obtain the node conditional probability distribution, and further input the node prior probability and conditional probability into the Bayesian network risk assessment model based on FMECA for reasoning calculation, so that the risk probability distribution of the local operation of the dispatching automation master station system can be accurately understood, thereby accurately reflecting the local risk situation of the dispatching automation master station system according to the probability distribution of local risk events; further, based on the change of probability, the accuracy of the model is verified through sensitivity analysis, which can conveniently and reasonably determine the correctness of the modeling, thereby ensuring the accuracy of the risk assessment results. Then, the utility function is used to rank the importance of local risk events, and then based on the evidence reasoning technology of weight distribution and matrix analysis, multiple local risk events are integrated to obtain the overall risk assessment results, which can accurately reflect the overall risk situation of the dispatching automation master station system by first evaluating locally and then integrating the whole.

[0033] As a probabilistic graphical model, Bayesian network consists of a structural model and parameter distribution. It can effectively combine expert knowledge and data information, clearly express the causal relationship between nodes, and is an effective uncertain information expression and reasoning tool. FMECA is a robust risk analysis method widely used in various high-risk industries. FMECA can analyze each device or system and data set that may fail and evaluate personal ownership, and can be applied to each system component or process. Combining the two effectively can more accurately assess risks. However, in traditional risk analysis methods, human errors are often ignored because they are difficult to quantify and their impact on risk assessment is ignored. To this end, the present invention incorporates human errors into the risk characterization parameter system, and then in the original risk characterization parameter system, fully considers the human risk factors, and on this basis, Bayesian modeling is performed. In this way, considering human errors in the dispatching automation master station system is of great significance for assessing the risk status of the dispatching automation master station system, and more accurate risk assessment results can be obtained.

[0034] The method has simple steps and strong practicality. It effectively combines the limitations of traditional failure mode effects and criticality analysis (FMECA) and the fact that the dispatching automation master station system usually ignores the human error factor, reconstructs a new risk characterization parameter system, incorporates human error into it, and adopts a hierarchical structure to more accurately evaluate risk events. In particular, the present application innovatively evaluates the importance of local risk events first and then sorts them, and then uses an evidence reasoning algorithm based on weight allocation and matrix analysis to effectively integrate the risk probability distribution results of each local risk event. That is, the overall risk situation of the system can be scientifically and reasonably reflected by integrating multiple local risk assessment results. In this way, the stringent requirements on computing power for directly evaluating the overall dispatching automation master station system can be avoided, and at the same time, the amount of computing in the evaluation process can be significantly reduced. This method can effectively save computing power, and on the basis of significantly reducing the amount of computing, quickly and accurately realize the risk assessment of the dispatching automation master station system. It can accurately and efficiently assess the risk status of the dispatching automation master station system from the perspective of the entire system. In this way, it can effectively avoid various types of faults generated during the operation of the dispatching automation master station system, effectively ensure the safe and reliable operation of the dispatching master station system, and provide an effective solution for the scientific and reasonable power grid dispatching process. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flow chart of the present invention;

[0036] Figure 2 It is a schematic diagram of the risk characterization parameter system in the present invention;

[0037] Figure 3It is a schematic diagram of the risk assessment model constructed by the Bayesian network in the present invention;

[0038] Figure 4 This is an example diagram of obtaining node parameters in the present invention;

[0039] Figure 5 It is a risk assessment bar chart of the dispatching automation master station system in the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below.

[0041] like Figures 1 to 5 As shown, the present invention provides a method for assessing the risk status of a dispatching automation master station system taking into account human errors. First, according to the operation status of the dispatching automation master station system, the main risk events are identified; then, a risk characterization parameter system taking into account human errors is constructed, and a Bayesian network risk assessment model based on FMECA is constructed in combination with a Bayesian network; the node variable state is determined, and the prior probability distribution of the node is solved by using a fuzzy trapezoidal function combined with expert knowledge, and the node conditional probability distribution is obtained by using an expert judgment matrix; the risk assessment result of the risk event is calculated according to Bayesian reasoning, and then the model performance is verified to ensure the correctness of the modeling; the risk events are ranked in importance by using a utility function; based on the evidence reasoning technology of weight distribution and matrix analysis, the risk assessment results of multiple risk events are integrated; according to the final fusion result, the system operation risk probability distribution can be understood, and the system risk status can be judged according to the distribution result, which specifically includes the following steps;

[0042] Step 1: Identify major risk events based on the operation status of the dispatching automation master station system;

[0043] The sensor group collects the operation data of the dispatching automation master station system, analyzes the operation of the dispatching automation master station system, and then identifies the risk components of the dispatching automation master station system based on the analysis results. The risk components specifically include communication failure M1, data acquisition failure M2, historical data failure M3 and monitoring control failure M4; as shown in Table 1;

[0044] Table 1: Risk events of dispatching automation master station

[0045] Risk event number Risk Events M1 communication failure M2 Data collection failure M3 History database failure M4 Monitor control failure

[0046] Step 2: Construct a risk characterization parameter system that takes into account human errors and build a risk assessment model using a Bayesian network;

[0047] The traditional failure mode effects and criticality analysis (FMECA) has the following limitations, mainly reflected in the incomplete risk characterization parameters and the lack of parameter importance differences. To solve this problem, an analysis is conducted based on the operation of the dispatching automation master station system, and human error risk characterization parameters, time delays, and additional losses are introduced. A Bayesian network risk assessment model based on FMECA is constructed to measure and determine the severity of the consequences of risk occurrence; the specific risk characterization parameter system diagram is shown in the figure below: Figure 2 As shown in the figure, the Bayesian network risk assessment model is constructed as shown in Figure 3 As shown;

[0048] Step 3: Determine the node variable status;

[0049] According to the constructed FMECA-based Bayesian network risk assessment model, the number of Bayesian nodes is determined, which is 7. Five states are set for each node, which are recorded as state 1, state 2, state 3, state 4 and state 5, representing the low probability, lower probability, medium probability, higher probability and high probability of the event occurring at the node, respectively.

[0050] Step 4: Determine the Bayesian network parameters;

[0051] First, the trapezoidal fuzzy function and expert knowledge are combined to obtain the corresponding trapezoidal function for each state of the node according to the fuzzy rating. Taking the risk event "historical database failure" as an example, Table 2 gives the fuzzy rating results of each node; the fuzzy rating is combined with the expert knowledge trapezoidal function to obtain the prior probabilities of the five states of the node. The node prior probability acquisition result is shown in the figure below. Figure 4 The final results are shown in Table 3. The expert knowledge evaluation matrix is ​​then used to obtain the conditional probability distribution of the nodes, and the results are shown in Table 5.

[0052] Table 2: Expert evaluation results of “historical database failure”

[0053] Risk Parameters Expert evaluation Trapezoidal fuzzy number O Low, Medium, Higher, High [1.2,4.2,6.7,9.1] D Lower, Lower, Medium, Higher [1.3,3.2,5.4,8.6] H Low, Lower, Medium, High [0.5,3.2,5.1,9.3] TD Low, Medium, Medium, High [0.4,4.2,5.6,8.1] EC Low, Lower, Medium, High [0.7,3.5,5.4,8.8]

[0054] Table 3: Node prior probabilities

[0055] Risk Parameters Prior probability O (0.0898,0.2107,0.2959,0.2354,0.1682) D (0.0798,0.2054,0.2566,0.2566,0.2016) H (0.0899,0.2146,0.2753,0.2463,0.1739) TD (0.0359,0.1611,0.2873,0.2873,0.2284) EC (0.0688,0.1993,0.2753,0.2619,0.1947)

[0056] Table 4: Expert decision-making knowledge matrix

[0057]

[0058]

[0059] Table 5: Conditional probabilities of some nodes

[0060]

[0061] Step 5: Calculate the risk assessment results of risk events based on Bayesian reasoning;

[0062] The obtained node prior probability and conditional probability are input into the established FMECA-based Bayesian network risk assessment model for reasoning. Specifically, the Bayesian network node X = {X1, X2, ... X n}, and obtain the probability distribution of risk events according to the joint probability distribution P(X);

[0063]

[0064] Where Pa(X i ) is X i The parent node of p(X i |Pa(X i )) represents the probability of the child node occurring under the condition that the parent node event occurs; the risk assessment results of each risk event are shown in Table 6 below;

[0065] Table 6: Risk assessment results of risk events

[0066]

[0067]

[0068] Step 6: Verify model performance and ensure modeling accuracy;

[0069] In order to verify the validity and reliability of the Bayesian network risk assessment model based on FMECA, a sensitivity analysis is required. If the constructed Bayesian network risk assessment model based on FMECA is reasonable, the sensitivity analysis should at least satisfy the following two axioms:

[0070] Axiom 1. A slight increase / decrease in the prior subjective probability of each input node should result in a relative increase / decrease in the posterior probability value of the output node.

[0071] Axiom 2: The total impact of the combined probability change from the x attribute (evidence) on the risk priority value should always be greater than the impact from the xy (y∈x) attribute (sub-evidence) set.

[0072] If the Bayesian network risk assessment model based on FMECA meets the requirements of the above two axioms, it proves that the correctness of the modeling meets the requirements;

[0073] Step 7: Rank the importance of risk events;

[0074] According to the probability distribution of the risk event nodes obtained in step 5, the state probability distribution of each risk event is converted into a corresponding importance representation value using the utility function, and the importance of the risk event is judged by comparing the importance representation values. Specifically, the importance representation value CI is obtained according to formula (2), and then the importance of the risk events is ranked according to the size of the importance representation value CI.

[0075] CI=P(C)×U m (2);

[0076] Where P(C) represents the state probability distribution of each risk event; U m represents the utility function, U m =(1, 25, 50, 75, 100) T ;

[0077] Calculation shows that the importance of risk events is ranked as follows: communication failure > historical database failure > monitoring and control failure > data acquisition failure.

[0078] Step 8: Based on the evidence reasoning technology of weight allocation and matrix analysis, integrate the risk assessment results of multiple risk events;

[0079] The above risk event assessment results can only evaluate the risk status of the dispatching automation master station system from a local perspective. Therefore, all risk event results are integrated using evidence reasoning based on weight distribution and matrix analysis to achieve a holistic risk status assessment. The specific process is as follows:

[0080] S81: Let the risk events be A and B respectively, let the matrix A = (a1, a2, a3, a4, a5), B = (b1, b2, b3, b4, b5);

[0081] S82: Calculate the weight of each risk event using the fuzzy comprehensive evaluation method, and multiply the corresponding weight by the corresponding matrix;

[0082] S83: Use formula (3) to convert the transposed matrix A after weight distribution T Multiply it with matrix B to get the fusion matrix M1;

[0083]

[0084] In the fusion matrix M1, the sum of the non-main diagonal elements is the conflict degree K between risk events A and B;

[0085] S84: Figure 5As shown, the probability distribution of the operation risk of the dispatching automation master station system is understood according to the fusion matrix M1, and the risk status of the dispatching automation master station system is judged according to the distribution results. If the risk of the dispatching automation master station system is too high, the risk events are corrected according to the importance ranking of the risk events until the risk of the dispatching automation master station is lower than the set risk threshold.

[0086] After obtaining the risk assessment results of each risk event in the dispatching automation master station, the assessment results of each risk factor are integrated using the evidence reasoning algorithm based on matrix analysis and weight distribution, and finally the following is obtained: Figure 5 The risk situation of the dispatching automation master system is shown. Figure 5 It shows that the risk index of the dispatching automation master station of the present invention is 10.28% low risk, 28.45% lower risk, 52.03% medium risk, 3.00% higher risk, and 6.24% high risk. At the same time, the risk index of the dispatching automation master station system calculated by the risk utility function is 41.7203. The probability of the master station system being under a lower risk is 38.73%, the probability of a medium risk is 52.03%, and the probability of being above a higher risk is 9.24%, indicating that the overall risk status of the system is below medium, and the risk level is relatively good, so there is no need for risk correction.

[0087] In the present invention, firstly, the main risk events are identified according to the operation status of the dispatching automation master station system, which can effectively reduce the amount of calculation in the subsequent evaluation process, help save transportation capacity, and help to quickly obtain the evaluation results. Then, human errors are incorporated into the risk status assessment process, a new risk characterization parameter system taking human errors into account is constructed, and a Bayesian network risk assessment model based on FMECA is constructed in combination with the Bayesian network, which can facilitate the use of the assessment module to more accurately and quickly reflect the system operation status, which is conducive to accurately measuring and distinguishing the severity of the consequences of risk occurrence. Next, determine the node variable state, and use the fuzzy trapezoidal function combined with expert knowledge to solve the node's prior probability distribution, use the expert judgment matrix to obtain the node conditional probability distribution, and further input the node prior probability and conditional probability into the Bayesian network risk assessment model based on FMECA for reasoning calculation, so that the risk probability distribution of the local operation of the dispatching automation master station system can be accurately understood, thereby accurately reflecting the local risk situation of the dispatching automation master station system according to the probability distribution of local risk events; further, based on the change of probability, the accuracy of the model is verified through sensitivity analysis, which can conveniently and reasonably determine the correctness of the modeling, thereby ensuring the accuracy of the risk assessment results. Then, the utility function is used to rank the importance of local risk events, and then based on the evidence reasoning technology of weight distribution and matrix analysis, multiple local risk events are integrated to obtain the overall risk assessment results, which can accurately reflect the overall risk situation of the dispatching automation master station system by first evaluating locally and then integrating the whole.

[0088] As a probabilistic graphical model, Bayesian network consists of a structural model and parameter distribution. It can effectively combine expert knowledge and data information, clearly express the causal relationship between nodes, and is an effective uncertain information expression and reasoning tool. FMECA is a robust risk analysis method widely used in various high-risk industries. FMECA can analyze each device or system and data set that may fail and evaluate personal ownership, and can be applied to each system component or process. Combining the two effectively can more accurately assess risks. However, in traditional risk analysis methods, human errors are often ignored because they are difficult to quantify and their impact on risk assessment is ignored. To this end, the present invention incorporates human errors into the risk characterization parameter system, and then in the original risk characterization parameter system, fully considers the human risk factors, and on this basis, Bayesian modeling is performed. In this way, considering human errors in the dispatching automation master station system is of great significance for assessing the risk status of the dispatching automation master station system, and more accurate risk assessment results can be obtained.

[0089] The method has simple steps and strong practicality. It effectively combines the limitations of traditional failure mode effects and criticality analysis (FMECA) and the fact that the dispatching automation master station system usually ignores the human error factor, reconstructs a new risk characterization parameter system, incorporates human error into it, and adopts a hierarchical structure to more accurately evaluate risk events. In particular, the present application innovatively evaluates the importance of local risk events first and then sorts them, and then uses an evidence reasoning algorithm based on weight allocation and matrix analysis to effectively integrate the risk probability distribution results of each local risk event. That is, the overall risk situation of the system can be scientifically and reasonably reflected by integrating multiple local risk assessment results. In this way, the stringent requirements on computing power for directly evaluating the overall dispatching automation master station system can be avoided, and at the same time, the amount of computing in the evaluation process can be significantly reduced. This method can effectively save computing power, and on the basis of significantly reducing the amount of computing, quickly and accurately realize the risk assessment of the dispatching automation master station system. It can accurately and efficiently assess the risk status of the dispatching automation master station system from the perspective of the entire system. In this way, it can effectively avoid various types of faults generated during the operation of the dispatching automation master station system, effectively ensure the safe and reliable operation of the dispatching master station system, and provide an effective solution for the scientific and reasonable power grid dispatching process.

Claims

1. A method for assessing the risk status of a dispatching automation master station system taking into account human errors, characterized in that: The steps include: Step 1: Identify major risk events based on the operation status of the dispatching automation master station system; The sensor group collects the operation data of the dispatching automation master station system, analyzes the operation of the dispatching automation master station system, and then identifies the risk components of the dispatching automation master station system based on the analysis results. The risk components specifically include communication failure M1, data collection failure M2, historical data failure M3 and monitoring control failure M4; Step 2: Construct a risk characterization parameter system that takes into account human errors and build a risk assessment model using a Bayesian network; According to the operation status of the dispatching automation master station system, the risk characterization parameters of human error, time delay and additional loss are introduced, and a Bayesian network risk assessment model based on FMECA is constructed to measure and judge the severity of the consequences of risk occurrence; Step 3: Determine the node variable status; According to the constructed FMECA-based Bayesian network risk assessment model, the number of Bayesian nodes is determined, and five states are set for each node, which are recorded as state 1, state 2, state 3, state 4 and state 5, representing the low probability, lower probability, medium probability, higher probability and high probability of the event occurring at the node, respectively; Step 4: Determine the Bayesian network parameters; First, the trapezoidal fuzzy function and expert knowledge are combined to obtain the corresponding trapezoidal function for each state of the node according to the fuzzy rating, and the fuzzy rating is combined with the expert knowledge trapezoidal function to obtain the prior probability of the five states of the node; then the expert knowledge evaluation matrix is ​​used to obtain the conditional probability distribution of the node; Step 5: Calculate the risk assessment results of risk events based on Bayesian reasoning; The obtained node prior probability and conditional probability are input into the established FMECA-based Bayesian network risk assessment model for reasoning. Specifically, the Bayesian network node X = {X1, X2, ... X n }, and obtain the probability distribution of risk events according to the joint probability distribution P(X); Where Pa(X i ) is X i The parent node of p(X i |Pa(X i )) represents the probability of the child node occurring under the condition that the parent node event occurs; Step 6: Verify model performance and ensure modeling accuracy; The correctness of the modeling is verified through sensitivity analysis; if a slight increase / decrease in the prior subjective probability of each input node leads to a relative increase / decrease in the posterior probability value of the output node, and if the total impact of the combined probability change from the x attribute on the risk priority value is always greater than the impact from the xy attribute set, where y∈x, then the correctness of the Bayesian network risk assessment model based on FMECA meets the requirements; Step 7: Rank the importance of risk events; According to the probability distribution of the risk event nodes obtained in step 5, the state probability distribution of each risk event is converted into a corresponding importance representation value using the utility function, and the importance of the risk event is judged by comparing the importance representation values. Specifically, the importance representation value CI is obtained according to formula (2), and then the importance of the risk events is ranked according to the size of the importance representation value CI. CI=P(C)×U m (2); Where P(C) represents the state probability distribution of each risk event; U m represents the utility function, U m =(1, 25, 50, 75, 100) T ; Step 8: Based on the evidence reasoning technology of weight allocation and matrix analysis, integrate the risk assessment results of multiple risk events; All risk event results are integrated through evidence reasoning based on weight allocation and matrix analysis to achieve risk status assessment from an overall perspective. The specific process is as follows: S81: Let the risk events be A and B respectively, let the matrix A = (a1, a2, a3, a4, a5), B = (b1, b2, b3, b4, b5); S82: Calculate the weight of each risk event using the fuzzy comprehensive evaluation method, and multiply the corresponding weight by the corresponding matrix; S83: Use formula (3) to convert the transposed matrix A after weight distribution T Multiply it with matrix B to get the fusion matrix M1; S84: Understand the probability distribution of the operation risk of the dispatching automation master station system according to the fusion matrix M1, and judge the risk status of the dispatching automation master station system according to the distribution results. If the risk of the dispatching automation master station system is too high, correct the risk events according to the importance ranking of the risk events until the risk of the dispatching automation master station is lower than the set risk threshold.

Images