Automatic model evaluation method and system of power system and storage medium

Abstract

The invention relates to the technical field of power system reliability evaluation, in particular to an automatic model evaluation method and system of a power system and a storage medium. Comprising the following steps: firstly, reflecting the influence of a fault state on the reliability of a power system by constructing a fault probability influence index; then, generating a power system fault state set by using a hybrid dynamic simulation method in combination with the fault probability influence index; sorting the fault states in the power system fault state set according to the fault probability influence index, and selecting the fault state having great influence on the reliability of the power system as a main fault state; thirdly, constructing an optimal load reduction model to calculate an optimal load reduction quantity in a main fault state; and finally, according to the optimal load reduction model and the system operation data in the main fault state, generating an optimal load reduction amount, and according to the optimal load reduction amount, calculating a reliability index of the power system. According to the method, rapid and accurate reliability evaluation can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of reliability assessment of power systems, and in particular to an automated model assessment method, system and storage medium for power systems. Background Art

[0002] With the rapid development of social economy and the continuous advancement of science and technology, the power system has undergone tremendous changes, from the initial small-scale, single-power simple system to today's large-scale, diversified, and highly interconnected complex network. Modern power systems not only include traditional thermal and hydropower generation, but also integrate renewable energy such as wind and solar energy, as well as various new energy storage technologies. At the same time, the introduction of smart grid technology has further improved the flexibility and controllability of the system, but also increased the complexity of the system. With the expansion of the scale of the power system and the complexity of its structure, accurate and efficient reliability assessment is required to ensure the stability of power supply. However, due to the exponential growth in the number of fault states and the uncertainty and high volatility of the system, the evaluation of the power system faces many challenges.

[0003] To meet these challenges, researchers have proposed a variety of solutions. For example, the Chinese patent with publication number CN113887989A provides a power system reliability assessment method, device, computer equipment and storage medium. First, a plurality of different power system state vectors are generated through the standard value of the discrete degree of load loss probability; then the cross entropy algorithm is used to iteratively calculate the system state vector to update the reliability indicator load loss probability; when the discrete degree of the latest reliability indicator load loss probability is less than the standard value of the discrete degree of load loss probability, the iteration ends; finally, the power system reliability is assessed based on the latest reliability indicator load loss probability, which can quickly and accurately obtain the reliability assessment results of the power system.

[0004] Although the existing technology has improved the efficiency of reliability assessment to a certain extent, it regards all possible fault states as equally important and fails to take into account that some fault states have a greater impact on system reliability, resulting in unreasonable allocation of computing resources. Therefore, there is room for further improvement in the efficiency of power system reliability assessment.

[0005] To this end, an automated model evaluation method, system and storage medium for a power system are proposed. Summary of the invention

[0006] The object of the present invention is to provide an automated model evaluation method, system and storage medium for an electric power system. First, a fault probability impact index is constructed to reflect the impact of a fault state on the reliability of the electric power system. Then, a hybrid dynamic simulation method is used in combination with the fault probability impact index to generate a set of fault states of the electric power system, so as to ensure a comprehensive simulation while paying more attention to the fault states that have a greater impact on the reliability of the electric power system. Further, according to the fault probability impact index, the fault states in the set of fault states of the electric power system are sorted, and the fault states that have a greater impact on the reliability of the electric power system are selected as the main fault states. Then, an optimal load reduction model is constructed to calculate the optimal load reduction amount under the main fault state. Finally, according to the optimal load reduction model and the system operation data under the main fault state, an optimal load reduction amount is generated, and the reliability index of the electric power system is calculated according to the optimal load reduction amount.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An automated model evaluation method for a power system, comprising:

[0009] Establishing a power system model, wherein the power system model includes a power generation system and a transmission system;

[0010] Acquire the topological structure data, equipment parameters and historical operation data of the power system according to the power system model; the topological structure data includes node information and transmission line information;

[0011] Construct a fault probability impact index to reflect the impact of fault conditions on power system reliability;

[0012] Using hybrid dynamic simulation method, a set of power system fault states is generated;

[0013] For each fault state in the power system fault state set, calculating the fault probability impact index, sorting the fault states in descending order according to the fault probability impact index, and selecting the first N fault states as the main fault states;

[0014] Constructing an optimal load reduction model; the optimal load reduction model is based on a stacked denoising autoencoder model;

[0015] Building a training data set based on the historical operation data of the power system to train the optimal load reduction model; the training data set includes system operation data and corresponding optimal load reduction amount;

[0016] generating the optimal load reduction amount according to the trained optimal load reduction model and the system operation data under the main fault state;

[0017] A reliability index of the power system is calculated according to the optimal load reduction amount.

[0018] Preferably, the steps for calculating the failure probability impact index are as follows:

[0019] Under each fault condition, the load flow of the power system is calculated to obtain the active power and reactive power of each transmission line in the power system;

[0020] Obtaining the power capacity limit of each transmission line according to the equipment parameters, comparing the reactive power with the power capacity limit, and calculating the number of overloaded lines;

[0021] Obtaining the number of available transmission lines and the number of power outage transmission lines in the power system under a fault state, and obtaining the line availability of each transmission line based on the historical operation data;

[0022] Calculating a probability of a fault state occurring based on the number of available transmission lines, the number of power outage transmission lines, and the line availability;

[0023] Calculating a fault state impact index according to the active power, the power capacity limit, the number of overloaded lines and the probability of occurrence of the fault state;

[0024] The fault state occurrence probability and the fault state impact index are multiplied to obtain a fault probability impact index.

[0025] Preferably, the direct current method is used to calculate the load flow of the power system, and the steps are as follows:

[0026] Establish the power balance equation and current balance equation of the node;

[0027] Substituting the current balance equation into the power balance equation and separating the real part and the imaginary part to obtain an active power equation and a reactive power equation;

[0028] Performing linear approximation operations on the active power equation and the reactive power equation respectively to obtain a simplified active power equation and a simplified reactive power equation;

[0029] According to the simplified active power equation and the simplified reactive power equation, active power matrix equations and reactive power matrix equations of all nodes are established, and numerical methods are used to solve them respectively to obtain the voltage angle and voltage amplitude of each node;

[0030] According to the voltage angle and the voltage amplitude, the active power and reactive power of each transmission line between nodes are calculated.

[0031] Preferably, the steps of generating the power system fault state set by the hybrid dynamic simulation method are as follows:

[0032] S1. Traverse all devices in the power system, create a fault state for each device, and obtain a single device fault state set;

[0033] S2. Based on the historical operation data, a fault state is created for each common device combination to obtain a combined device fault state set; the common device combination is a plurality of devices that fail at the same time;

[0034] S3. Merge the single device fault state set and the combined device fault state set to obtain a first fault state set;

[0035] S4. Obtain the mean time between failures and mean time to repair of each device based on the device parameters, and calculate the device failure probability p of each device within the time step Δt device ;

[0036] S5. Perform simulation cycle by dynamic Monte Carlo simulation method, the steps are as follows:

[0037] S51. Initialize all device states to normal and second fault state set And define the total simulation time T and the dynamic adjustment period a;

[0038] S52. For the c=1 time step, generate a random number between [0,1) for each device; if the random number is less than the device failure probability, change the device state to failure; if the random number is not less than the device failure probability, change the device state to normal; traverse all devices to obtain a random failure state;

[0039] S53. Add the random fault state to the second fault state set, c=c+1;

[0040] S54. Repeat steps S52 and S53. When c≤TΔt, and c is a multiple of a, for each fault state in the second fault state set, calculate the fault probability impact index; sort the fault states in descending order according to the fault probability impact index; for the faulty devices in the first 30% fault states, increase the device failure probability; for the faulty devices in the last 30% fault states, reduce the device failure probability; the device failure probability p device =p device μ, where μ is the adjustment factor;

[0041] S55. When c>TΔt, the simulation cycle ends and the second fault state set is obtained;

[0042] S6. Combine the first fault state set and the second fault state set to obtain the power system fault state set.

[0043] Preferably, the optimal load reduction model adopts a two-stage training strategy, including: using a denoising autoencoder to perform unsupervised pre-training layer by layer; and training the entire model through supervised learning.

[0044] Preferably, the system operation data includes: the active load demand and reactive load demand of each node and the power of each power generation node; the power generation node is a node with a generator in the power system; obtaining the system operation data of the power system under the main fault state includes: determining the active load demand and the reactive load demand of each node based on the historical data of the power system; under the main fault state, the direct current flow method is used to calculate the power generation power of the power generation node.

[0045] Preferably, the expected power shortage is used as an indicator for evaluating the reliability of the power system, and the calculation formula is as follows:

[0046]

[0047] Wherein, EENS represents the expected power shortage amount; E k represents the optimal load reduction amount when the fault state k occurs; p k represents the probability of occurrence of fault state k; t k represents the expected duration of fault state k; N represents the number of the main fault states.

[0048] Preferably, the expected duration is calculated by: for a single device failure, using the average repair time of the device as the expected duration; for simultaneous failure of multiple devices, using the device with the largest average repair time as the expected duration.

[0049] An automatic model evaluation system for a power system, comprising:

[0050] A model building module, used to build a power system model, wherein the power system model includes a power generation system and a transmission system;

[0051] A data acquisition module, used to acquire topological structure data, equipment parameters and historical operation data of the power system according to the power system model; the topological structure data includes node information and transmission line information;

[0052] Failure probability impact module, used to construct and calculate the failure probability impact index;

[0053] A fault state generation module, used for generating a set of fault states of a power system using a hybrid dynamic simulation method;

[0054] A fault state selection module is used to sort the fault states according to the fault probability impact index and select the first N fault states as the main fault states;

[0055] The optimal load reduction module is used to build an optimal load reduction model and calculate the optimal load reduction amount of the power system;

[0056] The reliability calculation module is used to calculate the reliability index of the power system according to the optimal load reduction amount.

[0057] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements an automatic model evaluation method for a power system.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. The fault probability impact index is introduced to evaluate the impact of fault conditions on the reliability of the power system. The comprehensive consideration of the probability of occurrence of the fault condition and the fault condition impact index provides a basis for improving the accuracy of power system evaluation. By calculating the fault probability impact index of each fault condition, the fault conditions are sorted to identify the main fault conditions with the greatest impact on system reliability. The most critical fault conditions can be effectively screened out, avoiding the huge computational burden brought by a comprehensive analysis of all possible fault conditions. By screening critical fault conditions, the evaluation efficiency of complex power systems is significantly improved.

[0060] 2. A hybrid dynamic simulation method is proposed to generate a set of fault states of the power system. This method combines deterministic analysis and random simulation. First, single equipment failures and common combined equipment failures are considered, and then more random fault states are generated through dynamic Monte Carlo simulation methods. In addition, a dynamic adjustment mechanism is introduced to regularly adjust the failure probability of the equipment according to the importance of the fault state, so that the simulation method can not only comprehensively cover various possible fault conditions, but also focus on the fault states that have a greater impact on the reliability of the power system, thereby improving the efficiency and accuracy of fault state generation, and thus improving the overall efficiency of power system evaluation.

[0061] 3. An optimal load reduction model based on stacked denoising autoencoders is proposed. A two-stage training strategy is adopted. First, the denoising autoencoder is used for unsupervised pre-training, and then the entire model is trained through supervised learning. The model can effectively handle the nonlinear relationships and complex dependencies in the power system, so as to quickly and accurately calculate the optimal load reduction under various fault conditions. Moreover, the model has better generalization ability, so that even when facing large-scale and complex power systems, the evaluation process can be completed quickly. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a flow chart of an automated model evaluation method for a power system provided by an embodiment of the present invention;

[0063] Figure 2 A schematic diagram of the structure of an automated model evaluation model for a power system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] 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.

[0065] Reliability assessment of power systems is an important foundation for ensuring the stability of power supply and improving system operation efficiency. With the continuous expansion of power system scale and the increasing complexity of structure, accurate and efficient reliability assessment becomes particularly important. Efficient assessment methods not only help to quickly identify weak links in the system, but also provide timely and reliable decision support for system planning, operation and maintenance, and emergency response.

[0066] The present invention provides an automated model evaluation method, system and storage medium for an electric power system, which can simulate fault states and sort and select fault states by distinguishing the importance of different fault states, thereby enabling rapid and accurate reliability evaluation. The effectiveness of the present invention will be illustrated in two embodiments.

[0067] Embodiment 1

[0068] See also Figure 1 , an automatic model evaluation method for a power system, characterized by comprising:

[0069] Establishing a power system model, wherein the power system model includes a power generation system and a transmission system;

[0070] Acquire the topological structure data, equipment parameters and historical operation data of the power system according to the power system model; the topological structure data includes node information and transmission line information;

[0071] Construct a fault probability impact index to reflect the impact of fault conditions on power system reliability;

[0072] Using hybrid dynamic simulation method, a set of power system fault states is generated;

[0073] For each fault state in the power system fault state set, calculating the fault probability impact index, sorting the fault states in descending order according to the fault probability impact index, and selecting the first N fault states as the main fault states;

[0074] Constructing an optimal load reduction model; the optimal load reduction model is based on a stacked denoising autoencoder model;

[0075] Building a training data set based on the historical operation data of the power system to train the optimal load reduction model; the training data set includes system operation data and corresponding optimal load reduction amount;

[0076] generating the optimal load reduction amount according to the trained optimal load reduction model and the system operation data under the main fault state;

[0077] A reliability index of the power system is calculated according to the optimal load reduction amount.

[0078] By constructing a fault probability impact index, the impact of fault conditions on system reliability can be accurately evaluated; the use of a hybrid dynamic simulation method to generate a set of fault conditions improves the comprehensiveness and efficiency of fault condition generation; the use of an optimal load reduction model can quickly and accurately calculate the optimal load reduction amount; overall, the efficiency and accuracy of power system reliability assessment are improved.

[0079] In the embodiment of the present application, the topological structure data includes node information and transmission line information, the node information includes: node number and node type, the node type includes power generation node, load node and connection point; the transmission line information includes: line number, starting node, ending node and line type; the equipment parameters include generator parameters, transformer parameters and transmission line parameters; the historical operation data includes equipment failure and maintenance data, load data, power flow and voltage data, power generation data and topology change records, etc.; some transmission line data are shown in Table 1.

[0080] Table 1 Example of relevant data of transmission line

[0081] Line number Starting Node Termination Node Power capacity limit / MVAr Availability 1 1 2 50 0.97 2 1 4 30 0.91 3 2 3 50 0.98 4 2 5 60 0.93 5 3 6 40 0.94 6 4 5 20 0.90

[0082] Furthermore, the steps for calculating the failure probability impact index are as follows:

[0083] Under each fault condition, the load flow of the power system is calculated to obtain the active power and reactive power of each transmission line in the power system;

[0084] Obtaining the power capacity limit of each transmission line according to the equipment parameters, comparing the reactive power with the power capacity limit, and calculating the number of overloaded lines;

[0085] Obtaining the number of available transmission lines and the number of power outage transmission lines in the power system under a fault state, and obtaining the line availability of each transmission line based on the historical operation data;

[0086] The probability of a fault state occurring is calculated according to the number of available transmission lines, the number of power outage transmission lines and the line availability, and the calculation formula is as follows:

[0087]

[0088] Among them, p k a represents the probability of occurrence of the fault state k; i and a j N represents the line availability of transmission line i and transmission line j respectively; al N represents the number of available transmission lines; ol Indicates the number of power transmission lines that are out of power;

[0089] The fault state impact index is calculated according to the active power, the power capacity limit, the number of overloaded lines and the probability of occurrence of the fault state, and the calculation formula is as follows:

[0090]

[0091] Among them, PI k The performance index representing the fault state k; N oll Represents the number of overloaded lines; ω i represents the weight coefficient of transmission line i; P i P represents the active power of transmission line i. i lim represents the capacity limit of the transmission line i; n represents the order of the fault state impact index;

[0092] The fault state occurrence probability and the fault state impact index are multiplied to obtain the fault probability impact index PIp k =PI k ·p k .

[0093] By comprehensively considering multiple factors, the fault impact assessment is more comprehensive; by calculating the probability of fault state occurrence, the assessment result is more in line with the actual situation; by combining the probability of fault occurrence and the degree of impact, a more reasonable basis for fault state ranking is provided.

[0094] Furthermore, the direct current method is used to calculate the load flow of the power system. The steps are as follows:

[0095] Establish the power balance equation and current balance equation of the node, the equations are as follows:

[0096] S l =P l +j·Q l =V l I l ;

[0097]

[0098] Among them, S l , P l and Q l denote the complex power, active power and reactive power of node l respectively; j denotes an imaginary unit, j 2 =-1; V l and V m Respectively represent the voltage amplitude of node l and node m; Y l sa represents the parallel admittance of node l; Y lm represents the line admittance connecting node l and node m; Y l sa and Y lm Obtained through device parameters;

[0099] Substituting the current balance equation into the power balance equation and separating the real part and the imaginary part to obtain an active power equation and a reactive power equation;

[0100] Performing linear approximation operations on the active power equation and the reactive power equation respectively to obtain a simplified active power equation and a simplified reactive power equation;

[0101] The simplified active power equation and the simplified reactive power equation are as follows:

[0102]

[0103]

[0104] Among them, δ l and δ m Represent the voltage angles of node l and node m respectively;

[0105] According to the simplified active power equation and the simplified reactive power equation, active power matrix equations and reactive power matrix equations of all nodes are established, and numerical methods are used to solve them respectively to obtain the voltage angle and voltage amplitude of each node;

[0106] According to the voltage angle and the voltage amplitude, the active power and reactive power of the transmission line between nodes are calculated, and the calculation formula is:

[0107] P lm =V l2 ·Y lm ·sinβ lm +V l ·V m ·Y lm ·sin(δ lm -δ lm );

[0108]

[0109] Among them, P lm and Q lm They represent the active power and reactive power of the transmission line from node l to node m respectively; δ lm and represents the voltage angle difference between node l and node m; β lm represents the admittance angle of the transmission line between node l and node m; represents the charging admittance of the transmission line between node l and node m, β lm and Obtained through device parameters.

[0110] The direct current flow method simplifies complex power system calculations and improves calculation efficiency. Through linear approximate operations, nonlinear factors are ignored, and node voltage and line power information can be quickly obtained, providing a basis for subsequent evaluation and analysis.

[0111] Furthermore, the steps of generating the power system fault state set by the hybrid dynamic simulation method are as follows:

[0112] S1. Traverse all devices in the power system, create a fault state for each device, and obtain a single device fault state set;

[0113] S2. Based on the historical operation data, a fault state is created for each common device combination to obtain a combined device fault state set; the common device combination is a plurality of devices that fail at the same time;

[0114] S3. Merge the single device fault state set and the combined device fault state set to obtain a first fault state set;

[0115] S4. Obtain the mean time between failures and mean time to repair of each device based on the device parameters, and calculate the device failure probability p of each device within the time step Δt device ;

[0116] S5. Perform simulation cycle by dynamic Monte Carlo simulation method, the steps are as follows:

[0117] S51. Initialize all device states to normal and second fault state set And define the total simulation time T and the dynamic adjustment period a;

[0118] S52. For the c=1 time step, generate a random number between [0,1) for each device; if the random number is less than the device failure probability, change the device state to failure; if the random number is not less than the device failure probability, change the device state to normal; traverse all devices to obtain a random failure state;

[0119] S53. Add the random fault state to the second fault state set, c=c+1;

[0120] S54. Repeat steps S52 and S53. When c≤TΔt, and c is a multiple of a, for each fault state in the second fault state set, calculate the fault probability impact index; sort the fault states in descending order according to the fault probability impact index; for the faulty devices in the first 30% fault states, increase the device failure probability; for the faulty devices in the last 30% fault states, reduce the device failure probability; the device failure probability p device =p device *μ, where μ is the adjustment factor;

[0121] S55. When c>TΔt, the simulation cycle ends and the second fault state set is obtained;

[0122] S6. Combine the first fault state set and the second fault state set to obtain the power system fault state set.

[0123] The hybrid dynamic simulation method combines deterministic analysis and stochastic simulation to improve the comprehensiveness of fault state generation. The dynamic adjustment mechanism is introduced to pay more attention to important fault states that have a greater impact on reliability, thereby improving evaluation efficiency. The diversity and representativeness of fault states are ensured by merging the sets of fault states generated by multiple methods.

[0124] Furthermore, the optimal load reduction model adopts a two-stage training strategy, including: using a denoising autoencoder to perform unsupervised pre-training layer by layer; and training the entire model through supervised learning.

[0125] The two-stage training strategy improves the model's ability to learn complex nonlinear relationships and enhances the model's generalization performance.

[0126] Furthermore, the system operation data includes: the active load demand and reactive load demand of each node and the power of each power generation node; the power generation node is a node with a generator in the power system; obtaining the system operation data of the power system under the main fault state includes: determining the active load demand and the reactive load demand of each node based on the historical data of the power system; under the main fault state, the direct current flow method is used to calculate the power generation power of the power generation node.

[0127] The input data of the model are based on the basic operating parameters of the power system, which are representative and relatively easy to obtain, so that the model can be adapted to different power systems.

[0128] Furthermore, the expected power shortage is used as an indicator to evaluate the reliability of the power system, and the calculation formula is as follows:

[0129]

[0130] Wherein, EENS represents the expected power shortage amount; E k represents the optimal load reduction amount when the fault state k occurs; p k represents the probability of occurrence of fault state k; t k represents the expected duration of fault state k; N represents the number of the main fault states.

[0131] It provides an intuitive reliability evaluation index to help decision making. The index takes into account the probability and duration of the fault state, making the evaluation result more comprehensive and accurate.

[0132] Furthermore, the calculation method of the expected duration includes: for a single device failure, using the average repair time of the device as the expected duration; for simultaneous failure of multiple devices, using the device with the largest average repair time as the expected duration; the average repair time is obtained based on device parameters.

[0133] A reasonable duration estimation method is provided for different types of faults; using the mean repair time as the basis improves the accuracy and operability of the estimation and provides a more accurate data basis for the calculation of reliability indicators.

[0134] Embodiment 2

[0135] See also Figure 2 , an automated model evaluation system for a power system, comprising:

[0136] A model building module, used to build a power system model, wherein the power system model includes a power generation system and a transmission system;

[0137] A data acquisition module, used to acquire topological structure data, equipment parameters and historical operation data of the power system according to the power system model; the topological structure data includes node information and transmission line information;

[0138] Failure probability impact module, used to construct and calculate the failure probability impact index;

[0139] A fault state generation module, used for generating a set of fault states of a power system using a hybrid dynamic simulation method;

[0140] A fault state selection module is used to sort the fault states according to the fault probability impact index and select the first N fault states as the main fault states;

[0141] The optimal load reduction module is used to build an optimal load reduction model and calculate the optimal load reduction amount of the power system;

[0142] The reliability calculation module is used to calculate the reliability index of the power system according to the optimal load reduction amount.

[0143] Furthermore, the steps of constructing the fault probability impact module and calculating the fault probability impact index are as follows:

[0144] Under each fault condition, the load flow of the power system is calculated to obtain the active power and reactive power of each transmission line in the power system;

[0145] According to the equipment parameters obtained by the data acquisition module, the power capacity limit of each transmission line is obtained, the reactive power is compared with the power capacity limit, and the number of overloaded lines is calculated;

[0146] Acquire the number of available transmission lines and the number of power outage transmission lines in the power system under a fault state, and acquire the line availability of each transmission line through the data acquisition module;

[0147] Calculating a probability of a fault state occurring based on the number of available transmission lines, the number of power outage transmission lines, and the line availability;

[0148] Calculating a fault state impact index according to the active power, the power capacity limit, the number of overloaded lines and the probability of occurrence of the fault state;

[0149] The fault state occurrence probability and the fault state impact index are multiplied to obtain a fault probability impact index.

[0150] Furthermore, the steps of the fault state generation module using the hybrid dynamic simulation method to generate a set of fault states of the power system are as follows:

[0151] S1. Traverse all devices in the power system, create a fault state for each device, and obtain a single device fault state set;

[0152] S2. Based on the historical operation data, a fault state is created for each common device combination to obtain a combined device fault state set; the common device combination is a plurality of devices that fail at the same time;

[0153] S3. Merge the single device fault state set and the combined device fault state set to obtain a first fault state set;

[0154] S4. Obtain the mean time between failures and mean time to repair of each device based on the device parameters, and calculate the device failure probability p of each device within the time step Δt device ;

[0155] S5. Perform simulation cycle by dynamic Monte Carlo simulation method, the steps are as follows:

[0156] S51. Initialize all device states to normal and second fault state set And define the total simulation time T and the dynamic adjustment period a;

[0157] S52. For the c=1 time step, generate a random number between [0,1) for each device; if the random number is less than the device failure probability, change the device state to failure; if the random number is not less than the device failure probability, change the device state to normal; traverse all devices to obtain a random failure state;

[0158] S53. Add the random fault state to the second fault state set, c=c+1;

[0159] S54. Repeat steps S52 and S53. When c≤TΔt, and c is a multiple of a, for each fault state in the second fault state set, calculate the fault probability impact index; sort the fault states in descending order according to the fault probability impact index; for the faulty devices in the first 30% fault states, increase the device failure probability; for the faulty devices in the last 30% fault states, reduce the device failure probability; the device failure probability p device =p device *μ, where μ is the adjustment factor;

[0160] S55. When c>TΔt, the simulation cycle ends and the second fault state set is obtained;

[0161] S6. Combine the first fault state set and the second fault state set to obtain the power system fault state set.

[0162] Furthermore, the optimal load reduction module constructs an optimal load reduction model based on the stacked denoising autoencoder model, and constructs a training data set based on the historical operation data obtained by the data acquisition module to train the optimal load reduction model; the training data set includes system operation data and the corresponding optimal load reduction amount; the optimal load reduction model adopts a two-stage training strategy, including: using a denoising autoencoder to perform unsupervised pre-training layer by layer; and training the entire model through supervised learning.

[0163] Furthermore, the reliability calculation module uses the expected power shortage as an indicator for evaluating the reliability of the power system, and the calculation formula is as follows:

[0164]

[0165] Wherein, EENS represents the expected power shortage amount; E k represents the optimal load reduction amount when the fault state k occurs; p k represents the probability of occurrence of fault state k; t k represents the expected duration of fault state k; N represents the number of the main fault states; for a single device failure, the average repair time of the device is used as the expected duration; for simultaneous failure of multiple devices, the device with the largest average repair time among the multiple devices is used as the expected duration; the average repair time is obtained based on the device parameters obtained by the data acquisition module.

[0166] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An automated model evaluation method for a power system, characterized in that: include: Establishing a power system model, wherein the power system model includes a power generation system and a power transmission system; Acquire the topological structure data, equipment parameters and historical operation data of the power system according to the power system model; the topological structure data includes node information and transmission line information; Construct failure probability impact index; Using a hybrid dynamic simulation method, combined with the fault probability impact index, to generate a set of power system fault states; For each fault state in the power system fault state set, calculating the fault probability impact index, sorting the fault states in descending order according to the fault probability impact index, and selecting the first N fault states as the main fault states; Constructing an optimal load reduction model; the optimal load reduction model is based on a stacked denoising autoencoder model; Building a training data set based on the historical operation data of the power system to train the optimal load reduction model; The training data set includes system operation data and corresponding optimal load reduction amount; Acquire the system operation data under the main fault state, and generate the optimal load reduction amount by using the trained optimal load reduction model; A reliability index of the power system is calculated according to the optimal load reduction amount.

2. The method for automatic model evaluation of a power system according to claim 1, characterized in that: The calculation steps of the failure probability impact index are as follows: Under each fault condition, the load flow of the power system is calculated to obtain the active power and reactive power of each transmission line in the power system; Obtaining the power capacity limit of each transmission line according to the equipment parameters, comparing the reactive power with the power capacity limit, and calculating the number of overloaded lines; Obtaining the number of available transmission lines and the number of power outage transmission lines in the power system under a fault state, and obtaining the line availability of each transmission line based on the historical operation data; Calculating a probability of a fault state occurring based on the number of available transmission lines, the number of power outage transmission lines, and the line availability; Calculating a fault state impact index according to the active power, the power capacity limit, the number of overloaded lines and the probability of occurrence of the fault state; The fault state occurrence probability and the fault state impact index are multiplied to obtain a fault probability impact index.

3. The method for automatic model evaluation of a power system according to claim 2, characterized in that: The direct current method is used to calculate the load flow of the power system. The steps are as follows: Establish the power balance equation and current balance equation of the node; Substituting the current balance equation into the power balance equation and separating the real part and the imaginary part to obtain an active power equation and a reactive power equation; Performing linear approximation operations on the active power equation and the reactive power equation respectively to obtain a simplified active power equation and a simplified reactive power equation; According to the simplified active power equation and the simplified reactive power equation, the active power matrix equation and the reactive power matrix equation of all nodes are established, and numerical methods are used to solve them respectively to obtain the voltage angle and voltage amplitude of each node; According to the voltage angle and the voltage amplitude, the active power and reactive power of each transmission line between nodes are calculated.

4. The method for automatic model evaluation of a power system according to claim 1, characterized in that: The steps of generating the power system fault state set by the hybrid dynamic simulation method are as follows: S1. Traverse all devices in the power system, create a fault state for each device, and obtain a single device fault state set; S2. Based on the historical operation data, a fault state is created for each common device combination to obtain a combined device fault state set; the common device combination is a plurality of devices that fail at the same time; S3. Merge the single device fault state set and the combined device fault state set to obtain a first fault state set; S4. Obtain the mean time between failures and mean time to repair of each device based on the device parameters, and calculate the device failure probability p of each device within the time step Δt device ; S5. Perform simulation cycle by dynamic Monte Carlo simulation method, the steps are as follows: S51. Initialize all device states to normal and second fault state set and the number of cycles c=1, and define the total simulation time T and the dynamic adjustment period a; S52. For the c-th time step, generate a random number between [0,1) for each device; If the random number is less than the device failure probability, changing the device state to failure; If the random number is not less than the device failure probability, change the device state to normal; traverse all devices to obtain a random failure state; S53. Add the random fault state to the second fault state set, c=c+1; S54. When c≤TΔt, repeat steps S52 and S53. If c is a multiple of a, for each fault state in the second fault state set, calculate the fault probability impact index; sort the fault states in descending order according to the fault probability impact index; for the faulty equipment in the first σ% fault state, increase the equipment failure probability; for the faulty equipment in the last σ% fault state, reduce the equipment failure probability; the equipment failure probability p device =p device μ, where μ is the adjustment factor; S55. When c>TΔt, the simulation cycle ends and the second fault state set is obtained; S6. Combine the first fault state set and the second fault state set to obtain the power system fault state set.

5. The method for automatic model evaluation of a power system according to claim 1, characterized in that: The optimal load reduction model adopts a two-stage training strategy, including: using a denoising autoencoder to perform unsupervised pre-training layer by layer; and training the entire model through supervised learning.

6. The method for automatic model evaluation of a power system according to claim 1, characterized in that: The system operation data includes: the active load demand and reactive load demand of each node and the power of each power generation node; the power generation node is a node with a generator in the power system; obtaining the system operation data of the power system under the main fault state includes: determining the active load demand and the reactive load demand of each node based on the historical data of the power system; under the main fault state, the direct current flow method is used to calculate the power of the power generation node.

7. The method for automatic model evaluation of a power system according to claim 1, characterized in that: The expected power shortage is used as an indicator to evaluate the reliability of the power system. The calculation formula is as follows: Wherein, EENS represents the expected power shortage amount; E k represents the optimal load reduction amount when the fault state k occurs; p k represents the probability of occurrence of fault state k; t k represents the expected duration of fault state k; N represents the number of the main fault states.

8. The method for automatic model evaluation of a power system according to claim 7, characterized in that: The calculation method of the expected duration includes: for a single device failure, using the average repair time of the device as the expected duration; for simultaneous failure of multiple devices, using the device with the largest average repair time as the expected duration; the average repair time is obtained based on device parameters.

9. An automated model evaluation system for a power system, characterized in that: include: A model building module, used to build a power system model, wherein the power system model includes a power generation system and a transmission system; A data acquisition module, used to acquire topological structure data, equipment parameters and historical operation data of the power system according to the power system model; the topological structure data includes node information and transmission line information; Failure probability impact module, used to construct and calculate the failure probability impact index; A fault state generation module, used for generating a set of fault states of a power system using a hybrid dynamic simulation method; A fault state selection module is used to sort the fault states according to the fault probability impact index and select the first N fault states as the main fault states; The optimal load reduction module is used to construct an optimal load reduction model and calculate the optimal load reduction amount of the power system; The reliability calculation module is used to calculate the reliability index of the power system according to the optimal load reduction amount.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, an automatic model evaluation method for a power system as described in any one of claims 1 to 8 is implemented.

Images