Reliability evaluation method, system and equipment for multi-energy-storage frequency modulation system and medium

Abstract

The invention discloses a reliability evaluation method and system for a multi-energy-storage frequency modulation system, and the method comprises the steps: combining obtained time with a thermal power output sequence, and obtaining a first sequence of a system operation state; constructing a mapping relation between a system operation state and a load loss amount, and calculating a reliability index of system frequency modulation to quantify an influence of a system frequency deviation on system reliability; the first sequence is input into a SeTformer model, a nonlinear mapping relation from a system state to a load loss amount is captured, calculation of the load loss amount is accelerated in combination with optimal transmission, and the precision and efficiency of an evaluation result can be effectively improved; and performing stability optimization on regularization parameters of the SeTform model by improving an aurora optimization algorithm so as to improve generalization capability and robustness of the SeTform model, and outputting an optimized load loss prediction value and a reliability index.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power frequency modulation, and in particular to a reliability evaluation method and system for a multi-energy storage frequency modulation system. BACKGROUND

[0002] The multi-energy storage frequency modulation system is a frequency modulation method gradually emerging in the power system in recent years, and its main purpose is to improve the frequency stability of the power system. However, when the thermal power unit is combined with the multi-energy storage device for frequency modulation, the system operating state is complex and changeable, and factors such as device failure and load fluctuation will have a significant impact on the system reliability.

[0003] However, the existing research on the reliability evaluation method of the multi-energy storage frequency modulation system includes the analytical method and the general generating function method. Although these methods can accurately describe the influence of load fluctuation and other factors on the long-term reliability of the power system, they cannot quantify the influence of system frequency deviation on the system reliability, and it is difficult to meet the actual demand for system operation reliability evaluation. At the same time, due to the large amount of power grid data, the prediction and evaluation calculation power and time consumption are huge, the efficiency is low, and the cost is high. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a reliability evaluation method and system for a multi-energy storage frequency modulation system to solve the problem that the existing evaluation method cannot quantify the influence of system frequency deviation on system reliability, the accuracy is general, and the prediction and evaluation calculation power and time consumption are huge, the efficiency is low, and the cost is high.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a reliability evaluation method for a multi-energy storage frequency modulation system, comprising:

[0008] obtaining the normal operation time and the fault repair time of the power equipment;

[0009] combining the obtained time with the thermal power output sequence to obtain a first sequence of system operating states;

[0010] constructing a mapping relationship between the system operating state and the loss of load, and calculating the reliability index of the system frequency modulation to quantify the influence of the system frequency deviation on the system reliability;

[0011] inputting the first sequence into the SeTformer model to capture the nonlinear mapping relationship between the system state and the loss of load, and combining the optimal transmission to accelerate the calculation of the loss of load to meet the accuracy requirement of the reliability index;

[0012] The regularization parameters of the SeTformer model are optimized in stability by improving the aurora optimization algorithm, and an optimized load loss prediction value and a reliability index are output.

[0013] As a preferred scheme of the reliability evaluation method of the multi-energy storage frequency modulation system, the obtained time is combined with the thermal power output sequence to obtain a first sequence of system operating states, including:

[0014] The average equipment normal operation time according to the equipment failure rate is extracted by the sequential Monte Carlo simulation method, and the average equipment repair time according to the repair rate is extracted after the failure occurs;

[0015] The thermal power output sequence is calculated, including:

[0016] The unbalanced power and frequency deviation factor of the system are calculated;

[0017] The upper and lower limits of the regulating power of the AGC unit are calculated;

[0018] Based on the state duration sequence and the thermal power output sequence, a first sequence N of system operating states is generated.

[0019] As a preferred scheme of the reliability evaluation method of the multi-energy storage frequency modulation system, the mapping relationship between the system operating state and the load loss is constructed, and the reliability index of the system frequency modulation is calculated, including:

[0020] The state evaluation model with the minimum system load shedding amount as the objective function is constructed to represent the mapping relationship;

[0021] The constraint conditions of the state evaluation model include power balance constraint, load shedding constraint and energy storage constraint;

[0022] The system load loss probability and the power shortage expectation are introduced as the system power supply reliability index, and are calculated in combination with the constraint parameters.

[0023] As a preferred scheme of the reliability evaluation method of the multi-energy storage frequency modulation system, the first sequence is input into the SeTformer model to capture the nonlinear mapping relationship between the system state and the load loss, including:

[0024] The n elements in the first sequence N are input into the SeTformer model and embedded into the RKHS, and the inner product of the kernel function value corresponding to the mapping of two points in the high-dimensional feature space is represented by the positive definite kernel function value k(x,x'), that is:

[0025] k(x,x') = <u(x),u(x')>F

[0026] Wherein, x, x' are any two elements in the first sequence N; u(x) is a mapping function from the original feature space X to the high-dimensional feature space F.

[0027] As a preferred scheme of the reliability evaluation method of the multi-energy frequency modulation system, wherein: the calculation of the loss of load is accelerated by combining the optimal transport algorithm, including:

[0028] The optimal transport algorithm is used to calculate the minimum cost function, and the input features and the reference features in the ideal operating state are aligned, expressed as:

[0029]

[0030] Wherein, C ij is the cost matrix; T ij is the transport plan matrix from position i to position j with the minimum cost; H(T) is the entropy of the transport plan; is the regularization parameter; g, h are the distributions related to the input features x and the reference features y in the ideal operating state, respectively;

[0031] The input feature vector x and y are mapped to RKHS, the alignment score between the input features x and the reference features y is calculated, and the optimal transport plan T is obtained by minimizing the transport cost;

[0032] The input features x are weighted and aggregated into m clusters to obtain the alignment matrix A:

[0033] A=TU

[0034] Wherein, U represents the mapping of the input features in RKHS.

[0035] The input features x are weighted and aggregated with the reference features y through the alignment matrix A, expressed as:

[0036] Ay(x)=m -1 / 2 A·u(x)

[0037] Wherein, A is the alignment matrix; u(x) is a mapping function from the original feature space X to the high-dimensional feature space F; m represents the features in y.

[0038] Each input feature x i is replaced by its corresponding embedding v(x i ):

[0039]

[0040] Wherein, T(v(x),y) is the optimal transport plan for aligning the input features and the reference features; v(x) is a finite-dimensional representation of the mapping function from the original feature space X to the high-dimensional feature space F;

[0041] To integrate the position information into the SeTformer model, an exponential penalty is applied based on the similarity of the position distance between the input set and the reference set, and t(v(x),y) is multiplied with the distance matrix M;

[0042] Wherein, the similarity matrix M is expressed as:

[0043]

[0044] Wherein, α, β are normalized positions in the input and reference set respectively; σ is a smoothing parameter, M ij is the similarity measure between the i-th position in the input sequence and the j-th position in the reference sequence.

[0045] The beneficial effects of the preferred technical solution are: through the optimal transmission theory to optimize the input features and reference features matching of SeTformer model, it can ensure that the model output conforms to the dynamic law of power system, and the loss of load automatically satisfies the energy conservation and frequency stability constraints.

[0046] As a preferred scheme of the reliability evaluation method of the multi-energy storage frequency modulation system, wherein: the stability of the regularization parameter of the SeTformer model is optimized by improving the aurora optimization algorithm, including:

[0047] Initialize the population, for each regularization parameter candidate value in the SeTformer model, construct a polarization cycle to obtain the charged particle value changing with time;

[0048] The solution space is searched by aurora elliptical path, and the calculation model is obtained by combining the local search and global search strategies, which is expressed as:

[0049] X new (i,j)=X(i,j)+r2×(v(t)W1+A0W2)

[0050]

[0051] Wherein, X new (i,j) is the position of the updated particle; X(i,j) is the current position of the energy particle; r1, r2 are random values between 0 and 1. W1 and W2 are the adaptive weights of the two strategies in the control equation respectively; v(t) is the speed of the charged particle; Levy(d) is the Levy flight path; d is the vector dimension; X avg is the centroid position of the high-energy particle detonation.

[0052] Each regularization parameter is obtained through the particle collision algorithm model to obtain the optimal regularization parameter combination.

[0053] As a preferred scheme of the reliability evaluation method of the multi-energy storage frequency modulation system, wherein: the stability optimization of the regularization parameters of the SeTformer model is performed through the improved aurora optimization algorithm, and further comprising: in the particle collision stage, the individual satisfying the preset fitness value is disturbed, and the adaptive disturbance function formula is as follows:

[0054] Delta p id = p id + eta i

[0055] Wherein, p id is the optimal position of the individual i in the tth iteration; Delta p id is the disturbance correction position of p id ; eta i is included in (-zeta d , zeta d ) is the correction value of the individual i in the tth iteration process, and the calculation formula of zeta d is as follows:

[0056]

[0057] Wherein, t is the current iteration number; is the average value of the optimal target position searched in the t-1th iteration; c i is a proportional coefficient; Delta r is the search area width; c is an adjustment constant.

[0058] The beneficial effects of the preferred technical scheme are that the PLO algorithm has stronger global search ability for non-convex and discrete parameter space, and the PLO can well couple the regularization parameters of the power system to satisfy the dynamic security constraints.

[0059] In a second aspect, the present application provides a reliability evaluation system of a multi-energy storage frequency modulation system, comprising:

[0060] The acquisition module is used to acquire the normal operation time and the fault repair time of the power equipment;

[0061] The sequence generation module is used to combine the acquired time with the thermal power output sequence to obtain a first sequence of system operation states;

[0062] The calculation module is used to construct the mapping relationship between the system operation states and the loss of load, and calculate the reliability index of system frequency modulation, so as to quantify the influence of system frequency deviation on system reliability;

[0063] The acceleration module is used to input the first sequence into the SeTformer model, capture the nonlinear mapping relationship between the system state and the loss of load, and combine the optimal transmission algorithm to accelerate the calculation of the loss of load;

[0064] An optimization module is configured to perform stability optimization on the regularization parameters of the SeTformer model by improving a polar optimization algorithm, and output an optimized load shedding amount prediction value and a reliability index.

[0065] In a third aspect, the present application provides an electronic device, comprising:

[0066] a memory and a processor;

[0067] The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, so as to implement the steps of the reliability evaluation method of the multi-energy storage frequency regulation system.

[0068] In a fourth aspect, the present application provides a computer readable storage medium storing computer executable instructions, which, when executed by a processor, implement the steps of the reliability evaluation method of the multi-energy storage frequency regulation system.

[0069] Compared with the prior art, the present application has the following beneficial effects: the present application combines sequential Monte Carlo simulation and a SeTformer model, the sequential Monte Carlo simulation method can generate a large number of system running state samples, the SeTformer model can learn the complex mapping relationship between the system state and the load shedding amount, and the combination of the two can effectively improve the accuracy and efficiency of the evaluation result; the improved PLO algorithm is used to optimize the regularization parameters of the SeTformer model, so as to improve the generalization ability and robustness of the SeTformer model; at the same time, the economic factors such as load shedding cost are considered when evaluating the system reliability index, so that the evaluation result is more comprehensive and practical. BRIEF DESCRIPTION OF DRAWINGS

[0070] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0071] Figure 1 The overall flowchart of the reliability evaluation method of the multi-energy storage frequency regulation system according to an embodiment of the present application.

[0072] Figure 2 The overall framework diagram of the reliability evaluation method of the multi-energy storage frequency regulation system according to an embodiment of the present application.

[0073] Figure 3 The SeTformer model operation flowchart in the reliability evaluation method of the multi-energy storage frequency regulation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0074] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0075] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a reliability assessment method for a multi-energy storage frequency modulation system, comprising:

[0076] S100: Obtaining the normal operation time and fault repair time of power equipment;

[0077] S200: combining the acquired time with the thermal power output sequence to obtain a first sequence of system operating states;

[0078] S300: Constructing a mapping relationship between the system operating status and the amount of load loss, and calculating the reliability index of the system frequency regulation to quantify the impact of the system frequency deviation on the system reliability;

[0079] S400: Input the first sequence into the SeTformer model to capture the nonlinear mapping relationship between the system state and the load loss amount, and combine it with the optimal transmission to accelerate the calculation of the load loss amount to meet the reliability indicator accuracy requirements;

[0080] S500: The stability of the regularization parameters of the SeTformer model is optimized by improving the Aurora optimization algorithm, and the optimized load loss prediction value and reliability index are output.

[0081] It should be noted that current reliability assessment methods, such as analytical methods and general generating function methods, are primarily based on quasi-steady-state assumptions, using probability distributions to describe the impact of slowly changing factors such as load fluctuations and equipment failures on reliability. However, system frequency deviation is a typical dynamic process, and its instantaneous fluctuations can trigger protection actions and lead to load losses. Traditional methods fail to couple frequency dynamic equations with reliability models and are unable to capture the direct impact of dynamic characteristics such as frequency deviation amplitude and rate of change on reliability.

[0082] Therefore, in view of the above specific problems, a system state evaluation model with the minimum system load shedding amount as the objective function is constructed through steps S100-S500 to quantitatively evaluate the system reliability. In view of the problem of a large increase in the number of samples required for system operation scenarios, the SeTformer model is used to capture the complex mapping relationship between the system state and the load shedding amount. The improved PLO algorithm is used to optimize the regularization parameters of the SeTformer model to improve the generalization ability and robustness of the SeTformer model.

[0083] Embodiment 2, with reference to Figures 1-3 For an embodiment of the present application, based on the above embodiment, a reliability evaluation method for a multi-energy storage frequency regulation system is provided.

[0084] In the embodiment of the present application, the normal operation time and the fault repair time of the power equipment are obtained in step S100.

[0085] In an optional embodiment, step S100 can obtain the time data by installing sensors such as vibration sensors, partial discharge monitoring, etc.; or by extracting through a work order system, operation and maintenance logs, etc.

[0086] It should be noted that when the thermal power unit is jointly frequency-regulated with multiple energy storage devices, the devices will exist in the conversion between normal and fault states, and the time information is prepared for subsequent construction of time sequence parameters.

[0087] In the embodiment of the present application, as Figure 2 In step S200, the obtained time is combined with the thermal power output sequence to obtain a first sequence of system operation states, including the following steps A1-A3:

[0088] A1: The average normal operation time of the equipment is extracted according to the equipment failure rate λ by the sequential Monte Carlo simulation method, and the average repair time of the equipment is extracted after the failure according to the repair rate μ;

[0089] Specifically, it can be represented as:

[0090]

[0091] Where, T tf is the normal operation time of the equipment; T pr is the fault repair time; λ and μ are the failure rate and repair rate of the equipment, respectively; β1 and β2 are random numbers between 0 and 1.

[0092] It should be noted that according to the extracted normal operation time T tf and the fault repair time T pr, obtain the duration sequence of each element operation state, provide data support for subsequent system operation reliability evaluation, and further more comprehensively evaluate the operation reliability level of the thermal power-multi energy storage frequency modulation system.

[0093] A2: Calculate the thermal power output sequence, including:

[0094] A2-1: Calculate the unbalanced power and frequency deviation factor of the system;

[0095] Specifically, the unbalanced power and frequency deviation factor of the system is calculated as follows:

[0096] ΔP i = ΔP s + ΔP CG

[0097]

[0098] Where, ΔP s is the unbalanced power of the system caused by the charging and discharging of the energy storage device; ΔP CG is the unbalanced power of the system caused by the outage of the thermal power unit; R i is the frequency modulation coefficient of the system; D i is the load regulation coefficient of the system.

[0099] A2-2: Calculate the upper and lower limits of the regulating power of the AGC unit;

[0100] It should be noted that when the unbalanced power of the system is between the upper and lower limits of the AGC unit regulating power, the system frequency will eventually return to the rated value; when the unbalanced power of the system exceeds the upper and lower limits of the unit regulating power, the system frequency cannot return to the normal range, and needs to be adjusted by cutting the load to make the power shortage return to the AGC adjustable range.

[0101] Specifically, the upper and lower limits of the regulating power of the AGC unit are calculated as follows:

[0102]

[0103] Where, N a is the total number of units; N f is the number of faulty units; P j,max is the maximum output of the jth normal running unit; P j is the actual output of the jth normal running unit; P j,min is the minimum output of the jth normal running unit.

[0104] A3: Based on the duration sequence of each state and the thermal power output sequence, generate the first sequence N of system operation state.

[0105] In the embodiments of the present application, the mapping relationship between the system operating state and the load shedding amount in step S300 is constructed, and the reliability index of system frequency modulation is calculated, including the following steps B1-B3:

[0106] B1: The mapping relationship is represented by constructing a state evaluation model with the minimum total system load shedding amount as the objective function;

[0107] Specifically, the system load shedding amount is calculated as follows:

[0108]

[0109] Wherein, P C is the total system load shedding amount; N D is the set of load nodes; P Ci is the load shedding amount of node i. B2: The constraint conditions of the state evaluation model include power balance constraint, load shedding constraint and energy storage constraint;

[0110] Specifically, the power balance constraint is expressed as:

[0111]

[0112] Wherein, M is the set of adjacent nodes of load node i; N D is the set of load nodes; P i is the power of the road connected to node i; P i,j is the transmission power from node i to j.

[0113] Specifically, the load shedding constraint is expressed as:

[0114] 0≤P Ci ≤P Di ,i∈N D

[0115] Wherein, P Ci is the load shedding amount of node i; P Di is the active load of node i.

[0116] Specifically, the capacity configuration constraint is expressed as:

[0117] max(∑W load )≤∑W power

[0118] Wherein, W load is the total load demand; W power is the total capacity configuration.

[0119] Specifically, for a specific operating state i, the system load shedding amount is P Ci ; when P Ci is equal to 0, the system can be reliably powered under state i; when PCi When it is greater than 0, the system cannot be reliably powered in state i, that is:

[0120]

[0121] Among them, F(P Ci ) is an indicator function that characterizes whether the operating state i can provide reliable power supply; when the value is 1, it means that reliable power supply is not possible.

[0122] It should be noted that steps B1-B2 are to quantify the impact of system frequency deviation on system reliability, and propose to build a system status assessment model and establish a mapping relationship between the system operating status and the amount of load loss, so as to reduce the operating cost of the power system and improve the economy and stability of the power system operation; by analyzing the generated system operating status one by one, it is determined whether it can guarantee the reliable supply of all power loads under the premise of meeting the constraints; based on the evaluation of all system operating states in the time series N, the reliability index of the system frequency regulation is calculated.

[0123] B3: Introducing the system load loss probability and power shortage expectation as system power supply reliability indicators, and combining them with constraint parameters for calculation;

[0124] Specifically, based on the evaluation of the operating status of all systems in sequence N, the system load loss probability and power shortage expectation are introduced as system power supply reliability indicators. The system load loss probability LOLP is the probability that the system cannot meet the load demand within a given time, and its calculation formula is as follows:

[0125]

[0126] Among them, T i is the duration of the running state i; T is the sum of the duration of all running states; F(P Ci ) is an indicator function that characterizes whether the operating state i can provide reliable power supply.

[0127] The expected power shortage (EENS) is used to describe the risk of system power outages. When the system frequency cannot return to the normal range after adjustment, load shedding is required to restore the system frequency. The expected power shortage (EENS) is calculated as follows:

[0128]

[0129] Where n is the number of system states to be considered; p i is the probability of system state i; T i is the duration of operating state i; ΔP m is the actual regulation capability of the system; ΔP i is the unbalanced power of the system.

[0130] In an alternative embodiment, the reliability index can also include a loss of load expectation (LOLE) and a loss of energy expectation (LOEE) and the like.

[0131] It should be noted that the reliability index is a supervisory signal used to evaluate the quality of the model output and guide the parameter optimization in the subsequent step, and the training target of the subsequent model is to make the predicted loss of load as close to the true value as possible. While the reliability index (such as LOLP, EENS) is a statistical quantity calculated based on the loss of load; it ensures that the model autonomously mines features related to reliability from raw state data, rather than relying on manually defined indicators as input, thus more in line with the principles of data-driven modeling. At the same time, the reliability index can indirectly constrain the learning direction of the model, for example: the acceleration process of the subsequent optimal transport algorithm needs to rely on the reliability index as the convergence standard, and when adjusting the SeTformer regularization parameter, the stability of the reliability index needs to be used as the optimization target to ensure the physical meaning and practicality of its prediction results in power system reliability evaluation.

[0132] Further, in order to make the calculation results of the reliability index meet the accuracy requirements, it is necessary to simulate the operation of the system for millions of hours and evaluate tens of thousands of system operating states; considering that with the increase of the power scale, the number of sources of uncertainty in the system increases and the degree continues to improve, the number of samples required to cover the system operating scenarios will also increase significantly, the SeTformer model is used in step S400 to accelerate the calculation of the loss of load.

[0133] In the embodiments of the present application, as Figure 3 , the first sequence is input into the SeTformer model in step S400 to capture the nonlinear mapping relationship between the system state and the loss of load, including:

[0134] C1: input the n elements in the first sequence N into the SeTformer model and embed them into the RKHS, and the kernel function value is represented by the inner product of the mapping of the two points in the high-dimensional feature space, that is:

[0135] k(x,x') = <u(x),u(x')>F

[0136] Where x and x' are any two elements in the first sequence N; u(x) is a mapping function from the original feature space X to the high-dimensional feature space F.

[0137] Specifically, u(x) can be infinite-dimensional, derived from R k to a finite-dimensional representation v(x), and the positive definite kernel function value κ(x,x') is represented as <v(x i ),v(x j '>). When κ is positive definite, for any x and x', κ(x,x') ≥ 0.

[0138] It should be noted that Setformer introduces a kernel cost function for optimal transport to maintain a non-negative attention matrix and use a nonlinear reweighting mechanism to emphasize important tokens in the input sequence, thereby better learning the complex mapping relationship between the system state and the load; at the same time, considering n elements in the input sequence N = {x1, …, xn}, in order to maintain linear calculation, the input feature vector is embedded into RKHS, where the point evaluation adopts the form of a linear function, and RKHS provides a robust function space for load prediction, making the model fault-tolerant to noise. Through the above positive definite function k, the data is mapped from the original feature space X to the high-dimensional feature space F. n} in the input sequence N = {x1, …, xn}, in order to maintain linear calculation, the input feature vector is embedded into RKHS, where the point evaluation adopts the form of a linear function, and RKHS provides a robust function space for load prediction, making the model fault-tolerant to noise. Through the above positive definite function k, the data is mapped from the original feature space X to the high-dimensional feature space F.

[0139] In the embodiments of the present application, the optimal transport algorithm is combined in step S400 to accelerate the calculation of the load, including the following steps:

[0140] C2: Calculate the minimum cost function by optimal transport (OT) to align (i.e., best match) between the input features and the reference features under the ideal operating state, denoted as:

[0141]

[0142] where C is the cost matrix; T is the transport plan matrix from position i to position j with the minimum cost; H(T) is the entropy of the transport plan; ij is a regularization parameter; g, h are the distributions related to the input features x and the reference features y under the ideal operating state, respectively. ij is a regularization parameter; g, h are the distributions related to the input features x and the reference features y under the ideal operating state, respectively.

[0143] C3: Map the input feature vector x and y to RKHS, and calculate the alignment score (OT kant ) between the input feature x and the reference feature y by the above optimal transport (OT), and obtain the optimal transport plan T by minimizing the transport cost.

[0144] The input feature x is weighted and aggregated into m clusters to obtain the alignment matrix A:

[0145] A = TU

[0146] where U represents the mapping of the input feature in RKHS.

[0147] It should be noted that the reference feature y is used to realize efficient element aggregation, and each element in the reference set is regarded as an "alignment unit", and the input feature x is aggregated by weighted summation on these units. The weight is the corresponding relationship between the input and the reference, which is calculated by OT, and is specifically:

[0148] ​The input feature x is weighted and aggregated with the reference feature y through the alignment matrix A, denoted as:

[0149] Ay(x)=m -1 / 2 A·u(x)

[0150] Wherein, A is an alignment matrix; u(x) is a mapping function from the original feature space X to the high-dimensional feature space F; m represents the features in y;

[0151] For infinite-dimensional or high-dimensional u(x), the transmission plan is approximated by sampling rows and columns and mapping the input from the feature space F to the linear subspace F1. To achieve this approximation, each input feature x i is replaced by its corresponding embedding v(x i ):

[0152]

[0153] Wherein, T(v(x),y) is the optimal transmission plan for aligning the input feature and the reference feature; v(x) is a finite-dimensional representation of the mapping function from the original feature space X to the high-dimensional feature space F.

[0154] C4: To integrate location information into the SeTformer model, an exponential penalty is applied based on the similarity of location distance between the input set and the reference set, and T(v(x),y) is multiplied by the distance matrix M;

[0155] Wherein, the similarity matrix M is denoted as:

[0156]

[0157] Wherein, α, β are the normalized positions in the input and reference sets respectively; σ is a smoothing parameter, M ij is the similarity measure between the i-th position in the input sequence and the j-th position in the reference sequence.

[0158] It should be noted that by optimizing the input feature and reference feature of the SeTformer model through optimal transport theory, the model output can conform to the dynamic rules of the power system, the loss of load automatically satisfies the energy conservation and frequency stability constraints, and errors such as "generation greater than load" that violate physical laws are avoided. At the same time, OT can alleviate data sparsity and noise interference through implicit high-dimensional mapping (RKHS).

[0159] Further, the core goal of the SeTformer model is to capture the complex mapping relationship between the system state and the load loss, thereby accelerating the load loss calculation, therefore, the S500 of the present application proposes to optimize the regularization parameters of the SeTformer model by using the aurora optimization algorithm (PLO), so as to improve the generalization ability and robustness of the model.

[0160] In the embodiments of the present application, the stability optimization of the regularization parameters of the SeTformer model in step S500 is improved by the aurora optimization algorithm, including:

[0161] D1: initialize the population, which can be specifically represented as:

[0162]

[0163] Wherein, N represents the size of the candidate solution contained in the population; D represents the scalable dimension of the solution space; LB and UB represent the upper and lower boundaries of the solution space respectively; R is a random number between 0 and 1.

[0164] D2: for each regularization parameter candidate value in the SeTformer model, a polarization cycle is constructed to obtain the time-varying value of the charged particle;

[0165] Specifically, i.e., the circular motion. Each iteration of the cycle represents an exploration of the parameter space. When these particles approach the earth, they encounter the resistance of the earth's magnetic field and radiate in various directions under its influence. In this process, the charged particles close to the earth interact with the earth's magnetic field and experience a rotating motion along the magnetic field lines, which is described by the Lorentz force, and the velocity of the charged particle is represented as:

[0166]

[0167] Wherein, m is the mass of the charged particle; B is the earth's magnetic field strength; q is the charge carried by the charged particle; V is the initial speed of the particle; t is the time of the particle rotating motion.

[0168] Considering the damping effect of the atmosphere on the particles, in order to improve the accuracy of the equation, the damping phenomenon is introduced into the equation of the change of the particle velocity with time, and a damping factor α is introduced, and the equation of the change of the charged particle with time is:

[0169]

[0170] Wherein, C is the integral constant; α is the damping factor, representing the decay rate of the particle velocity.

[0171] D3: search the solution space by aurora elliptical step, and combine the local search and global search strategies to obtain the calculation model, which is represented as:

[0172] X new (i,j) = X(i,j) + r2 x (v(t) W1 + A0 W2)

[0173]

[0174] wherein X new (i,j) is the updated position of the particle; X(i,j) is the current position of the energetic particle; r1, r2 are random values taken between [0, 1]. W1 and W2 are the adaptive weights of the two strategies in the control equation respectively; v(t) is the velocity of the charged particle; Levy(d) is the Levy flight path; d is the vector dimension; X avg is the position of the center of mass of the high-energy particle detonation.

[0175] It should be noted that, in order to effectively search the solution space, the above local search stage takes fine adjustments and small steps to improve the quality of the solution, making it closer to the optimal solution. The global search stage searches the solution space with larger steps to obtain the globally optimal solution.

[0176] D4: pass each regularization parameter through the particle collision algorithm model to obtain the optimal regularization parameter combination;

[0177] Specifically, each regularization parameter is regarded as a "particle" moving in the parameter space, and the parameter space is explored to find the optimal or approximately optimal regularization parameter combination by simulating the collision and interaction of parameter particles. In this process, as these particles enter the atmosphere and converge within the auroral oval, collisions occur more frequently, causing the shape of the aurora to change constantly. The mathematical model is as follows:

[0178] X new (i,j) = X(i,j) + (X(i,j) - X(a,j)) sin(πr3)

[0179] wherein X(a,j) represents any particle in the particle cluster; r3 is a random value between [0, 1].

[0180] It should be noted that the PLO algorithm simulates the formation process of the aurora, and its core idea is to dynamically adjust the position of the whole body, use the guide direction and the exploration direction, and find the optimal solution in the global search space.

[0181] In an alternative embodiment, the optimization of the regularization parameters of the SeTformer model in step S500 can also be performed by adaptive moment estimation (Adam), combined with momentum and adaptive learning rate, to dynamically adjust the parameter update step size.

[0182] In an alternative embodiment, the optimization of the regularization parameters of the SeTformer model in step S500 can also be performed by particle swarm optimization (PSO).

[0183] It should also be noted that the PLO algorithm is preferred in the present application because it has stronger global search capability for non-convex and discrete parameter space. The regularization parameters of SeTformer, such as Dropout rate, L1 / L2 coefficient, etc., are usually discrete or mixed type and the objective function can be highly non-convex, while PSO needs additional processing in mixed parameter space and Adam cannot be directly applied to discrete parameters depending on gradient information; at the same time, the regularization parameters of the power system need to meet dynamic security constraints, PLO can be well coupled, and other traditional algorithms need to rely on penalty functions with larger calculation.

[0184] In the embodiments of the present application, the stability optimization of the regularization parameters of the SeTformer model in step S500 by improving the aurora optimization algorithm further comprises D5:

[0185] D5: In the particle collision stage, the individuals satisfying the preset fitness value are subjected to disturbance processing, and the adaptive disturbance function formula is as follows:

[0186] Δp id = p id + η i

[0187] wherein p id is the optimal position of individual i in the tth iteration; Δp id is the disturbance correction position of p id ; η i ∈(-ζ d ,ζ d ) is the correction value of individual i in the tth iteration, and the calculation formula of ζ d is as follows:

[0188]

[0189] wherein t is the current iteration number; is the average value of the optimal target position searched in the t-1th iteration; c i is the proportional coefficient; Δr is the search region width; and c is the adjustment constant.

[0190] It should be noted that in the algorithm optimization process, after a certain number of iterations, the search region of the entire population will automatically converge to the optimal solution set. In order to speed up the convergence speed of the search region and improve the accuracy of the optimal solution, the adaptive disturbance strategy described above is used in the present application, which can significantly improve the accuracy and efficiency of system reliability evaluation.

[0191] In summary, the combination of the above steps not only improves the evaluation accuracy and efficiency, but also quantifies the impact of system frequency deviation on system reliability, and improves the economic efficiency of system operation. In addition, the method can be extended to reliability evaluation problems in other fields, and has certain universality and controllability, which can help improve the quality of power system operation, enhance system reliability, and improve economic benefits, etc.

[0192] Embodiment 3, the above is a schematic scheme of a reliability evaluation method of a multi-energy storage frequency modulation system. It should be noted that the technical scheme of the reliability evaluation system of the multi-energy storage frequency modulation system belongs to the same concept as the technical scheme of the reliability evaluation method of the multi-energy storage frequency modulation system described above. The technical scheme of the reliability evaluation system of the multi-energy storage frequency modulation system in this embodiment is not described in detail. The details can be seen from the description of the technical scheme of the reliability evaluation method of the multi-energy storage frequency modulation system.

[0193] The embodiment also provides a system for reliability evaluation of a multi-energy storage frequency modulation system, comprising:

[0194] The acquisition module is configured to acquire the normal operation time and the fault repair time of the power equipment.

[0195] The sequence generation module is configured to combine the acquired time with the thermal power output sequence to obtain a first sequence of system operation states.

[0196] The calculation module is configured to construct a mapping relationship between the system operation state and the loss of load, and calculate the reliability index of system frequency modulation, so as to quantify the impact of system frequency deviation on system reliability.

[0197] The acceleration module is configured to input the first sequence into the SeTformer model to capture the nonlinear mapping relationship between the system state and the loss of load, and accelerate the calculation of the loss of load by combining the optimal transport algorithm.

[0198] The optimization module is configured to stabilize the regularization parameters of the SeTformer model by improving the aurora optimization algorithm, and output the optimized loss of load prediction value and the reliability index.

[0199] The embodiment also provides an electronic device suitable for the reliability evaluation of a multi-energy storage frequency modulation system, comprising a memory and a processor. The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the reliability evaluation method of the multi-energy storage frequency modulation system proposed in the above embodiment.

[0200] The embodiment also provides a storage medium having a computer program stored thereon. The program is executed by a processor to implement the reliability evaluation method of the multi-energy storage frequency modulation system proposed in the above embodiment.

[0201] The storage medium proposed in the embodiment belongs to the same inventive concept as the reliability evaluation method for implementing the multi-energy frequency modulation system proposed in the above embodiment. The technical details not described in the embodiment can be seen from the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0202] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH, a hard disk or an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method of various embodiments of the present application.

[0203] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A reliability assessment method for a multi-energy storage frequency modulation system, characterized in that: include: Obtain the normal operation time and fault repair time of power equipment; The acquired time is combined with the thermal power output sequence to obtain the first sequence of system operation status; Construct a mapping relationship between the system operating status and the amount of load loss, and calculate the reliability index of the system frequency regulation to quantify the impact of system frequency deviation on system reliability; The first sequence is input into the SeTformer model to capture the nonlinear mapping relationship between the system state and the load loss amount, and the optimal transmission is combined to accelerate the calculation of the load loss amount to meet the reliability indicator accuracy requirements; The stability of the regularization parameters of the SeTformer model is optimized by improving the Aurora optimization algorithm, and the optimized load loss prediction value and reliability index are output.

2. The reliability assessment method of a multi-energy storage frequency modulation system according to claim 1, characterized in that: The acquired time is combined with the thermal power output sequence to obtain the first sequence of system operation status, including: The average equipment normal operation time is sampled according to the equipment failure rate through the Sequential Monte Carlo simulation method, and the average equipment repair time is sampled according to the repair rate after the failure occurs; Calculate thermal power output sequence, including: Calculate the system's unbalanced power and frequency deviation factors; Calculate the upper and lower limits of the AGC unit's regulating power; Based on the duration sequence of each state and the thermal power output sequence, a first sequence N of system operating states is generated.

3. The reliability assessment method of a multi-energy storage frequency modulation system according to claim 1 or 2, characterized in that: Construct a mapping relationship between the system operating status and the amount of load loss, and calculate the reliability index of the system frequency regulation, including: The mapping relationship is expressed by constructing a state evaluation model with the minimum load shedding of the entire system as the objective function; The constraints of the state assessment model include power balance constraints, load shedding constraints and energy storage constraints; The system load loss probability and power shortage expectation are introduced as system power supply reliability indicators, and are calculated in combination with constraint parameters.

4. The reliability assessment method of a multi-energy storage frequency modulation system according to claim 3, characterized in that: The first sequence is input into the Setformer model to capture the nonlinear mapping relationship between the system state and the load loss, including: Input n elements in the first sequence N into the SeTformer model and embed them into RKHS. The positive definite kernel function value k(x, x′) is used to represent the inner product of the kernel function value corresponding to the mapping of two points in the high-dimensional feature space, that is: k(x,x′)=<u(x),u(x′)> F Wherein, x and x′ are any two elements in the first sequence N; u(x) is the mapping function from the original feature space X to the high-dimensional feature space F.

5. The reliability assessment method of a multi-energy storage frequency modulation system according to claim 4, characterized in that: Combined with the optimal transmission algorithm to accelerate the calculation of lost load, including: The minimum cost function is calculated by the optimal transmission algorithm to align the input features with the reference features under the ideal operating state, which is expressed as: Among them, C ij is the cost matrix; T ij is the transmission plan matrix from location i to location j with minimum cost; H(T) is the entropy of the transmission plan; is the regularization parameter; g and h are the distributions related to the input feature x and the reference feature y under the ideal operating state, respectively; Map the input feature vectors x and y to RKHS, calculate the alignment score between the input feature x and the reference feature y, and obtain the optimal transmission plan T by minimizing the transmission cost; The input features x are weighted and aggregated into m clusters to obtain the alignment matrix a: A=TU Among them, U represents the mapping of input features in RKHS. The input feature x is weightedly aggregated with the reference feature y through the alignment matrix A, expressed as: Ay(x)=m -1 / 2 A·u(x) Where A is the alignment matrix; u(x) is the mapping function from the original feature space X to the high-dimensional feature space F; m represents the feature in y; For each input feature x i Replace it with its corresponding embedding v(x i ): Where T(v(x),y) is the optimal transmission plan for aligning input features and reference features; v(x) is the finite-dimensional representation of the mapping function from the original feature space X to the high-dimensional feature space F; To incorporate position information into the SeTformer model, an exponential penalty is applied to the similarity between the input set and the reference set based on the position distance, and T(v(x),y) is multiplied by the distance matrix M; Among them, the similarity matrix M is expressed as: Among them, α and β are the normalized positions in the input and reference sets respectively; σ is the smoothing parameter, and M ij is the similarity measure between the i-th position in the input sequence and the j-th position in the reference sequence.

6. The reliability assessment method of a multi-energy storage frequency modulation system according to claim 5, characterized in that: The stability of the regularization parameters of the SeTformer model is optimized by improving the Aurora optimization algorithm, including: Initialize the population and construct a polarization cycle for each candidate value of the regularization parameter in the Setformer model to obtain the time-varying value of the charged particles; By searching the solution space through the aurora ellipse trail, the calculation model is obtained by combining the local search and global search strategies, which is expressed as: X new (i,j)=X(i,j)+r2×(v(t)W1+A0W2) Among them, X new (i, j) is the updated particle position; X(i, j) is the current position of the energy particle; r1 and r2 are random values ​​between [0, 1]. W1 and W2 are the adaptive weights of the two strategies in the control equation; v(t) is the velocity of the charged particle; Levy(d) is the Levy flight path; d is the vector dimension; X avg is the center of mass of the high-energy particle detonation. Each regularization parameter is passed through the particle collision algorithm model to obtain the optimal regularization parameter combination.

7. The reliability assessment method of a multi-energy storage frequency modulation system according to claim 6, characterized in that: The stability of the regularization parameters of the SeTformer model is optimized by improving the Aurora optimization algorithm, and the method also includes: perturbation processing of individuals that meet the preset fitness value during the particle collision stage. The adaptive perturbation function formula is as follows: Δp id =p id +n i Among them, p id is the optimal position of individual i in the tth iteration; Δp id For p id The disturbance correction position of η i ∈(-ζ d ,ζ d ) is the correction value of individual i in the t-th iteration process, ζ d The calculation formula is: Where t is the current iteration number; is the average value of the optimal target position searched within t-1 iterations; c i is the proportional coefficient; Δr is the width of the search area; c is the adjustment constant.

8. A system for reliability assessment of a multi-energy storage frequency modulation system, applied to the method according to any one of claims 1 to 7, characterized in that: include: An acquisition module is used to obtain the normal operation time and fault repair time of power equipment; A sequence generation module is used to combine the acquired time with the thermal power output sequence to obtain a first sequence of system operation status; The calculation module is used to construct a mapping relationship between the system operating status and the load loss amount, and calculate the reliability index of the system frequency regulation to quantify the impact of the system frequency deviation on the system reliability; An acceleration module, configured to input the first sequence into a SeTformer model, capture the nonlinear mapping relationship between the system state and the load loss amount, and accelerate the calculation of the load loss amount in combination with an optimal transmission algorithm; The optimization module is used to optimize the stability of the regularization parameters of the SeTformer model by improving the Aurora optimization algorithm, and output the optimized load loss prediction value and reliability index.

9. An electronic device comprising: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the reliability assessment method of the multi-energy storage frequency regulation system according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the reliability assessment method of a multi-energy storage frequency modulation system according to any one of claims 1 to 7.

Images