Method for evaluating operation reliability of electrical coupling system

Abstract

The invention discloses an electrical coupling system operation reliability evaluation method, which comprises the steps of analyzing a network topology, an energy flow direction and an element logic relationship of an electrical coupling system, and constructing a Bayesian network model of the system from a power supply source to a load point; establishing an element reliability model, and obtaining different failure rate information of the non-key element by changing different age limits and load rates; taking different failure rate information of the non-key elements as input data of nodes of the Bayesian network model, outputting an analytical model coefficient, and obtaining a reliability index in combination with a key element combination state probability; training an LLSVM model by taking different failure rate information of the non-critical elements as input and taking an analytical model coefficient as output; and during online application prediction, obtaining different failure rates of corresponding non-key elements by combining element long-term and short-term prediction results with the element reliability model, inputting the different failure rates into the trained LLSVM model, and outputting a reliability index. According to the method, effective, rapid and efficient reliability evaluation of the electrical coupling system is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated energy systems, and more particularly to a method for evaluating the operational reliability of an electrical coupling system. Background Art

[0002] At present, with the sharp rise in energy demand worldwide, the over-exploitation and utilization of traditional fossil fuels are accelerating the process of environmental deterioration. Multi-energy complementary integrated energy system is one of the important measures to achieve the goal of low-carbon emissions, and the integrated electricity-gas system (IEGS) is one of the typical operating modes. However, with the close integration of electricity and natural gas systems, the complex interaction between systems increases the difficulty of operation and poses a challenge to safe and reliable operation. Reliability assessment of the integrated electricity-gas energy system can measure the risk of energy interruption with quantitative indicators, provide guidance for system planning and operation, and aim to improve system stability and safety. It is of great significance to ensure the stability of energy supply and optimize system operation, and is a key issue in current research in the energy field. Therefore, how to establish an effective, fast and efficient reliability assessment method is the key to ensuring the safe and stable operation of the electricity-gas coupled integrated energy system.

[0003] In response to the above problems, at present, from a methodology perspective, the operational reliability assessment of the electric-gas coupling system mainly adopts three methods: model-driven method, data-driven method, and model-data hybrid-driven method. The model-driven method mainly relies on the establishment of an accurate system model and evaluates reliability through theoretical calculations. This method has certain difficulties in dealing with high-dimensional nonlinear problems. The data-driven method mainly uses artificial intelligence technology to model time-varying and supply-demand uncertainties, and realizes real-time perception of the system's operating status and short-term risk prediction. The model-data hybrid-driven method combines the advantages of model-driven and data-driven, which not only ensures the accuracy of the assessment but also improves the processing speed. In terms of failure rate modeling, the modeling and algorithm basically do not consider the system operating status such as line flow and pipeline air pressure, and do not consider the impact of changes in environmental parameters on the short-term state probability of components. Statistical values ​​or means are usually used.

[0004] From the current development status at home and abroad:

[0005] Literature 1: Zhu Jizhong, Luo Tengyan, Wu Wanli, et al. Review of reliability assessment of integrated energy system operation I: Model-driven method [J]. Journal of Electrical Engineering, 2022, 37(11): 2761-2776. Analytical method and Monte Carlo simulation method are mainstream reliability assessment methods, which provide effective means for quantitatively evaluating the reliability indicators of distribution systems. However, these methods often face the problems of increased model complexity and decreased solution efficiency when dealing with high-dimensional, multi-state and high-uncertainty systems. Similarly, Huo Limin, Zhu Yongli, Fan Gaofeng, et al.

[0006] Reference 2: A new method for power system reliability assessment based on Bayesian network [J]. Automation of Electric Power Systems, 2003, 27(5): In the field of natural gas systems, researchers are also constantly exploring reliability assessment methods suitable for their special operating characteristics, such as methods based on Markov models, graph theory and Bayesian networks. These methods have improved the accuracy and flexibility of the assessment to a certain extent.

[0007] Literature 3: Yang Yongxiang, Zhu Yuancheng, Qian Taotao, Tang Qingliang, Wang Ying, Chen Xingtian, Wang Yutong. Multi-factor driven time-varying fault failure model of distribution transformer and its application [J]. Smart Grid, 2019, 9(6): 253-262. In terms of failure rate modeling, a variety of historical and future environmental factors, including meteorological conditions (such as weather, temperature, wind speed and direction, and sunshine intensity) and power system operating parameters (such as load, equipment age, line current, bus voltage stability, and system frequency fluctuations) are analyzed in detail to significantly affect the short-term reliability of transformers and transmission lines. Through detailed analysis, this study provides an important perspective for understanding and predicting the performance changes of power facilities in complex environments.

[0008] Reference 4: Sun Yuanzhang, Cheng Lin, Liu Haitao. Power system operation reliability assessment based on real-time operation status [J]. Power System Technology, 2005, (15): 6-12. Considering the influence of aging effect and operation status on the reliability of IES components, a real-time model of the failure rate of the power supply components and transmission components of the electric-gas coupling system under load rate and operating years is established, which lays a theoretical foundation for the real-time operation reliability assessment.

[0009] However, the above-mentioned literature has not really established an effective, fast and efficient reliability assessment method. Therefore, how to establish an effective, fast and efficient reliability assessment method and ensure the safe and stable operation of the electric-gas coupled integrated energy system is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0010] In view of the fact that the existing model-driven method has certain computational difficulties when dealing with high-dimensional nonlinear problems, and the shortcomings of the data-driven method, such as large errors and weak interpretability, the model-data hybrid driven method combines the advantages of model-driven and data-driven methods, which not only ensures the accuracy of the evaluation but also improves the processing speed. However, it does not consider the impact of changes in row status and environmental parameters on the short-term state probability of components. In view of this, the present invention provides a method for evaluating the operational reliability of an electrical coupling system.

[0011] In order to achieve the above object, the present invention adopts the following technical solution:

[0012] A method for evaluating the operational reliability of an electrical coupling system comprises the following steps:

[0013] S10. Analyze the network topology, energy flow and component logic relationship of the electrical coupling system, and construct a Bayesian network model of the system from energy supply to load point;

[0014] S20, based on the Bayesian network model, establish a component reliability model, and obtain different failure rate information of non-critical components by changing different years and load rates;

[0015] S30, using the different failure rate information of the non-critical components as input data of the nodes of the Bayesian network model, outputting analytical model coefficients, and obtaining reliability indicators in combination with the probability of the combined state of the critical components;

[0016] S40, using the non-critical component different failure rate information as LLSVM training data input, using the analytical model coefficients as LLSVM training data output, and training the LLSVM model;

[0017] S50, when applying online prediction, the long-term and short-term prediction results of the components are combined with the component reliability model to obtain the corresponding non-critical components with different failure rates, which are input into the trained LLSVM model to output the reliability index.

[0018] It can be seen from the above technical solutions that compared with the prior art, the present invention has the following technical advantages:

[0019] 1) The proposed fault models and operating status differences of different equipment take into account the real-time characteristics of the equipment, and the calculated results are more convincing.

[0020] 2) Compared with pure model-driven methods, this method can handle more complex models, convert network topology into a Bayesian network, and achieve higher-dimensional model solutions.

[0021] 3) Compared with pure data-driven, the present invention decomposes the analytical model into the part affected by key components and the part affected by non-key components. The data-driven method of the present invention only uses LSSVM prediction for the part affected by non-key components, while the unavailability rate of key components uses the real value, which makes the component calculation more accurate and more interpretable.

[0022] 4) When the component failure rate changes, the graphic-analog-digital hybrid drive method proposed in the present invention does not require remodeling. The change in the unavailability rate of key components only requires recalculating the probability of each combination state. The change in the unavailability rate of non-key components only requires introducing the trained LSSVM model and re-obtaining the analytical model coefficients, which can achieve rapid updating of the operation reliability evaluation indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0024] Figure 1 The topology diagram of the gas distribution system for the case analysis;

[0025] Figure 2 Schematic diagram of the Bayesian network model for the gas distribution system transformation;

[0026] Figure 3 A flow chart for realizing a hybrid drive of graphics, analog and digital provided in an embodiment of the present invention;

[0027] Figure 4 The topological structure diagram of the electric-gas coupling system for case analysis;

[0028] Figure 5 It is a schematic diagram of the error analysis of the reliability assessment results of the analytical model decomposition;

[0029] Figure 6a Comparative analysis chart of LOLP results evaluated by hybrid drive, single data drive and analytical method;

[0030] Figure 6b Comparative analysis chart of EENS results evaluated by hybrid drive, single data drive and analytical method;

[0031] Figure 7a This is a schematic diagram of the LOLP evaluation results of the operating reliability within a certain day;

[0032] Figure 7b This is a schematic diagram of the EENS evaluation results of operating reliability within a certain day;

[0033] Figure 8 Schematic diagram of reliability evaluation results considering aging effects;

[0034] Fig. 9 Schematic diagram of reliability assessment results considering coupling effects;

[0035] Fig.10a This is a schematic diagram of the LOLP rapid evaluation results of component replacement reliability achieved by the method of the present invention;

[0036] Fig.10b This is a schematic diagram of the EENS rapid evaluation results of component replacement reliability achieved using the method of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] The embodiment of the present invention discloses a method for evaluating the operation reliability of an electrical coupling system, comprising the following steps:

[0039] S10. Analyze the network topology, energy flow and component logic relationship of the electrical coupling system, and construct a Bayesian network model of the system from energy supply to load point;

[0040] S20, based on the Bayesian network model, establish a component reliability model, and obtain different failure rate information of non-critical components by changing different years and load rates;

[0041] S30, using the different failure rate information of the non-critical components as input data of the nodes of the Bayesian network model, outputting analytical model coefficients, and obtaining reliability indicators in combination with the probability of the combined state of the critical components;

[0042] S40, using the non-critical component different failure rate information as LLSVM training data input, using the analytical model coefficients as LLSVM training data output, and training the LLSVM model;

[0043] S50, when applying online prediction, the long-term and short-term prediction results of the components are combined with the component reliability model to obtain the corresponding non-critical components with different failure rates, which are input into the trained LLSVM model to output the reliability index.

[0044] The present invention proposes a rapid reliability assessment method for an electric-gas integrated energy system based on a Bayesian network and a least squares support vector machine (LSSVM). The method first constructs a logical relationship model of the system from energy supply to load points by analyzing the network topology and energy flow, and establishes a Bayesian network accordingly. When screening key components, the contribution of the betweenness index and the Bayesian contribution are comprehensively considered to ensure the rationality of the selection of key components. Furthermore, the reliability index is decomposed into the key component influencing part and the non-key component influencing part. The mapping relationship between the operating status of the non-key component and the reliability index coefficient is established through the LSSVM regression model. Combined with the key component fault information, the rapid update of the system operation reliability is achieved. The example test verifies the accuracy and efficiency of the method proposed in the present invention.

[0045] The technical solution of the present invention is described in detail below:

[0046] The method for evaluating the operation reliability of an electric-pneumatic coupling system based on a hybrid drive of graphics-analog-digital is proposed in the present invention, and the specific steps are as follows:

[0047] (1) Modeling of electric-gas coupling system based on Bayesian network

[0048] By analyzing the network topology, energy flow and logical relationship of components, a logical relationship model of the system from energy supply to load points was constructed.

[0049] According to the different functions of distribution network components, when constructing the Bayesian network, they can be divided into four types: component nodes, load nodes, system nodes and joint nodes. The relationship between nodes can be divided into "and", "or", "cause and effect" and "joint".

[0050] against Figure 1 The gas distribution system shown in the figure takes the gas source, pipeline and compressor as the first layer of the Bayesian network, the GL load element as the second layer, and the system element as the third layer. S1 is used as the main power source and P2G as the backup power source. For GL2, according to the power flow analysis, there are two gas supply paths, namely S1-P1-P2-P3-P4-GL2 and P2G-C1-GL2. For line 1, any failure of S1 and P1~P4 will make the path unavailable. For load GL2, when the first path is not available, the second path can be used for gas supply, which will not cause the load to stop operating. Therefore, the first-layer elements on a line and the load are in an "and" relationship, and the relationship between the two lines of the same load is "or". For system nodes, the different system states are caused by different load states, so there is a causal relationship between load nodes and system nodes. This case does not consider the failure and state of operating elements, so the joint relationship is not considered. "And", "Or", "Cause and Effect". According to the above network topology, energy flow and component logic relationship, a Bayesian network model of the system from energy supply to load point is constructed, such as Figure 2 shown.

[0051] For the gas-electricity coupling system, the gas turbine GT can convert the excess natural gas in the gas distribution system into electrical energy to supply power to the power distribution system. At this time, the gas turbine, gas load point and electric load power supply path are equivalent to the Bayesian parent node, and the electric load node is the child node. Only when the three parent nodes are normal can the electric load be powered, so the four form a logical "and" relationship. Similarly, when the natural gas system fails and causes insufficient energy supply, a P2G device can be used to convert electrical energy into natural gas.

[0052] (2) Analytical model for reliability assessment of electric-gas coupled system operation

[0053] ① Reliability modeling of components in electric-pneumatic coupling systems considering aging effects and load rates

[0054] The failure probability of various energy supply equipment is closely affected by its operating status, operating time and climate and environmental factors. The present invention considers the impact of operating status and service time on the unavailability rate of energy supply equipment.

[0055] 1) Energy supply component unavailability model

[0056] The unavailability of energy supply components such as generator sets, P2G equipment, and gas turbines in the electric-gas coupling system is mainly affected by the load rate. High load or long-term heavy load operation will cause the internal temperature of the energy supply component to rise, thereby shortening its service life; while low load operation is likely to increase component wear and may bring safety hazards. Therefore, the following energy supply component unavailability model can be established:

[0057]

[0058] In the formula is the basic unavailability rate of the energy supply equipment, and They are the maximum unavailability rate under low load and the maximum unavailability rate under heavy load, which can be obtained based on experience and historical data; and are the failure rate and load rate of the energy supply equipment at time t, and They are the light load and heavy load critical values ​​respectively. When the load rate is between the two, the failure rate is the basic unavailability rate.

[0059] 2) Transmission Component Unavailability Model

[0060] In the electric-gas coupling system, the transmission elements include power transmission lines and natural gas pipelines. Studies have shown that the unavailability rates of these two types of transmission elements are time-varying with the service time and load rate. After long-term operation, power transmission lines will cause aging and wear. Under high load conditions, the current, voltage and other parameters of the line will change greatly, which may easily lead to local overload or equipment damage, thereby increasing the unavailability rate. Natural gas pipelines may experience aging and corrosion problems after being in service for too long. When the load rate is high, the pipeline system has a large operating load, which may easily cause fatigue and loss.

[0061] The influence of line power and pipeline pressure on the unavailability of transmission lines and pipelines can be described by the following formula:

[0062]

[0063] Where: F is the statistical failure probability of the line (pipeline); l 、F l rated and F l maxThey are line power (pipeline pressure), rated power (pressure) and transmission limit power (pressure).

[0064] Considering the influence of line aging effect on line (pipeline) unavailability rate, assuming that the line (pipeline) service time is t ε , t ε Annual unavailability rate It can be described as:

[0065]

[0066] Where f(t ε ) is selected as the Weibull distribution function, which is generally used for statistical reliability or life testing. This embodiment uses a two-parameter Weibull distribution function, and its probability density function is:

[0067]

[0068] Among them, x is a random variable, λ>0 is the scale parameter, and k>0 is the shape parameter.

[0069] Will As the statistical failure probability of the line (pipeline), by bringing it into the load rate variation model, we can get: the model of the line (pipeline) based on aging effect and operating status,

[0070]

[0071] ② IES component importance analysis based on complex network theory and Bayesian reasoning

[0072] "IES elements" refer to the various components in the Integrated Energy System.

[0073] In complex network theory, betweenness centrality is an important indicator to measure the importance of nodes or lines. The betweenness index is defined as the ratio of the number of paths passing through a node in all the shortest paths in the network to the total number of shortest paths. In this embodiment, the betweenness index is used as one of the IES component importance analysis indicators. The branch vulnerability factor of the electric and gas systems caused by the betweenness index is defined as follows:

[0074]

[0075] V e (k), V g (p) The branch vulnerability factors n of the electricity and gas systems caused by the betweenness index areeg 、n ed are the number of energy supply equipment and loads in the power system, n gg 、n gd They are the number of energy supply equipment and loads in the natural gas system respectively; W dg They are power generation weight, power load weight, gas source weight, and gas load weight. e 、V G are the norms (Euclidean lengths) of the vulnerability vectors of the electric and gas system branches, respectively, and are used for normalization. e When the load nodes are on the shortest path, on the contrary Same reason.

[0076] The betweenness index only considers the influence of source load distribution, branch energy flow and shortest path, which mainly affects the expected power shortage EENS index, while the load loss probability LOLP index is affected by the failure rate of the component itself and the line length and impedance. The larger the failure rate of the component itself, the greater the impact on the reliability of the system LOLP index. Generally speaking, the line impedance and the failure rate and length are positively correlated.

[0077] Define Bayesian contribution:

[0078] Bayesian reasoning is used to analyze the causes of system failures. This method can be used to calculate the likelihood of each component (such as each component in the system, energy supply equipment, etc.) causing system failures. This is characterized by the vulnerability of each component to causing failures in the system. The present invention selects key components in the subsequent selection by comprehensively considering the two types of contributions and taking the top 10% of the importance ranking as key components.

[0079] ③ Establishment of the analytical model decomposition of the operation reliability of the electric-pneumatic coupling system considering key components

[0080] Loss of load probability (LOLP) indicates the probability that the system cannot supply energy to the load within a certain period of time, and expected short supply (EENS) indicates the amount of electricity that is expected to be unsatisfactory within a certain period of time. The former reflects the reliability of the system power supply, and the latter reflects the power supply guarantee capability of the system, and is widely used in reliability evaluation. These two indicators are also applicable to the reliability calculation of natural gas systems.

[0081] Traditional reliability assessment is the process of calculating the reliability index of a system or node under certain network topology, operating status, reliability parameters, etc. For example, a literature proposed a method to decompose the reliability index calculation process, that is, assuming that the system has N components, the reliability index calculation result is N components, 2 NThe result of the combination of load loss probability and expected power shortage. Obviously, calculation 2 N The result under this combination state is unrealistic. Therefore, it is necessary to select key components according to the results of importance analysis. Assume that M components are selected as key components and the remaining NM components are non-key components. Their unavailability rates are U1, U2, ..., U M , then the reliability index of the system can be U1, U2, ..., U M A nonlinear function of:

[0082] ξ LOLP,e =f1(U1,U2,…,U M )

[0083] ξ LOLP,g =f2(U1,U2,…,U M )

[0084] ξ EENS,e =f3(U1,U2,…,U M )

[0085] ξ EENS,g =f4(U1,U2,…,U M ) (7)

[0086] In the formula, ξ LOLP,e , LOLP,g are the load loss probabilities of the power system and the natural gas system, ξ EENS,e , EENS,g They are the expected supply shortage energy of the power system and the natural gas system respectively.

[0087] ξ LOLP,e As an example, the establishment process of the analytical model is explained. Assuming that all components have only two working states, available and unavailable, and the failures of important components between components to the system reliability index are independent of each other, the selected M key components have a total of L = 2 M Event F1-FL represents L combinations of states of key components, which constitute a complete event group. Combining conditional probability and total probability, the reliability index is decomposed into the superposition of failure rates caused in each state.

[0088]

[0089] In the formula is the set of load loss events when the state of M key components is j, P(F j ) represents the probability of the key component being in state j, P(E h ∣F j ) represents the probability of system load loss when the key component is in state j. P(E h ∣Fj ) can be obtained by the non-sequential MCS method, or by setting the state of key components through the GeNIe Bayesian visualization software simulation platform. The second method is selected in this embodiment for higher accuracy.

[0090] The failure rates of M key components can be accurately calculated by the aging and load rate reliability modeling established above (i.e., formula (1) and formula (5)), and then the probability of L combination states, i.e., P(F j ) value. h ∣F j ) calculation is to calculate the unavailability rate through the reliability modeling model of NM non-critical components, input it into the GeNIe Bayesian visualization simulation software platform and set the P(E) corresponding to the jth operating state output h ∣F j ) value, it is extremely difficult to obtain. From the analytical model, it can be seen that P(E h ∣F j ) value depends only on the unavailability rate of non-critical components. Relevant research has revealed that about 10% of the key components in the system contribute up to 80% to 95% of the reliability index, while the remaining non-critical components have a relatively small impact on the system reliability. Therefore, through data-driven, the P(E) of non-critical components under different operating conditions is obtained. h ∣F j )value.

[0091] Similarly, the expected short supply energy (EENS) analytical model calculation method can be obtained.

[0092]

[0093] In the formula, P(E h |F1)P(F1) is ξ LOLP , H e (s) is the load loss of the power system in state s.

[0094] (3) Reliability evaluation of the electric-pneumatic coupling system with analog-digital hybrid drive

[0095] Let K j =P(E h ∣F j ), E j =H e (s)P(E h ∣F j ), K j 、E jIt is called the analytical model coefficient, and its value changes with the load factor level and operating years. However, it is difficult to express the value of the analytical model coefficient in terms of load factor level and operating years. In this embodiment, the unavailability rate of non-critical components is obtained through the operating reliability model, and its value is input into the GENIE software to obtain some analytical model coefficients. The mapping relationship between the load factor level and operating years and the analytical model is obtained through the LSSVM regression model, so as to realize the rapid update of the system analytical model coefficient when the load changes, and then realize the rapid update of the system operation reliability index.

[0096] ①GeNIe software obtains analytical model coefficients and reliability indicators

[0097] In this embodiment, the analytical model coefficients are obtained through GENIE software (Bayesian network model) as data input for the LSSVM regression model. The specific implementation steps are as follows:

[0098] 1) First, select key components based on their contribution. The number of key components selected affects the accuracy of the system operation reliability index and the modeling cost. Suppose M key components are selected and the remaining NM components are non-key components, then there are L = 2 M A combination state.

[0099] 2) Calculate the probability of each combination state. In order to avoid missing the combination state, use M binary numbers to represent it. '0' is the component failure state, '1' is the component normal state. When M = 6, the combination state is 000000, 000001, ..., 111110, 111111. Add 1 to the decimal number represented to get the number of each combination state. There are 64 combinations in total. Determine the number at each binary position. When it is '0', take the unavailability rate, and when it is '1', take the availability rate. Multiply them to get the probability of each combination state, that is, P(F) of the analytical model. j )value.

[0100] 3) According to the changes in operating years and load factors, the unavailability rate of the corresponding non-critical components is calculated and input into the corresponding components of the GENIE software.

[0101] 4) Set the state in step 2) in the GENIE software to obtain the system failure rate, i.e., K j value, and the failure rate of each load and calculate E j value.

[0102] 5) P(F) calculated in step 2) j ) and K calculated in step 4) j and E j Substitute the analytical model to obtain the reliability index.

[0103] 6) Repeat steps 2) to 5) to complete the indicator results under all predetermined operating years and load rates as data-driven sample data.

[0104] ②Data-driven method based on LSSVM

[0105] The process of using GENIE software to obtain the reliability index of the analytical model introduced in the previous section is difficult to obtain sample data. Therefore, the algorithm used for data-driven must have a good training effect on small samples. LSSVM is an improved version of the support vector machine (SVM). It uses the least squares method to solve the optimization problem of SVM, thereby simplifying the optimization process of SVM and reducing the computational complexity. LSSVM has excellent generalization ability and can achieve good prediction performance on samples outside the training set. It performs well in application scenarios where the sample acquisition cost is high and good generalization performance is required.

[0106] Assume that the training sample set {(x i ,y i ),i=1,2,...,N s}, x i is the input data, y i is the output sample, N s is the total number of samples. LSSVM uses the principle of structural risk minimization of the sample set to establish a regression model:

[0107]

[0108] Where: J(.) is the loss function; e error i is the training error; It is a nonlinear mapping function that maps the input data from the original space to the high-dimensional feature space; ω is the weight; b is the bias; γ is the regularization parameter.

[0109] Therefore, the LSSVM regression model output is:

[0110]

[0111] Where: α i is the Lagrange multiplier; α i and b are the model parameters of LSSVM; k(x i ,x j ) is the RBF kernel function; σ is the kernel function parameter; m is the number of samples in the training set; i and j refer to two different sample indexes in the input data.

[0112] In the LSSVM (least squares support vector machine) model, the selection of regularization parameters and kernel function parameters is crucial, and they significantly affect the accuracy of model prediction. Improper parameter selection easily leads to a decrease in model performance, which is manifested as underfitting or overfitting. The present embodiment uses the LS-SVM lab toolbox on the MATLAB platform and its built-in parameter-finding function to effectively determine these two key parameters to ensure that the model achieves the optimal prediction effect. The simplex method is used to find the optimal parameters, and cross-validation is used to evaluate the model.

[0113] (4) Graphics-Analog-Digital Hybrid Drive Implementation Process

[0114] In order to simultaneously achieve the accuracy and speed of the reliability assessment of the operation of the electric-gas regional integrated energy system, the present invention proposes a method based on graph-analog-digital hybrid drive, combining the Yes network graph, analytical method, and LLSVM algorithm, and adopting a mode combining "offline modeling and online evaluation".

[0115] In the offline modeling stage;

[0116] 1) By analyzing the network topology, energy flow and logical relationship of components, the logical dependency relationship between components is deeply analyzed, and then an accurate Bayesian network model diagram is constructed. The component reliability model is established and the different failure rate information of non-critical components is obtained by changing different years and load rates as the input data of the Bayesian network node. At the same time, the analytical model coefficient is output, and the reliability index is obtained by combining the probability of the combination state of key components.

[0117] 2) The different failure rate information of non-critical components in step (2) is used as the LLSVM training data input, and the analytical model coefficients are used as the LLSVM training data output to train the LLSVM model.

[0118] Online assessment session:

[0119] 3) Long-term and short-term prediction results are combined with the reliability model to obtain the reliability index of the trained LLSVM model with different failure rates of non-critical components. Its architecture is as follows Figure 3 shown.

[0120] In combination with specific examples, the specific steps of the present invention are as follows:

[0121] (1) Parameter selection

[0122] In order to verify the effectiveness of the proposed model, an example is set up based on the IES demonstration project case in a certain area in the east and the literature Yang Chenxi, Gao Li'ai, Tang Wei. Reliability evaluation of gas-electric coupling system based on Bayesian network timing simulation[J]. Electric-Pneumatic Transmission, 2021, 51(11). The topological diagram of the test example is shown in the figure. Figure 4As shown in the figure, after taking the energy supply as the key component, the area is divided according to the topology structure and transformer, T1-T7, T2-T6-T5, T3-T4, and the gas distribution system. The reliability of this area is not affected by the lines in other areas (K j and E j The value is not affected by other areas). To verify the accuracy of the proposed hybrid drive model, area 1 is selected for analysis. The energy supply equipment includes 3 traditional units, 2 gas units, 1 power-to-gas (P2G) equipment, and 1 gas source. The calorific value of natural gas is 10 (kW·h) / m 3 The system reliability index is set as follows: the failure rate and repair time of power lines and transformers are 0.065 times / (year.km) and 5h, 0.065 times / year and 200h.

[0123] (2) Method error analysis

[0124] 1) Analysis of reliability assessment results based on analytical model decomposition

[0125] By analyzing the model decomposition method, the separate influence of key components and non-key components is realized. For region 1, the first 6 components are selected as key components, namely G1, G2, G_S1, L64, L62, and L1, and the remaining 26 components are selected as non-key components. The initial failure rate of key components is used to calculate P(F j )(j=1,2,...,L) value.

[0126] By inputting the unavailability rate information corresponding to different years and different load rate information into the GENIE software, the fault data of the system and each load can be obtained, and the K under each combination state can be obtained by setting the state of the component. j 、E L value, and calculate P(F j ) value into the analytical model, the reliability assessment result based on the analytical model decomposition can be obtained as follows: Figure 5 As shown, Figure 5 Only part of the data is shown: subscript 1 is the analytical model decomposition value, and subscript 2 is the analytical method value (i.e., Bayesian network exact reasoning).

[0127] By comparing with the evaluation results of the analytical method, the accuracy of the proposed analytical model can be verified. The results obtained by the analytical method are regarded as the true value of the reliability index and as the basis for error calculation. The calculation formula of the relative error is defined as:

[0128]

[0129] δ represents the relative error, which measures the difference between the model prediction result and the true value. AN represents the model prediction value, XAM Represents the true value.

[0130] From the 12 selected data samples, the maximum error of the LOLP index in this area is δ LOLP (0,0.3)=0.3876%, the maximum error of EENS index δ EENS (0,0.3)=2.904%, the LOLP mean error is 0.0655%, and the EENS mean error is 1.586%, which verifies that the analytical model decomposition method proposed in this paper has high accuracy.

[0131] 2) Error analysis of reliability evaluation results of graphic-analog-digital hybrid drive

[0132] This method obtains the reliability model by changing the operating years and load rate, and then obtains the failure data of each component as a data sample, which is input into the GENIE software to obtain the analytical model coefficient. The non-critical component failure information is used as input, and the analytical model coefficient is used as output. After obtaining the LLSVM training model, the mapping relationship between the operating years and load rate and the reliability index can be obtained, and finally the operating reliability evaluation results under different operating years and load rates are obtained. The operating years are selected as 20, 25, 30 and 35, and the error analysis of this method, single data drive and analytical method is compared. The output results are as follows Figure 6a , Figure 6b As shown,

[0133] The maximum errors of LOLP and EENS of this method are 0.04665% and 3.2014% respectively, while the maximum errors of LOLP and EENS driven by data are 0.5225% and 6.6368% respectively. The mean errors of LOLP and EENS of this method are 0.00592% and 1.91112% respectively, while the mean errors of LOLP and EENS driven by data are 0.31746% and 3.56389% respectively. Overall, the method proposed in this invention performs better than the data-driven method in two key indicators, and both the maximum error and the mean error show higher prediction accuracy and reliability. This is because this method decomposes the analytical model into the part affected by key components and the part affected by non-key components. The data-driven method of this method only uses LSSVM prediction for the part affected by non-key components, while the unavailability rate of key components uses the real value. It further verifies that key components have a greater impact on system reliability indicators.

[0134] (3) Operation reliability assessment results

[0135] 1) Operation reliability evaluation results considering load uncertainty and aging effects

[0136] The IES operation reliability evaluation method proposed in this method takes hours as the time scale. Taking the next 24 hours as an example, considering the load uncertainty, the four regions in a certain day with an operation life of 25 years are obtained. Without considering coupling, that is, the gas turbine failure rate is the initial failure rate. The system operation reliability evaluation results are as follows Figure 7a , Figure 7b shown.

[0137] Line aging has a great impact on system operation reliability. Considering the load rate of 0.5 in the next 50 years, the reliability evaluation results of area 1 are as follows: Figure 8 shown.

[0138] The gas turbines in area 1 and area 3 are powered by gas systems GL5 and GL7 respectively. After obtaining the reliability evaluation results of GL5 and GL7, the reliability evaluation results considering coupling can be realized. Fig. 9 shown.

[0139] 2) Rapid reliability assessment when failure rate changes

[0140] After replacing a new component, the change in the operating life of the component causes its unavailability to change, but its topology does not change. The graph-analog-digital hybrid drive method proposed in this method does not require remodeling. The change in the unavailability of the key component only requires recalculating the probability of each combination state. The mapping relationship between non-critical components and system failure rates has been established before. The change in the unavailability of non-critical components only requires its unavailability to be brought into the LLSVM model and the analytical model coefficients to be re-obtained, so that the evaluation results can be quickly updated. After replacing the key components L1 and L8, the rapid evaluation results of the operating reliability are as follows Fig.10a , Fig.10b shown.

[0141] An embodiment of the present invention provides an electrical coupling system operation reliability assessment method. By analyzing network topology, energy flow and component logical relationship, a logical relationship model of the system from energy supply to load point is constructed, and a Bayesian network (graph) is established based on this. Then, the component unavailability model and the reliability index analysis model are introduced into the Bayesian network to obtain the reliability index (graph-model). Finally, a mapping relationship (number) between the operating status of non-critical components and the reliability index coefficient is established through the LSSVM regression model. Combined with the key component fault information, the rapid update of the system operation reliability is realized, which belongs to the technical field of integrated energy system.

[0142] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0143] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating the operational reliability of an electrical coupling system, characterized in that: The following steps are involved: S10. Analyze the network topology, energy flow and component logic relationship of the electrical coupling system, and construct a Bayesian network model of the system from energy supply to load point; S20, based on the Bayesian network model, establish a component reliability model, and obtain different failure rate information of non-critical components by changing different years and load rates; S30, using the different failure rate information of the non-critical components as input data of the nodes of the Bayesian network model, outputting analytical model coefficients, and obtaining reliability indicators in combination with the probability of the combined state of the critical components; S40, using the non-critical component different failure rate information as LLSVM training data input, using the analytical model coefficients as LLSVM training data output, and training the LLSVM model; S50, when applying online prediction, the long-term and short-term prediction results of the components are combined with the component reliability model to obtain the corresponding non-critical components with different failure rates, which are input into the trained LLSVM model to output the reliability index.

2. The method for evaluating the operating reliability of an electrical coupling system according to claim 1, characterized in that: The step S10 comprises: Analyze the network topology of the electrical coupling system and determine the connection relationship between various components; Analyze the energy flow and determine the flow relationship of energy from the energy supply to each load point; Analyze the logical relationship between components, including the "and", "or" and "cause and effect" relationships between components; Based on the above analysis results, a logical relationship model from energy supply to load point is constructed; the logical relationship module describes the interdependence between various components of the system and the impact on the system operation status; The logical relationship model is converted into a Bayesian network.

3. The method for evaluating the operational reliability of an electrical coupling system according to claim 1, characterized in that: The step S20 comprises: S201. Based on the betweenness index and Bayesian contribution index, select the key components that have a greater impact on system reliability; S202, decomposing the reliability index into a critical component influence part and a non-critical component influence part; S203. Considering the aging effect and the load factor, the reliability model of the key components and the reliability model of the non-key components are established respectively, and the unavailability rate data of the key components and the non-key components under different operating conditions are obtained accordingly.

4. The method for evaluating the operational reliability of an electrical coupling system according to claim 3, characterized in that: The step S201 includes: The branch vulnerability factor of the power and gas systems caused by the betweenness index is defined as follows: Where V e (k), V g (p) Branch vulnerability factors of the electricity and gas systems caused by the betweenness index; n eg 、n ed are the number of energy supply equipment and loads in the power system, n gg 、n gd They are the number of energy supply equipment and loads in the natural gas system respectively; W dg They are power generation weight, power load weight, gas source weight, and gas load weight respectively; V e 、V G are the norms of the vulnerability vectors of the electric and gas system branches, respectively, and are used for normalization processing; when line k is located at the generator node g e When the load nodes are on the shortest path, on the contrary When the pipeline p is located at the gas source node g g When the load nodes are on the shortest path, on the contrary Define Bayesian contribution: Use Bayesian reasoning to analyze the causes of system failures and calculate the probability of each component causing system failures; characterize the vulnerability of each component in causing failures in the system; According to the branch vulnerability factor and the Bayesian contribution, M key components having a greater impact on system reliability are selected from the N components.

5. The method for evaluating the operational reliability of an electrical coupling system according to claim 4, characterized in that: The step S202 includes: Determine M components as critical components, and the remaining NM components as non-critical components, whose unavailability rates are U1, U2, ..., U M , then the reliability index of the system is U1, U2, ..., U M A nonlinear function of: x LOLP,e =f1(U1,U2,…,U M ) ξ LOLP,g =f2(U1,U2,…,U M ) ξ EENS,e =f3(U1,U2,…,U M ) x EENS,g =f4(U1,U2,…,U M ) In the formula, ξ LOLP,e , LOLP,g are the load loss probabilities of the power system and the natural gas system, ξ EENS,e , EENS,g are the expected power shortages of the electric power system and the natural gas system respectively; Among them, the failure rates of M key components are analyzed to obtain the key component impact part; The remaining NM non-critical factors have an impact on system reliability. The relationship between their operating status and reliability indicators is established through a data-driven approach to obtain the impact of non-critical components.

6. The method for evaluating the operational reliability of an electrical coupling system according to claim 5, characterized in that: The step S203 includes: Considering the aging effect and load factor, the unavailability model of energy supply components is constructed: In the formula, is the basic unavailability rate of the energy supply equipment; and They are the maximum unavailability rate under low load and the maximum unavailability rate under heavy load respectively; and are the failure rate and load rate of the energy supply equipment at time t, and are the light load and heavy load critical values ​​respectively. When the load rate is between the two, the failure rate is the basic unavailability rate; Constructing a transmission element unavailability model: Where: is the statistical failure probability of the line or pipeline; F l 、F l rated and F l max They are line power or pipeline pressure, rated power or rated pressure, transmission limit power or transmission limit pressure respectively; As a statistical failure probability of a line or pipeline; The failure rates of M key components are calculated by using the above energy supply component unavailability model and transmission component unavailability model, and then 2 M The probability value of the combination state is taken as the unavailability rate of the key component; NM non-critical components, and obtain the non-critical component unavailability data of non-critical components under different operating conditions through data driving.

Images