A method and system for formulating a maintenance strategy to improve the reliability of a stability control system

Abstract

The application discloses a kind of maintenance strategy formulation method and system for improving the reliability of stability control system, comprising the following steps: first, collect system-level abnormal fault data in stability control system, establish fault tree model according to fault data, analyze the influence of potential failure of each component in stability control system on the reliability of overall stability control system;Second, based on the constructed fault tree model, the improved Apriori algorithm is used to analyze the influencing factors of potential failure in stability control system, and strong association rules between influencing factors and potential failure are constructed, to provide a theoretical basis for maintenance strategy to reduce failure rate;Finally, the maintenance strategy is optimized using Markov model method to improve the failure repair rate, and the stability control system failure repair time is reduced by finite time optimization control algorithm.The method of the present application aims to reduce the failure rate and failure repair time of stability control system to improve the reliability of stability control system.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power system safety and stability control, and particularly relates to a maintenance strategy making method and system for improving the reliability of a stability control system. BACKGROUND

[0002] The stability control system is a crucial component of the power system, which is responsible for maintaining the stability and reliability of the power system under various operating conditions. With the continuous expansion of the power system, the increasing number of devices, and the access of new energy and distributed power resources, the challenges faced by the stability control system are also increasing.

[0003] Traditional stability control system reliability evaluation methods mainly rely on static models, which usually assume that the state of system components is constant and do not consider factors such as device aging, maintenance activities, and external environmental changes. Static models cannot accurately capture dynamic changes, limiting the accuracy and practicality of the evaluation results.

[0004] Fault root cause analysis is a key link in the operation and maintenance of the power system, which involves identifying the initial signals of the fault from a large amount of data, understanding the causal relationship between the fault and other system behaviors, and inferring the direct and indirect causes of the fault. Traditional fault analysis methods rely on expert experience and rule-driven logic, which work well for simple faults, but when faced with complex faults and large-scale data, and when the system fails, the repair time often cannot meet the user's needs.

[0005] Therefore, there is an urgent need for a new stability control system maintenance strategy to capture the dynamic changes of the stability control system, quickly identify the causes of the fault, and shorten the repair time of the system failure. SUMMARY

[0006] To solve the problem that the static model in the prior art cannot accurately capture dynamic changes and the repair time is too long when facing complex faults, the application discloses a maintenance strategy making method and system for improving the reliability of a stability control system.

[0007] Technical scheme:

[0008] A maintenance strategy making method for improving the reliability of a stability control system, comprising:

[0009] Step 1, collect system-level fault data in the stability control system, including station communication abnormalities, system-level logic error prevention abnormalities, and inter-station / device communication abnormalities, and perform data preprocessing;

[0010] Step 2, according to the pre-processed fault data, a fault tree model is established, the upper station of the stability control system and the lower station controlled by the upper station are combined to form a hierarchical system, and the potential faults of each component in the upper station and the lower station and the corresponding top event are inferred according to the fault tree model; the top event is the most serious fault state in the fault tree model.

[0011] Step 3, the improved Apriori algorithm is used to analyze the influencing factors of the potential faults in the stability control system, whether there is a strong association rule between the influencing factors and the potential faults is judged, and the maintenance strategy is formulated according to the influencing factors, potential faults and the top event corresponding to step 2;

[0012] Step 4, the maintenance strategy is formulated, when the maintenance strategy does not meet the demand, the Markov model is used to construct a finite time optimization control algorithm, and the optimal repair time is determined.

[0013] Further, the data preprocessing of step 1 is to map the fault data into a Boolean matrix B containing only 0 and 1 a×b :

[0014]

[0015] In the formula, C cd = 0 or 1, c = 1, 2,..., a, d = 1, 2,..., b is a Boolean matrix element, a total of b fault records, a represents the length of the union set of b times of abnormal items, each column of the matrix represents a fault of the stability control system and its influencing factors and potential faults, and each row represents a fault data and information of the stability control system; if B a×b = 1 indicates that the fault b appears in the a-th data, and if B a×b = 0 indicates that the fault b does not appear in the a-th data.

[0016] Further, step 2 establishes a fault tree model, adopts a fault tree symbol, establishes a tree structure, forms a top-down system fault identification mode; sets the fault state of the system which is not expected to appear as the core target of fault analysis, connects these top events with the bottom events through a logic gate, and thus constructs a fault tree model.

[0017] Further, the Boolean structure expression of the logic gate is:

[0018] T = M1 + M2 + M3 + M4 + M5 + M6 + M7 +...

[0019] = [(X1+X2)+(X3X4)]+{[(X5+X6+X7)X8]+(X7X9)}+

[0020] (X5+X6)+(X 10 X11 )+X 12 +(X 13 +X 14 )+X5+...

[0021] After simplifying the above formula, we get:

[0022] T=X1+X2+(X3X4)+(X7X8)+(X7X9)+

[0023] X5+X6+(X 10 X 11 )+X 12 +X 13 +X 14 +...

[0024] In the formula, T is the top event; M i ,i=1,2,3... is the main event; X i ,i=1,2,3... is the bottom event, that is, the cut set; the cut set contains the failure events of components, and one cut set may contain multiple component failure events.

[0025] Further, the top event includes the main event, and the main event includes the bottom event; any fault tree model contains a limited number of minimum cut sets, and the minimum cut set is unique for the top event.

[0026] Further, the improved Apriori algorithm process of step 3 is:

[0027] The union of the abnormal items in each column of the stable control device fault Boolean matrix B generates a candidate frequent Z-1 item set, and the number S d of the "1" elements in each column of the Boolean matrix B is counted to determine whether the number of appearances meets the minimum threshold value a, and the pruning operation is performed, and if the number of appearances of the "1" elements in the d column meets S d ≤b×a%, the Boolean matrix is deleted to generate a frequent L-1 item set.

[0028] The potential fault item columns of the Boolean matrix B are subjected to logical "and" operation to generate a candidate frequent Z-2 item set; the number of "1" elements in each subset of the candidate frequent Z-2 fault item set is compared with b×a%, and if the number of "1" elements in the subset is less than b×a%, the subset is deleted to obtain a frequent L-2 item set; the number of single potential fault abnormal items in the frequent L-2 item set is counted, and if the number is less than 2, the subset containing the single abnormal item is deleted, and the fault frequent L-2 item set is updated.

[0029] The logical "and" operation is performed again on the potential fault item column of the Boolean matrix B to generate a candidate frequent Z-3 fault item set; the number of "1"s in each subset element in the candidate frequent Z-3 fault item set is compared with b x a%, if the number of "1"s in the subset is less than b x a%, the subset is deleted to obtain the frequent L-3 item set; the number of single potential fault abnormal items in the frequent L-3 item set is counted, if the number is less than 3, the subset containing the single abnormal item is deleted, and the fault frequent L-3 item set is updated;

[0030] By analogy, the fault frequent item set L-k, k > 3, is updated, and it is judged whether L is an empty set, if yes, the fault frequent item set is output, if not, the logical "and" operation is continuously performed on each subset in the fault frequent item set until the algorithm termination condition is met.

[0031] Further, the step 3 judges whether there is a strong association rule between the influencing factor and the potential fault, and the step includes:

[0032] (3.1) Each security and stability control system fault data is an item set Y, denoted as:

[0033] Y = {y1, y2,..., y n}

[0034] In the formula, y i , i = 1, 2,..., n represents each influencing factor and potential fault in the security and stability control system fault data;

[0035] (3.2) The influencing factor A and the potential fault B are two sub-item sets in Y, for an association rule The support degree S(A) of the influencing factor A, the support degree of the association rule caused by the influencing factor to the potential fault and the confidence degree C(A ) are as follows:

[0036]

[0037] In the formula, C(A) is the number of the item set element of the influencing factor A appearing in the stability control system fault data set, C(A U B) represents the number of the item set element of the influencing factor A and the item set element of the potential fault B appearing at the same time; CD(dateset) is the number of all stability control system fault data; P(A) is the probability of the item set of the influencing factor A appearing in the stability control system fault data set;

[0038] (3.3) The minimum threshold values S min and C min of the support degree and the confidence degree are set, an m-item set is called a frequent m-item set when the support degree of the m-item set meets the minimum support degree S min , and the association rule the support and the confidence satisfy the minimum threshold S simultaneously min , C min , the strong association rule is called This set of association rules is strong association.

[0039] Further, the maintenance strategy is for step 3 to obtain the influencing factors and potential faults with strong association rules, and to develop a follow-up operation scheme for the stable control system, including:

[0040] Monitor the influencing factors and set a warning threshold. When the monitoring indicators reach the warning threshold, a warning signal is sent out and maintenance is arranged;

[0041] Set up a backup component for the corresponding component with potential faults, and switch to use it when a fault occurs.

[0042] Further, the maintenance strategy is for step 3 to obtain the influencing factors and potential faults with strong association rules, and to develop a follow-up operation scheme for the stable control system, including:

[0043] (4.1) Draw a state space diagram and establish a state transition matrix Q according to the state space diagram:

[0044]

[0045] In the formula, the off-diagonal element q ij (i≠j) represents the probability of transition from state i to state j; the diagonal element represents the self-transition probability of state i, and the formula is:

[0046]

[0047] (4.2) Solve the state probability matrix P t of the stable control system at time t, and solve the system state probability at the next time according to the probability of the system being in each state:

[0048] P t = [p1 p2…p n ]

[0049]

[0050] (4.3) The repair rate refers to the probability that a component in a fault state at time t is repaired within a unit of time thereafter, usually represented by μ:

[0051]

[0052] In the formula, T is the time when the component is repaired; P is the state probability matrix of the stable control system.

[0053] The fault repair time T is represented as:

[0054]

[0055] where ξ is a random variable greater than 0 and less than 1, the failure repair time T can be effectively reduced when the failure repair rate μ is improved, and the maintenance strategy is optimized based on this;

[0056] (4.4) Based on the data and information of the fault state of the stability control system and the state information during normal operation, a smooth positive time-varying function ρ(t) is constructed:

[0057]

[0058] where ρ(0) = 1 and lim t→T ρ(t) = ρ T and 0 < ρ T ≤ 1 is a constant; when t ≥ T, ρ(t) = ρ T ; ρ T , λ and T are normal numbers;

[0059] (4.5) Design a performance function of a finite time control algorithm independent of the initial condition, combined with the characteristics of ρ(t):

[0060]

[0061] where δ is a normal number; is a monotonically decreasing function; For When t ≥ T is true

[0062] (4.6) Define the tracking error e = y - y d to determine the error of the stability control system after T time, when the tracking error is limited by , the tracking error should satisfy and satisfy the following characteristics:

[0063]

[0064] (4.7) Construct the performance function:

[0065]

[0066] ρ1(t) = ρ(t) + γ(t) tanh(Λ(t))

[0067] where γ(t) is continuously differentiable, γ(0) = 0 and 0 ≤ γ(t) < 1 - ρ(t); Λ(t) and Λ(0) = 0 are corrected by the following formula:

[0068]

[0069] where m1>0 and m2>0 are constants; Γ i (t),i=1,2 represent gains and are non-negative; v is the degree of failure within the expectation; is a severe failure outside the expectation.

[0070] The application also discloses a computer method for formulating a maintenance strategy for improving the reliability of a stability control system, comprising a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to realize the steps of the method.

[0071] The application has the beneficial effects that: a fault tree model is constructed based on fault data of the stability control system, an improved Apriori algorithm is used to analyze influencing factors of potential faults in the stability control system, strong association rules between the influencing factors and the potential faults are constructed, a theoretical basis is provided for the maintenance strategy, so as to reduce the failure rate, improve the fault repair rate, and reduce the fault repair time of the stability control system through a limited time optimization control algorithm.

[0072] The application optimizes the maintenance strategy by using the Markov model method, simulates the random change of the device state, considers the influence of device aging and maintenance activities, so as to provide more accurate reliability evaluation and better predict the behavior of the system in the future period of time, and provide more reliable decision support for power grid planning and operation. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 is a flowchart of the method for formulating a maintenance strategy for improving the reliability of a stability control system according to the application;

[0074] Figure 2 is a flowchart of the improved Apriori algorithm according to the application;

[0075] Figure 3 is a logical operation schematic diagram of the fault tree model in the embodiment of the application. DETAILED DESCRIPTION

[0076] The application will be further illustrated below in combination with the drawings and specific embodiments, and it should be understood that the following specific embodiments are only used to illustrate the application and not used to limit the scope of the application, and after reading the application, various equivalent modifications of the application by those skilled in the art all fall within the scope defined by the appended claims.

[0077] The method for formulating a maintenance strategy for improving the reliability of a stability control system according to the application is shown in Figure 1 The flowchart of the method according to the application comprises the following steps:

[0078] Step 1, collect system-level fault data such as abnormal communication between devices in the station, system-level logic anti-misoperation abnormality, abnormal communication between stations / devices, etc., and perform data preprocessing.

[0079] As the second line of defense for power grid safety, the security and stability control system ensures the safe and stable operation of the power system. The security and stability control system obtains system-level abnormal fault data such as abnormal communication between devices in the station, abnormal communication between devices in the station, system-level logic anti-misoperation abnormality, etc. through self-checking and preprocessing, and screens out abnormal fault data of the main events and underlying events corresponding to each abnormal fault.

[0080] First, the fault data of the stability control system is preprocessed to map into a Boolean matrix B containing only 0 and 1 a×b :

[0081]

[0082] In the formula, C cd = 0 or 1, c = 1, 2, …, a, d = 1, 2, …, b is a Boolean matrix element, a total of b fault records, a represents the abnormal item set length generated by b times of faults, and each column of the matrix represents a stability control system fault and its influencing factors and potential faults. Each row represents a stability control system fault data and information. If B a×b = 1 indicates that fault b occurs in the a-th data, and if B a×b = 0 indicates that fault b does not occur in the a-th data.

[0083] Step 2, establish a fault tree model according to the preprocessed fault data, form a hierarchical system with the upper station of the stability control system and the lower station controlled by the upper station, and infer the potential faults of each component in the upper station and the lower station and the corresponding top event according to the fault tree model; the top event is the most serious fault state in the fault tree model.

[0084] The component refers to each device or unit that constitutes the stability control system, including the relay protection device, the communication module in the station, the communication link between stations, the main control system, etc.; the fault of the component will affect the system operation step by step.

[0085] The fault tree model is constructed using fault tree symbols, establishing a tree structure to form a top-down system fault identification mode. The fault state that the system does not want to appear is set as the core target of fault analysis. These top events and underlying events are connected through logic gates to construct a fault tree. Through fault tree analysis, the specific top event corresponding to the potential fault of each component can be obtained, avoiding the occurrence of potential faults that cause the hierarchical system to fail, communication abnormalities, and misoperation, etc., thereby affecting the reliability of the stability control system.

[0086] The fault tree model analysis steps are first to determine the top event, second to determine the main events leading to the top event, then to find the minimal cut sets and determine the bottom events, and finally to connect the event sets through Boolean logic gates for quantitative analysis. The Boolean structure expression is:

[0087] T = M1 + M2 + M3 + M4 + M5 + M6 + M7 +...

[0088] = [(X1 + X2) + (X3 X4)] + {[(X5 + X6 + X7) X8] + (X7 X9)} +

[0089] (X5 + X6) + (X 10 X 11 ) + X 12 + X 13 + X 14 ) + X5 +...

[0090] After simplifying the above formula, we get:

[0091] T = X1 + X2 + (X3 X4) + (X7 X8) + (X7 X9) +

[0092] X5 + X6 + (X 10 X 11 ) + X 12 + X 13 + X 14 +...

[0093] In the formula, T is the top event; M i , i = 1, 2, 3... is the main event; X i , i = 1, 2, 3... is the bottom event, that is, the cut set. The cut set is the failure event of the component, and a cut set may contain multiple component failure events. The top event includes the main event, and the main event includes the bottom event, so the top event is equal to the sum of all bottom events, and when all component failures in the bottom event occur, the top event must occur. Any fault tree model contains a limited number of minimal cut sets, and the minimal cut set is unique for the top event.

[0094] As Figure 3 shown in the fault tree model in this embodiment, specifically:

[0095] A top event T contains two main events M1 and M2, and the main event M1 contains three bottom events X1, X4 and X5, and the main event M2 contains two bottom events X2 and X3, and they are all connected by an OR gate.

[0096] When the component failure causes one or more of the underlying events X1, X4 and X5 to occur, it inevitably causes the main event M1 to occur, and also inevitably causes the top event T to occur. Similarly, when the component failure causes one or more of the underlying events X2 and X3 to occur, it inevitably causes the main event M2 to occur, and also inevitably causes the top event T to occur.

[0097] For example, the failure of the communication component in the lower station will cause the upper station to be unable to obtain real-time data, thereby causing the system to fail, that is, in this case, the fault tree model in the application outputs the fault component as "the communication component of the lower station", the potential fault as "the upper station being unable to obtain real-time data", and the corresponding top event as "system failure".

[0098] Step 3: Use the improved Apriori algorithm to analyze the influencing factors of the potential fault in the stable control system, judge whether there is a strong association rule between the influencing factors and the potential fault, and formulate a maintenance strategy according to the influencing factors, the potential fault and the top event corresponding to step 2.

[0099] The traditional Apriori algorithm mainly relies on two rules: 1) If a transaction A is added to the item set Y of the non-potential fault frequent potential fault item set, then the new potential fault item set is also not a frequent item set; 2) The non-empty subset of the potential fault frequent item set is also a potential fault frequent item set. However, there are deficiencies in the algorithm process. When screening the fault frequent item set, the stable control system fault data needs to be scanned frequently, resulting in high time complexity.

[0100] By improving the Apriori algorithm, the number of scanning the stable control system fault data is reduced, thereby reducing the time complexity. The improved Apriori algorithm is as shown in Figure 2 .

[0101] The union of the abnormal items in each column of the stable control device fault Boolean matrix B generates a candidate frequent Z-1 item set, and the number S d of the "1" elements appearing in each column of the Boolean matrix B is counted. d If the number of "1" elements appearing in column d satisfies S ≤b×α%, the Boolean matrix is deleted, and a frequent L-1 item set is generated.

[0102] The Boolean matrix B potential fault item column is subjected to logical "and" operation, and a candidate frequent Z-2 item set is generated. The number of "1" elements in each subset in the candidate frequent Z-2 fault item set is compared with b×α%. If the number of "1" elements in the subset is less than b×α%, the subset is deleted, and a frequent L-2 item set is obtained. The number of single potential fault abnormal items in the frequent L-2 item set is counted. If the number is less than 2, the subset containing the single abnormal item is deleted, and the fault frequent L-2 item set is updated.

[0103] The logical AND operation is performed again on the potential fault item column of the Boolean matrix B to generate a candidate frequent Z-3 fault item set; the number of elements "1" in each subset in the candidate frequent Z-3 fault item set is compared with b x a%, if the number of elements "1" in the subset is less than b x a%, the subset is deleted to obtain the frequent L-3 item set; the number of single potential fault abnormal items in the frequent L-3 item set is counted, if the number is less than 3, the subset containing the single abnormal item is deleted, and the fault frequent L-3 item set is updated.

[0104] By analogy, the fault frequent item set L-k is updated, and it is judged whether L is empty or not, if yes, the fault frequent item set is output, if no, the logical AND operation is continuously performed on each subset in the fault frequent item set until the algorithm termination condition is met. Finally, the strong association rule between the influencing factors and the potential fault is obtained. The strong association rule provides a theoretical basis for the maintenance strategy, thereby reducing the failure rate.

[0105] The association rule can reflect the interdependence and correlation between the potential fault of the security and stability control system and the influencing factors. If there is a certain correlation between the potential fault of the system and one or more influencing factors, the correlation is mined through data mining, the most possible influencing factor of the potential fault is quickly found out, or the potential fault is predicted from the occurrence of the influencing factor. If each security and stability control system fault data is an item set, denoted as:

[0106] Y = {y1, y2,..., y n}

[0107] In the formula, y i ,i = 1, 2,..., n represents each influencing factor and potential fault in the security and stability control system fault data. For a security and stability control system fault data item set Y, if the influencing factor A and the potential fault B are two sub-item sets therein, the factor-potential fault association rule can be expressed in the form of The support degree is the percentage of A U B in the number of all security and stability control system fault data, and the confidence degree is the percentage of both A and B in the number of security and stability control system fault data.

[0108] For an association rule The support degree of the influencing factor A is:

[0109]

[0110] In the formula, C(A) is the number of the influencing factor A item set elements appearing in the stability control system fault data set; CD(dateset) is the number of all stability control system fault data; and P(A) is the probability of the stability control system fault data set appearing in the stability control system fault data set.

[0111] Support of the association rule that the impact factor causes the potential failure is:

[0112]

[0113] In the formula, C(A∪B) represents the number of simultaneous occurrence of the impact factor A item set element and the potential failure B item set element.

[0114] Support of the association rule that the impact factor causes the potential failure is:

[0115]

[0116] The minimum threshold values of the support and the confidence are set, if the minimum threshold values of the support and the confidence are S min and C min respectively, an m-item set is called a frequent m-item set when the support of the m-item set meets the minimum support S min , the support and the confidence of the association rule meet the minimum threshold values S min and C min at the same time, and the set of the association rules is called strong association.

[0117] The maintenance strategy process of the communication link priority formulated in the embodiment is as follows: firstly, the potential failure "in-station communication component failure" and the impact factor "inter-station communication link interruption" have strong association by the fault tree model and the improved Apriori algorithm, and the two may cause the top event "system-level communication interruption".

[0118] According to the analysis result, the priority maintenance strategy formulated includes: installing a temperature sensor to monitor the temperature of the in-station communication equipment in real time, and sending a warning and arranging maintenance when the temperature exceeds the standard; setting a redundant path of the inter-station communication link, so that the standby link can be quickly switched when the main link is interrupted.

[0119] Optionally, an additional maintenance strategy is formulated to regularly check the signal transmission stability and delay condition of the inter-station communication link; and the plan is dynamically adjusted according to the monitoring data, such as expanding or upgrading in advance when the communication link load increases. The implementation of the strategy significantly reduces the probability of communication link interruption, and quickly responds when a failure occurs, ensuring the reliability and stability of the stability control system.

[0120] Step 4, when the maintenance strategy does not meet the demand, a finite time optimization control algorithm is constructed by using a Markov model to optimize the maintenance strategy and improve the repair rate. The repair rate refers to the probability that an element in failure at time t is repaired in the unit time thereafter, which is usually represented by μ.

[0121] ​

[0122] In the formula, T is the time when the element is repaired; P is the state probability matrix of the stability control system.

[0123] The Markov model is used to analyze the possible states of the running system, and then the various states of the stability control system caused thereby are determined, a state space graph is drawn, and a state transition matrix Q is established according to the state space graph:

[0124]

[0125] In the formula, the non-diagonal element q ij (i≠j) represents the probability of transition from state i to state j; the diagonal element represents the self-transition probability of state i, and the formula is:

[0126]

[0127] The state probability matrix P of the stability control system at time t is solved t According to the probability of the system being in each state, the running state probability of the system at the next time is solved:

[0128] P t =[p1 p2…p n ]

[0129]

[0130] The Markov model is used to analyze the influence of different repair strategies on the state transition of the stability control system, so as to find the optimal strategy for making the stability control system transition from the fault state to the normal running state, and then improve the fault repair rate μ. The fault repair time T can be represented as:

[0131]

[0132] In the formula, ξ is a random variable greater than 0 and less than 1. When the fault repair rate μ is improved, the fault repair time T can be effectively reduced.

[0133] The finite time optimization control algorithm is constructed. When the finite time control algorithm is performed, a smooth positive time-varying function ρ(t) is constructed based on the data of the fault state of the stability control system and the state information when the system is normally running:

[0134]

[0135] In the formula, ρ(0)=1 and lim t→T ρ(t)=ρ T , and 0<ρ T ≤1 is a constant; when t≥T is consistent with the predetermined time T, ρ(t)=ρ T ; ρ Tλ and T are positive constants, designed according to user requirements. To ensure that the repair time of the stability control system meets the needs of both the user and the power grid, a suitable repair time T and ρ are determined based on historical data, user requirements, and power grid requirements. T This represents the system state converging to the desired performance level within a fixed time period, and is set as the performance requirement for the stable control system to operate normally.

[0136] Design a performance function for a finite-time control algorithm that is independent of initial conditions, taking into account the characteristics of ρ(t):

[0137]

[0138] In the formula, δ is a positive constant; It is a monotonically decreasing function; for When t≥T holds

[0139] Finally, based on the operating data of the stability control system during normal operation and failure, the tracking error e = yy is defined. d To determine the operating state error of the stability control system after time T, when using... When limiting tracking error, the tracking error should meet the following requirements: And satisfy the following characteristics:

[0140]

[0141] The initial value of the tracking error e is arbitrary. After a predetermined time T, the tracking error enters the performance boundary consisting of constant values, thereby enabling the stability control system to meet the stable operation conditions.

[0142] Furthermore, to prevent unexpected failures in the stability control system that could lead to the loss of preset performance, a system is constructed that is more robust than... A more conservative performance function has a boundary value that is greater than... A wider range of transient performance is introduced, sacrificing only when the stability control system malfunctions severely and beyond expectations, ensuring that the tracking error never reaches the performance boundary. This is achieved by introducing a correction signal Λt based on the finite-time control algorithm.

[0143]

[0144] ρ1(t)=ρ(t)+γ(t)tanh(Λ(t))

[0145] In the formula, γ(t) is continuously differentiable, γ(0) = 0 and 0 ≤ γ(t) < 1 - ρ(t). The system is corrected by the following formula for Λ(t) and Λ(0) = 0:

[0146]

[0147] where m1>0 and m2>0 are constants; Γ i (t), i = 1, 2 represent the gain and are non-negative; v is the degree of failure within the expectation; is a serious failure outside the expectation. The addition of Λ(t) realizes that the serious failure outside the expectation The constraint feedback of the modified finite-time control algorithm is introduced. By derivation of the first differential equation in the above formula, Γ(0) = 0, Γ(t) ≥ 0, and γ(t) is selected to satisfy ρ1(t) ≤ 1 and ρ1(0) = 1. Γ i (t), i = 1, 2 represent the gain and are non-negative; v is the degree of failure within the expectation; When it is within the expectation, that is, the failure, the modified finite-time control algorithm will restore to the preset performance of the finite-time control algorithm, otherwise, the modified finite-time control algorithm with wider boundary is used.

[0148] Based on the fault state data and information of the stability control system, the state information in normal operation, the repair time T satisfying the user demand and system demand, the finite-time optimization control algorithm is used, and the modified signal is introduced considering the failure outside the expectation, so that the modified finite-time control algorithm has a wider performance boundary, ensures that the stability control system can reach the required performance boundary within a limited time T under the failure within the expectation or the serious failure outside the expectation, thereby reducing the fault repair time of the stability control system.

Claims

1. A method for developing a maintenance strategy to improve the reliability of a stability control system, characterized in that, Comprise: Step 1, collect system-level fault data in the stability control system, including abnormal communication in the station, abnormal system-level logic error prevention, abnormal inter-station / inter-device communication, and perform data preprocessing; Step 2, establish a fault tree model according to the preprocessed fault data, construct a hierarchical system with the upper station of the stability control system and the lower station controlled by the upper station, and obtain the potential faults of each component in the upper station and the lower station and the corresponding top event according to the fault tree model; The top event is the most serious fault state in the fault tree model; Step 3, use the improved Apriori algorithm to analyze the influencing factors of potential faults in the stability control system, judge whether there is a strong association rule between the influencing factors and the potential faults, and formulate a maintenance strategy according to the influencing factors, potential faults and the corresponding top event in step 2 which exist strong association rules; Step 4, when the maintenance strategy does not meet the demand, use Markov model to build a finite time optimization control algorithm to optimize the maintenance strategy; Optimizing the maintenance strategy by using Markov model is based on the maintenance strategy formulated in step 3, which maintains a low failure rate while shortening the fault repair time, which specifically includes: (4.1) drawing a state space graph, establishing a state transition matrix according to the state space graph : where the off-diagonal elements represent the probability of transitioning from state to state ; the diagonal elements represent the self-transition probability of state , and the formula is: (4.2) Solve the steady control system's state probability matrix at the moment According to the probability of the system in each state, solve the system running state probability at the next moment: (4.3) The repair rate is the probability that an element that is in failure at time t is repaired in the next unit of time, usually denoted by (4.4) The failure rate is the probability that an element that is in working order at time t fails in the next unit of time, usually denoted by (4.5) The hazard rate is the ratio of the failure rate to In the formula, is the time when the element is repaired; is the state probability matrix of the stability control system; Fault repair time is represented as: In the formula, is a random variable greater than 0 and less than 1, which can effectively reduce the fault repair time when improving the fault repair rate . , and the maintenance strategy is optimized based on this.

2. The method of claim 1, wherein, The data pre-processing described in step 1 is mapping the failure data into a Boolean matrix containing only 0s and Is : In the formula, is a Boolean matrix element, and is a fault record, represents is the length of the union of abnormal items generated by the second fault, and each row of the matrix represents a stable control system fault and its influencing factors and potential faults, and each row represents a stable control system fault data and information; if represents a fault appears in the first data, if represents a fault does not appear in the first data.

3. The method of claim 2, wherein, Step 2, the fault tree model is established by using fault tree symbols to establish a tree structure and form a top-down system fault identification mode; The fault state that the system does not want to appear is set as the core target of fault analysis, and these top events and bottom events are connected through logical gates to build a fault tree model.

4. The method of claim 3, wherein, The expression of the logical gate is: In the formula, is a top event; is a bottom event, i.e. a cut set; a cut set can contain multiple component failure events.

5. The method of claim 4, wherein, The top event includes a main event, and the main event includes a bottom event; any fault tree model contains a limited number of minimal cut sets, and the minimal cut set is unique for the top event.

6. The method of claim 1, wherein, The improved Apriori algorithm process in step 3 is: Stable control device fault boolean matrix Each column of abnormal item set union produces candidate frequent Item set, statistical boolean matrix The number of times of each column "1" element appearing , judge whether the number of times meets the minimum threshold , prune the operation, if The number of times of column "1" element appearing meets , delete the boolean matrix, and generate frequent Item set; Boolean matrix Perform a logical AND operation on the potential fault item list to generate candidate frequent faults. Fault itemsets; frequently selected candidate itemsets The sum of the number of "1" elements in each subset of the fault item set If the element "1" in the subset is less than 1, then the subset is considered to be less than 1. Delete the subset, and you will get frequent failures. Itemsets; frequent statistical failures Individual potential fault anomaly items in the item set are prone to failure. If the number of items in the set is less than 2, delete the subset containing the single potential fault item, and update the fault-prone items. Itemset; Again, on the Boolean matrix Perform a logical AND operation on the potential fault item list to generate candidate frequent faults. Fault itemsets; frequently selected candidate itemsets The sum of the number of "1" elements in each subset of the fault item set If the element "1" in the subset is less than 1, then the subset is considered to be less than 1. Delete the subset, and you will get frequent failures. Itemsets; frequent statistical failures Individual potential fault anomaly items in the item set are prone to failure. If the number of items in the set is less than 3, delete the subset containing the single potential fault item and update the frequently faulty items. Itemset; Similarly, update the frequent itemsets with faults , k > 3, and determine whether the empty set is empty, and output the frequent itemsets with faults, otherwise, continue the logical "and" operation of each subset in the frequent itemsets with faults until the algorithm termination condition is met.

7. The method of claim 6, wherein, Step 3, judging whether there is a strong association rule between the influencing factors and the potential faults, specifically includes: (3.1) Each security and stability control system fault data is a set Y, denoted as: In the formula, represents each influencing factor and potential failure in the safety and stability control system failure data; (3.2) Influencing factors and potential failures are two subsets of items in Y, for an association rule , the support of influencing factors S(A), the support of association rules caused by influencing factors leading to potential failures S(A ), the confidence of association rules In the formula, is the influencing factor the number of item set elements appearing in the stable control system fault data set, represents the influencing factor the item set element and the potential fault the number of simultaneous appearances of item set elements; is the number of all stable control system fault data; is the influencing factor the probability of the appearance of the item set in the stable control system fault data set; (3.3) Set minimum threshold of support and confidence , , one item set support meets the minimum support when called frequent - item set, the support and confidence of association rules minimum threshold , , called this set of strong association rules.

8. The method of claim 7, wherein, The maintenance strategy formulated according to the influencing factors, potential faults and the corresponding top event in step 2 which exist strong association rules includes: Monitor the influencing factors, set a warning threshold, and send a warning signal when the monitoring index reaches the warning threshold to arrange maintenance; Set up a spare component for the corresponding component with potential faults, and switch to use when a fault occurs.

9. The method of claim 1, wherein, Step 4 also includes: (4.4) Constructing a smooth positive time-varying function based on data and information of the fault state of the stability control system and the state information in normal operation : wherein and ; and are constants; when coincide with a predetermined time ; ; , and are normal numbers; (4.5) Design a performance function for a finite-time control algorithm that is independent of initial conditions, combining the properties of ​ wherein is a normal number; is a monotonically decreasing function; ; for when holds , ; (4.6) defining a tracking error to determine the steady control system in the moment after the operating state error, when the tracking error is limited, the tracking error should meet and meet the following characteristics: and meet the following characteristics: (4.7) Build a performance function: wherein continuously differentiable, and ; with the system is corrected by the equation: wherein and are constants; represents the gain and is non-negative; is the degree of failure within expectation; is the serious failure outside expectation.

10. A maintenance strategy making system for improving reliability of a stability control system, characterized by, A computer program product comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program to implement the steps of any one of the methods of claims 1-9.

Images