Measurement point collaborative selection method based on system parameter checking scene

Abstract

The invention provides a measuring point collaborative selection method based on a system parameter checking scene, relates to the technical field of power system simulation, and solves the problem that the checking effect of the system parameter checking scene in the prior art is unstable. The method comprises the steps of S1, carrying out scene and measurement point collaborative optimization selection based on the identifiability, S2, carrying out identifiability index construction and comprehensive scoring, S3, carrying out scene and measurement point collaborative optimization selection based on the identifiability, and S4, carrying out parameter checking and model updating on an optimal combination. According to the scheme, under the condition of limited calculation and test cost, the accuracy and engineering applicability of the parameter checking result of the electric power system are improved, and higher parameter precision and a better track fitting effect are obtained with limited simulation and measurement cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system simulation, and in particular to a kind of measurement point collaborative selection method based on system parameter checking scene. BACKGROUND

[0002] Power system simulation is an important tool for studying and optimizing power system operation, and the accuracy of simulation results directly affects the reliability of dispatching decisions and safety checking, so timely and effective adjustment of model error related parameters is crucial. Currently, methods based on sensitivity identification usually analyze the response of variable trajectories to each parameter based on a given number of disturbance scenarios and fixed measurement points, and construct the parameter set to be checked. This method has certain advantages in revealing the influence of parameters on system dynamic characteristics, but often defaults the fault type, fault location, measurement point distribution and time window to be pre-selected by experience, lacks systematic and quantitative evaluation of which scenes, which measurement points and which time periods are more conducive to parameter identification, resulting in that available information is not fully utilized, identifiable capacity depends on experience configuration, and stability and universality are insufficient.

[0003] On this basis, ordinary optimization algorithms (such as particle swarm optimization, genetic algorithm, etc.) and recursive filtering methods such as extended Kalman filter and ensemble Kalman filter are widely used for parameter correction, which usually aims to minimize the error between simulation values and measured values, and constantly adjusts parameters. Although these methods can theoretically introduce multi-event and multi-measurement point data, in actual application, all available scenes and measurement points are simply superimposed, without distinguishing the parameter identifiable contribution of different "scene-measurement point-time window" combinations through Fisher information, condition number, etc. On the one hand, in high-dimensional parameter space, problems such as ill-conditioned sensitivity matrix and strong parameter collinearity may occur, making the optimization result sensitive to scene selection, and even multiple solutions and instability may occur; on the other hand, under the engineering conditions of limited simulation times, test opportunities and online computing resources, the existing technology cannot answer which scenes and measurement points should be prioritized for parameter checking under given resource constraints, and lacks a scene and measurement point collaborative selection mechanism based on identifiable degree optimization, which greatly restricts the further improvement of parameter checking effect. SUMMARY

[0004] To solve the problem of unstable checking effect of the existing technology system parameter checking scene, the purpose of the present application is to provide a kind of measurement point collaborative selection method based on system parameter checking scene, electronic equipment and storage medium.

[0005] To solve the above technical problems, in a first aspect, according to some embodiments, the present application provides a kind of measurement point collaborative selection method based on system parameter checking scene, comprising:

[0006] S1, candidate scene and measurement point set construction and trajectory sensitivity calculation, including: selecting a measurement point set according to a fault scene or a disturbance scene, determining a time window, calculating the sensitivity of a parameter, splicing the sensitivity of each parameter direction to obtain a sensitivity matrix;

[0007] S2, construction and comprehensive scoring of distinguishability index, the distinguishability index at least including an information quantity index, a weakest direction distinguishability index, and a condition number index, and performing normalization processing according to a candidate set to form a distinguishability comprehensive score;

[0008] S3, scene and measurement point collaborative optimization selection based on distinguishability, including: introducing a decision variable, constructing an optimization model according to a constraint condition, and obtaining a preferred scene-measurement point-time window combination set with high comprehensive distinguishability and meeting resource constraints after optimization solving;

[0009] S4, parameter checking and model updating on the preferred combination, including: constructing a comprehensive error evaluation function, performing iterative minimization through a parameter estimation algorithm to obtain a parameter checking result, writing the parameter checking result back to a simulation model, and obtaining a power system simulation model corrected after checking.

[0010] Optionally, in some embodiments, the selecting a measurement point set according to a fault scene or a disturbance scene, determining a time window, calculating the sensitivity of a parameter, splicing the sensitivity of each parameter direction to obtain a sensitivity matrix, specifically includes:

[0011] Suppose that a candidate combination is determined by a fault / disturbance scene, a selected measurement point set, and a time window, and a corresponding trajectory or trajectory feature vector is represented by formula one as follows:

[0012] (1)

[0013] wherein, is a target parameter vector;

[0014] By applying an increment to each parameter, the overall sensitivity of the trajectory to the parameter under the combination is calculated in a double difference manner, and is represented by formula two as follows:

[0015] (2)

[0016] wherein, is a unit vector of the i-th parameter direction, indicates the sensitivity vector of the trajectory to the parameter under the combination;

[0017] ​​​​By concatenating the sensitivities of each parameter direction, a combined result is obtained. Sensitivity matrix S K .

[0018] Optionally, in some embodiments, the information content indicator includes:

[0019] Let the measurement noise covariance be... , No. Information matrix F of the combination k It is expressed using Formula 3, as follows:

[0020] (3)

[0021] Normalize the noise variance for different measurement types, and assume... ,get ;

[0022] The information content index is expressed using Formula 4, as follows:

[0023] (4)

[0024] in This is a regularization term used to avoid matrix singularities or ill-conditioned matrices. Reflection Combination The larger the value of the information volume provided in the parameter space, the higher the overall recognizability.

[0025] Optionally, in some embodiments, the weakest direction discernibility index specifically includes:

[0026] Let the information matrix be eigenvalues The weakest direction discernibility index is represented by Formula 5, as follows:

[0027] (5)

[0028] The weakest direction discernibility index is used to reflect the information strength in the most difficult-to-discern direction. The larger the value, the less discernible the parameters are in all directions.

[0029] Optionally, in some embodiments, the condition number index specifically includes:

[0030] Let the information matrix be eigenvalues The condition number index is expressed using Formula Six, as follows:

[0031] (6)

[0032] The larger the condition number index, the more severe the collinearity among the parameters and the worse the identifiability.

[0033] Optionally, in some embodiments, the normalization process based on the candidate set to form a comprehensive identifiability score specifically includes:

[0034] Information content indicators Weakest direction identifiability index Condition number index The candidate set is normalized, as shown in Formula 7, as follows:

[0035] (7)

[0036] in To prevent small amounts from being divided by zero, These are the minimum and maximum values ​​of the corresponding indicators among all candidate combinations;

[0037] combination The overall recognizability score is represented by Formula 8, as follows:

[0038] (8)

[0039] in These are weighting coefficients, set according to project requirements or determined through experience-based parameter tuning; The larger the value, the stronger the ability of the scene-measuring point-time window combination to identify the target parameters.

[0040] Optionally, in some embodiments, the introduction of decision variables, the construction of an optimization model based on constraints, and the optimization solution yielding a set of preferred scenario-measuring point-time window combinations with high comprehensive identifiability and satisfying resource constraints, specifically including:

[0041] Assume there is a total Each candidate scenario-measurement point-time window combination is used to introduce decision variables. It is represented by Formula Nine, as follows:

[0042] (9)

[0043] Given the maximum number of possible combinations Maximum number of available measurement points Under equal constraints, an optimization model is constructed, expressed by Formula 10, as follows:

[0044] (10)

[0045] For combination The overall recognizability score;

[0046] The constraint condition is expressed by formula eleven as follows:

[0047] (11)

[0048] wherein, represents the number of measuring points contained in the i-th combination, represents the maximum number of available measuring points.

[0049] Optionally, in some embodiments, the constructing the comprehensive error evaluation function specifically comprises:

[0050] Supposing that the measured trajectory feature vector is and the simulated trajectory feature vector is under the selected combination, the constructing the comprehensive error evaluation function is expressed by formula twelve as follows:

[0051] (12)

[0052] wherein, is a weight matrix, used for adjusting the relative importance of different scenes and different measurement types in error evaluation.

[0053] In a second aspect, the embodiments of the present application further provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method of any one of the first aspect.

[0054] In a third aspect, the embodiments of the present application further provide a computer readable storage medium, having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the method of any one of the first aspect.

[0055] ​The technical scheme of the present application has at least the following beneficial technical effects: the technical scheme of the present application, through sensitivity analysis and Fisher information measurement on various fault scenarios, measurement point configuration and time window, constructs comprehensive distinguishability scores and automatically selects the optimal combination under resource constraints, and then performs parameter checking on these combinations with the highest information amount, thereby achieving higher parameter accuracy and better trajectory fitting effect with limited simulation and measurement cost. The calculation example verification shows that the present application can effectively improve the accuracy and reliability of the power system simulation model parameters, and provides a feasible way for engineering field to build a standardized parameter checking scenario library and carry out efficient model maintenance; by constructing a candidate scenario library containing multiple types of disturbance faults, operating modes and measurement configurations, and based on simulation trajectory sensitivity analysis, Fisher information amount, condition number and other distinguishability indexes between the scenario-measurement point-time window combination and the target parameter are established, and the optimal scenario, measurement point and time window combination for parameter checking are optimized and selected under the resource constraints such as simulation times and available measurement points, thereby improving the accuracy and engineering applicability of the power system parameter checking results under limited calculation and test cost. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the drawings needed to be used in the embodiments 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.

[0057] Figure 1 is a schematic block diagram of a measurement point cooperative selection method based on a system parameter checking scenario provided by an embodiment of the present application.

[0058] Figure 2 is a 3-machine 9-node system for calculation example verification provided by an embodiment of the present application.

[0059] Figure 3 is a comparison diagram of simulation error improvement effect provided by an embodiment of the present application.

[0060] Figure 4 is a schematic block diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0062] Furthermore, in the following description, descriptions of well-known structures and techniques are omitted to avoid unnecessarily obscuring the concept of the present application.

[0063] Furthermore, the technical features involved in the different embodiments of the application described below can be combined with each other as long as there is no conflict.

[0064] In the current power system parameter checking research, the general practice is to construct simulation-measured error trajectories under a given small number of disturbance scenarios, fixed measurement points and unified time windows, and then use sensitivity analysis, optimization algorithms or recursive filtering methods to correct the model parameters. This process can to some extent realize the automatic adjustment of parameters, but the default "which scenarios, which measurement points, which time period" is pre-selected by experience, lacking systematic quantitative evaluation of the parameter identifiability of different fault types, fault locations, operating modes and measurement configurations. As a result, on the one hand, some disturbance scenarios are not enough to stimulate the target parameters or have highly redundant information, resulting in ill-conditioned sensitivity matrix, serious parameter collinearity, and easy occurrence of multiple solutions, unstable results, etc. On the other hand, under the condition of limited simulation times, test opportunities and online computing resources, it is impossible to answer the question "which scenarios and measurement points should be prioritized to achieve the best checking effect under limited resources", making it difficult to balance the accuracy and efficiency of parameter checking.

[0065] Therefore, the present application proposes a power system parameter checking scene and measurement point collaborative selection method based on parameter identifiability optimization, which constructs Fisher information, condition number and other identifiability indicators by simulating and analyzing the sensitivity of candidate combinations containing multiple fault / disturbance types, different operating modes and various measurement configurations, quantitatively evaluates the parameter identification information carried by each "scene-measurement point-time window" combination, and establishes a collaborative optimization model under the constraints of simulation times, available measurement points and other resources to automatically select the optimal scene set and measurement configuration for parameter checking. In this way, without changing the existing parameter checking algorithm framework, the parameter identifiability under the condition of unit resources is maximized, and the accuracy, stability and engineering applicability of the parameter checking results of complex power systems are improved.

[0066] The technical concept of the present application is as follows:

[0067] (1) Candidate scene and measurement configuration construction and identifiability analysis

[0068] A candidate "scenario-measuring point-time window" set containing multiple fault types, fault locations, fault clearing times, and multiple measuring point combinations is constructed by making full use of the historical disturbance records, operating modes, and measurement configuration information of the power system. The target parameters are perturbed slightly around the prior parameters, and simulation is carried out to extract the corresponding trajectories or trajectory features, forming the parameter sensitivity matrix under each combination. On this basis, the Fisher information, the minimum eigenvalue, and the condition number are calculated to quantify the influence of different "scenario-measuring point-time window" combinations on the identification ability of the target parameters and the parameter collinearity, providing basic data for subsequent optimization selection.

[0069] (2) Scenario and measuring point collaborative selection based on identifiability optimization and parameter checking

[0070] On the basis of obtaining the identifiability indicators of each candidate combination, engineering constraints such as the upper limit of the number of simulations, the number of available measuring points, and the coverage range of the operating mode are introduced to construct a collaborative optimization model with the goal of improving the comprehensive identifiability score, and the preferred scenario set, measuring point combination, and time window configuration for parameter checking are automatically selected. Subsequently, the existing parameter checking algorithms (such as optimization, recursive filtering, or deep learning methods) are called on the preferred combination to correct the target parameters; by carrying out checking only on the scenario and measuring point combination with the highest information, the simulation and calculation costs are significantly reduced while ensuring the parameter identifiability and the accuracy of the results, thereby improving the efficiency and engineering applicability of power system parameter checking.

[0071] As shown in Figure 1 , the embodiment of the present application provides a measuring point collaborative selection method based on system parameter checking scenarios, which includes:

[0072] S1, candidate scenario and measuring point set construction and trajectory sensitivity calculation, including: selecting a measuring point set according to a fault scenario or a disturbance scenario, determining a time window, calculating the sensitivity of the parameters, splicing the sensitivity of each parameter direction to obtain a sensitivity matrix;

[0073] S2, identifiability indicator construction and comprehensive score, the identifiability indicators at least include information quantity indicators, weakest direction identifiability indicators, and condition number indicators, and are normalized according to the candidate set to form an identifiability comprehensive score;

[0074] S3, scenario and measuring point collaborative optimization selection based on identifiability, including: introducing decision variables, constructing an optimization model according to the constraint conditions, and obtaining a preferred scenario-measuring point-time window combination set with high comprehensive identifiability and satisfying resource constraints after optimization solution;

[0075] S4, parameter checking and model updating on the preferred combination, comprising: constructing a comprehensive error evaluation function, obtaining parameter checking results by iterative minimization through parameter estimation algorithm, writing the parameter checking results back to the simulation model, and obtaining the corrected power system simulation model.

[0076] The following will be described in detail.

[0077] 1. Candidate scenario and measurement point set construction and trajectory sensitivity calculation

[0078] There are many fault types, fault locations and operating modes in power systems. Directly carrying out parameter checking on all possible scenarios and all measurement points not only has huge calculation overhead, but also contains a large number of combinations with insufficient excitation or redundant information for target parameters. Therefore, it is necessary to first construct a limited scale of candidate "scenario-measurement point-time window" set, and to quantitatively analyze the parameter sensitivity thereof.

[0079] Let the th candidate combination be determined by the fault / disturbance scenario, the selected measurement point set and the time window, and the corresponding trajectory or trajectory feature vector be denoted as:

[0080] (1)

[0081] wherein, is the target parameter vector.

[0082] By applying an increment to each parameter, the overall sensitivity of the trajectory to the parameter under the combination is calculated by using the double difference method:

[0083] (2)

[0084] wherein, is the unit vector in the direction of the th parameter, denotes the sensitivity vector of the trajectory to the parameter under the combination . By splicing the sensitivity in each parameter direction, the sensitivity matrix Sk of the combination is obtained.

[0085] In order to facilitate comparison between different combinations and different parameters, the sensitivity can be normalized by column.

[0086] 2. Construction of distinguishability index and comprehensive score

[0087] After obtaining the sensitivity matrix of each candidate combination, in order to quantitatively depict the distinguishing ability of different "scenario-measurement point-time window" combinations to the target parameters, it is necessary to construct a distinguishability index. Assuming that the measurement noise covariance is , the Fisher information matrix of the jth combination is defined as:

[0088] (3)

[0089] In engineering applications, if the noise variances of different measurement types have been normalized, we can let In this case, we have

[0090] Based on the above, the following discriminability indices are constructed:

[0091] (1) Information volume index

[0092] (4)

[0093] where is a regularization term to avoid matrix singularity or ill-conditioning; reflects the information "volume" provided in the parameter space, and the larger the value, the higher the overall discriminability.

[0094] (2) Discriminability in the weakest direction (minimum eigenvalue) Let be the eigenvalues of , then we define:

[0095] (5)

[0096] This index reflects the information intensity in the "most difficult to identify direction", and the larger the value, the less likely it is that the discriminability in all parameter directions will be too low.

[0097] (3) Condition number index (collinearity penalty)

[0098] (6)

[0099] The larger the value, the more serious the collinearity between parameters, and the worse the discriminability.

[0100] In order to achieve a comprehensive evaluation of multiple indices, we normalize , and according to the candidate set. Let:

[0101] (7)

[0102] where is a small quantity to prevent division by zero, are the minimum and maximum values of the corresponding index in the entire candidate combination, respectively.

[0103] ​​​​Final definition of combination Overall recognizability score:

[0104] (8)

[0105] in These are weighting coefficients, which can be set according to project requirements or determined through experience-based parameter tuning. The larger the value, the stronger the ability of the "scene-measurement point-time window" combination to identify the target parameters.

[0106] 3. Collaborative optimization selection of scene and measurement points based on identifiability

[0107] The overall recognizability score of each candidate combination was obtained. Afterwards, engineering constraints such as the number of simulations, the number of usable scenarios, and the number of available measurement points need to be considered, and the scenarios and measurement points need to be selected in a coordinated and optimized manner.

[0108] Assume there is a total A candidate "scenario-measurement point-time window" combination is introduced, and decision variables are defined as follows:

[0109] (9)

[0110] Given the maximum number of possible combinations Maximum number of available measurement points Under equal constraints, the following optimization model can be constructed:

[0111] (10)

[0112] The constraints include:

[0113] (11)

[0114] in, Indicates the first The number of measurement points included in each combination. Additional constraints such as operational mode coverage and geographical regional balance can also be added as needed.

[0115] The above optimization problem can be solved using integer programming, greedy search, genetic algorithms, etc.: For a medium-sized candidate set, integer linear programming can be used directly to obtain the optimal solution; for a large-scale candidate set, methods based on... The greedy or heuristic algorithm is used for pre-selection, and then combined with the local search algorithm for fine adjustment, so as to obtain the near-optimal combination of scene and measurement point within an acceptable computational complexity range.

[0116] After optimization, a set of optimal "scenario-measurement point-time window" combinations that have the highest overall identifiability and satisfy resource constraints can be obtained. , as the data basis of the subsequent parameter verification stage.

[0117] 4. Parameter verification and model updating on preferred combinations

[0118] After the scene and measurement point are selected, the specific parameter verification algorithm can be performed on the preferred combination set The present application does not limit the type of parameter verification algorithm, which can be flexibly combined with existing optimization, recursive filtering or deep learning methods.

[0119] Let the measured trajectory feature vector be and the simulated trajectory feature vector be The following comprehensive error evaluation function can be constructed:

[0120] (12)

[0121] wherein is the weight matrix, used to adjust the relative importance of different scenes and different measurement types in error evaluation.

[0122] Through the particle swarm algorithm or other parameter estimation algorithm, the iterative minimization of is performed to obtain the parameter verification result Since the solution is only performed on the scene and measurement point combination with the highest distinguishability, the degree of ill-conditioning of the sensitivity matrix is alleviated, the parameter collinearity effect is reduced, and the stability and physical reasonableness of the verification result are significantly improved.

[0123] Writing back to the simulation model, the corrected power system simulation model after verification can be obtained, which is used for subsequent stability analysis, safety verification and planning evaluation, etc.

[0124] To further illustrate the technical effects of the embodiments of the present application, the present application verifies the effectiveness of the scheme through an example.

[0125] According to the IEEE standard 3-machine 9-node system, an example analysis and verification is performed, as shown in Figure 2 The generator and excitation system parameters given in the standard model are taken as the "true parameters", denoted as To simulate the parameter drift caused by long-term operation of the equipment, certain random perturbations are applied to part of the key dynamic parameters to obtain the "prior parameter" set , which is used as the unverified simulation model.

[0126] In this example, six typical dynamic parameters are selected as target verification parameters, including the inertia time constant of three generators and the excitation-related time constant and gain. Part of the parameter settings are shown in Table 1.

[0127] Table 1 Real value and prior value of partial dynamic parameters of 3-machine 9-node system

[0128]

[0129] In the simulation, the generator adopts a fourth-order model, the excitation system adopts a second-order model, and the load adopts a constant power model. The fault scenario is uniformly three-phase short circuit: the fault occurs at 1.0 s, the fault duration is 0.10 s; after the fault is cleared, the line where the fault occurs is removed, and the system continues to operate until 5 s, and the simulation step is 0.01 s.

[0130] The trajectory obtained under the real parameters is regarded as an "actual trajectory", and the trajectory obtained under the prior parameters is regarded as an "unverified simulation trajectory".

[0131] According to the method of the application, a plurality of "scene-measuring point-time window" combinations are constructed and the distinguishability is analyzed, and the following are mainly set:

[0132] The fault positions are selected as nodes 4, 5 and 7, and the system is operated under rated load and 1.2 times rated load;

[0133] The measuring points include: active power of 3 generators, voltage amplitude and phase angle of adjacent buses of the fault, and active power of key tie lines;

[0134] The time window is divided into two types: 0-1 s after fault removal (fast transient interval) and 1-3 s (damped oscillation interval).

[0135] Through the combination of the above factors, a total of 12 candidate "scene-measuring point-time window" combinations are formed, the sensitivity matrix and the corresponding distinguishability index are calculated for each combination, and the comprehensive score is obtained. Part of the results are shown in Table 2.

[0136] In order to verify the advantages of the application, three verification schemes are set up for comparison:

[0137] Scheme A: traditional single-scene verification:

[0138] A single typical fault scene is selected, and the active power of 3 generators and the voltage data at the fault point are used to construct the objective function within the 0-3 s time window, and the particle swarm optimization algorithm is used to verify the 6 target parameters.

[0139] Scheme B: traditional multi-scene superposition verification:

[0140] Select all 12 candidate combinations, all scene and point corresponding error trajectory uniform superposition to build the objective function, also using particle swarm algorithm for joint check, not to be identifiable or scene screening optimization.

[0141] Scheme C: the invention of identifiable optimization check:

[0142] According to the method of the present application, based on the identifiable score Select the preferred combination Build the objective function, use the same particle swarm algorithm parameters and maximum iteration times as scheme A, B for checking.

[0143] The difference between the three schemes is only "which scene and measuring point is used", and the rest of the algorithm configuration is exactly the same.

[0144] Table 2 gives the checking results of some target parameters under the three schemes. It can be seen that the correction accuracy of the invention scheme is obviously improved compared with the traditional method, and the result is closer to the true value.

[0145] Table 2 Comparison of parameter checking results of three schemes:

[0146]

[0147] Take the average relative error of 6 target parameters as the overall index (denoted as MAPE), and the following information can be obtained:

[0148] Unchecked prior parameters: MAPE is about 15.7%;

[0149] Scheme A (traditional single scene): MAPE is about 7.8%;

[0150] Scheme B (all scene superposition): MAPE is about 5.8%;

[0151] Scheme C (the method of the present application): MAPE is about 2.5%.

[0152] It can be seen that without changing the checking algorithm itself, only by optimizing and selecting the "scene-measuring point-time window" identifiable, the average error of the present application can be effectively compressed to about 2%-3% level.

[0153] In addition, in order to intuitively show the improvement effect of simulation error, the active power of generator 1 Trajectory comparison, as shown in Figure 3 The figure gives 4 curves, denoted as (a), (b), (c), (d).

[0154] Trajectory (a): using the true parameters The active power trajectory obtained by simulation is taken as a benchmark of "actual trajectory".

[0155] Trajectory (b): simulation result under prior parameters The trajectory obtained by simulation represents the simulation result without parameter checking. It can be seen that the overshoot after the fault is large, and the oscillation frequency and the decay speed are obviously deviated from the actual trajectory.

[0156] Trajectory (c): simulation result under the traditional single-scene checking scheme. Compared with the un-checked condition, the amplitude of the first swing period and the steady-state recovery value have been improved to a certain extent, but there is still a relatively obvious phase deviation and residual error in the oscillation decay stage.

[0157] Trajectory (d): simulation result under the method of the application. The peak value after fault removal, swing frequency and oscillation decay process are highly consistent with the actual trajectory, the steady-state recovery value is basically coincided, and the root mean square error of simulation-actual trajectory is obviously smaller than scheme (c).

[0158] Effect analysis:

[0159] 1. The application no longer defaults to specify a single fault scene and fixed measuring point by experience, but based on the sensitivity matrix and Fisher information, the minimum eigenvalue and the condition number and other distinguishability indexes, a plurality of "fault location-operation mode-measuring point combination-time window" are comprehensively scored, and then a small number of preferred combinations are automatically selected from them. In the example, 12 groups of candidate combinations are constructed in the 3-machine 9-node system, after the distinguishability analysis, only 3 groups of combinations with the highest information amount are selected for parameter checking: including the node 7 fault, the rated and 1.2 times load, the observation of the transient interval of the active power of 3 generators and the voltage / phase angle of the adjacent bus of the fault, and the node 5 fault, the strong 1-3s damping oscillation combination. The results show that the application can automatically identify the "most sensitive to target parameters, most distinguishable" scene and measuring point configuration, avoid the traditional method of using a large number of scenes which are insufficiently excited or information redundant to target parameters, and improve the pertinence and effectiveness of parameter checking from the source.

[0160] 2. By calculating the Fisher information matrix and its eigen-spectrum of different combinations, the invention realizes the quantitative evaluation of the distinguishability of each combination parameter. The example results show that when the fault occurs at node 7 and the measurement covers the generator active power and the voltage / phase angle of the adjacent bus, the information quantity index is higher, the minimum eigenvalue is significantly larger than that of other combinations, and the condition number of the sensitivity matrix is smaller, indicating that in these working conditions, the target parameters (such as the inertia constant, direct-axis transient time constant, and excitation gain of each unit) can be effectively distinguished in multiple directions; on the contrary, when the fault is far away from the target area and the measurement points are concentrated on the far-end bus, the condition number is larger, showing strong collinearity, indicating that its contribution to the identification of the target parameters is limited. Through the distinguishability optimization, the invention retains the representative coverage of typical working conditions (different fault points, different load levels, and different dynamic stages) under the constraint of a limited number of scenarios, while eliminating redundant combinations with low distinguishability, making the parameter checking problem more numerically robust.

[0161] 3. In the preferred scenario and measurement point combination, the invention uses the same optimization algorithm as the traditional method to check the 6 target parameters, but has obvious advantages in parameter accuracy and result consistency. In the example, the average relative error (MAPE) of the uncalibrated prior parameters is about 15.7%; the traditional single-scenario method reduces to about 7.8%; in the case of adding all 12 groups of scenarios without screening, although the error is further reduced to about 5.8%, the cost is a large amount of redundant simulation and strong ill-conditioned. After using the distinguishability optimization scheme of the invention, the average relative error is about 2.5%, and the key parameters are significantly close to the true value. For example, the inertia constant of generator 1 from the prior 4.02s to 3.53s, close to the true value 3.50s; the inertia constant of generator 2 from 3.28s to 3.95s, close to the true value 4.00s; the excitation gain of generator 1 from 230 to about 203, with only a small deviation from the true value 200. Compared with the traditional single-scenario or simple multi-scenario superposition method, the invention realizes higher parameter checking accuracy and better parameter consistency while keeping the calculation scale controllable.

[0162] 4. From the simulation results, taking the active power trajectory of generator 1 as an example, Figure 2The four curves in the figure correspond to: trajectory (a) is the actual trajectory, trajectory (b) is the uncalibrated simulation result, trajectory (c) is the traditional single-scene calibration result, and trajectory (d) is the calibration result of the method. By comparison, it can be directly observed that: the uncalibrated trajectory (b) has a large overshoot, a deviation in oscillation frequency and a slow decay rate in the first swing cycle after the fault; the trajectory (c) obtained by the traditional method has improved overshoot and steady-state value, but there is still a significant phase deviation and residual error in the oscillation decay stage; the trajectory (d) after calibration by the method has a significantly improved degree of coincidence with the actual trajectory (a) in peak value, swing frequency and damping decay curve, the steady-state recovery value is basically coincident, and the root mean square error of the trajectory is significantly reduced compared with the uncalibrated model, and the preferred combination is further reduced compared with the traditional single-scene method. It can be seen from the iteration statistical information in the example that, under the premise of using the same optimization algorithm and iteration upper limit, the method reduces the scenes with low distinguishability and redundant measurements, so that the total simulation call number is reduced by more than one order of magnitude compared with the simple multi-scene superposition scheme, and the calibration accuracy is higher, which proves the comprehensive advantages of the method in parameter calibration effect and calculation efficiency.

[0163] The application proposes a "scene-measuring point-time window" integrated modeling and sensitivity analysis process. Based on the power grid topology, operating mode and measurement configuration, a variety of fault scenes and observation time period combinations are systematically constructed, and trajectory sensitivity matrix and Fisher information and other indicators are calculated for each combination, which directly links fault conditions, measuring point configuration and parameter distinguishability, and realizes quantitative evaluation of the distinguishability of each candidate combination. A comprehensive distinguishability score and a resource-constrained collaborative optimization mechanism are constructed. After normalization and weighting of information quantity, minimum eigenvalue, condition number and other indicators, a single score is formed Under the constraints of "optional scene number, available measuring point number and simulation number", the combination selection is converted into a 0-1 optimization problem, and a small number of "scene-measuring point-time window" combinations with the highest parameter distinguishability are automatically selected. A method for constructing a unified parameter calibration objective function based on the preferred combination is proposed. A multi-scene weighted error function is constructed only on a small number of combinations selected by distinguishability optimization, and these high-information-content scenes are used as calibration benchmarks, which not only reduces redundant simulation and calculation overhead, but also alleviates the problem of sensitivity matrix ill-conditioning, and can be directly integrated with various existing optimization, filtering or deep learning parameter calibration algorithms. An extensible parameter calibration scene library and an adaptive updating mechanism are formed. By evaluating and recording the distinguishability scores of each typical operating condition offline, a standardized "parameter calibration scene library" for different regions and device types can be constructed, and the preferred combination can be quickly updated when the topology or operating mode changes, providing reusable and expandable design and protection objects for long-term maintenance of simulation model parameters in engineering sites.

[0164] The embodiment of the application also provides an electronic device 400, such asFigure 4 As shown, it comprises a memory 401, a processor 402 and a computer program stored in the memory 401 and executable on the processor, wherein the processor 402 implements the steps of the method according to any one of the above embodiments when executing the program.

[0165] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the steps of the method according to any one of the above embodiments.

[0166] The embodiment of the present application also provides a computer program product, which comprises a computer program stored in a computer readable storage medium, and when a processor of an electronic device reads the computer program from the computer readable storage medium, the processor executes the computer program, so that the electronic device executes the steps of the method according to any one of the above embodiments.

[0167] Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0168] It should be understood that the above specific embodiments of the present application are only used for illustrative or explanatory purposes, and do not constitute a limitation on the present application. Therefore, any modification, equivalent replacement, improvement, etc. made without departing from the spirit and scope of the present application shall be included in the protection scope of the present application. In addition, the appended claims of the present application are intended to cover all variations and modifications falling within the scope and boundary of the appended claims, or the equivalent forms of such scope and boundary.

Claims

1. A method for collaborative selection of measurement points based on system parameter verification scenarios, characterized in that, include: S1. Construction of candidate scenarios and measurement point sets and calculation of trajectory sensitivity, including: selecting a set of measurement points based on fault scenarios or disturbance scenarios, determining the time window, calculating the sensitivity of parameters, and splicing the sensitivity of each parameter direction to obtain a sensitivity matrix. S2. Construction and comprehensive scoring of identifiability indicators, wherein the identifiability indicators include at least information content indicators, weakest direction identifiability indicators, and condition number indicators, and are normalized according to the candidate set to form a comprehensive identifiability score. S3. Collaborative optimization selection of scenarios and measurement points based on identifiability, including: introducing decision variables, constructing an optimization model based on constraints, and after optimization and solution, obtaining a set of preferred scenario-measurement-time window combinations with high comprehensive identifiability and meeting resource constraints; S4. Parameter verification and model update on the preferred combination, including: constructing a comprehensive error evaluation function, iteratively minimizing it through a parameter estimation algorithm to obtain parameter verification results, writing the parameter verification results back to the simulation model, and obtaining the verified and corrected power system simulation model.

2. The method according to claim 1, characterized in that, The process of selecting a set of measurement points based on the fault or disturbance scenario, determining the time window, calculating the sensitivity of the parameters, and concatenating the sensitivities of each parameter direction to obtain a sensitivity matrix specifically includes: Let the first Each candidate combination is determined by the fault / disturbance scenario, the selected set of measurement points, and the time window. The corresponding trajectory or trajectory feature vector is represented by Formula 1, as follows: (1) in, , where is the target parameter vector; By applying an increment to each of the parameters The overall sensitivity of the trajectory to the parameters under this combination is calculated using a bilateral difference method, and is expressed by Formula 2 as follows: (2) in, For the first A unit vector in each parameter direction. Indicate combination Lower trajectory parameters The sensitivity vector; By concatenating the sensitivities of each parameter direction, a combined result is obtained. Sensitivity matrix S K .

3. The method according to claim 1, characterized in that, The information content indicators include: Let the measurement noise covariance be... , No. Information matrix F of the combination k It is expressed using Formula 3, as follows: (3) Normalize the noise variance for different measurement types, and assume... ,get ; The information content index is expressed using Formula 4, as follows: (4) in This is a regularization term used to avoid matrix singularities or ill-conditioned matrices. Reflection Combination The larger the value of the information volume provided in the parameter space, the higher the overall recognizability.

4. The method according to claim 3, characterized in that, The weakest direction discernibility index specifically includes: Let the information matrix be eigenvalues The weakest direction discernibility index is represented by Formula 5, as follows: (5) The weakest direction discernibility index is used to reflect the information strength in the most difficult-to-discern direction. The larger the value, the less discernible the parameters are in all directions.

5. The method according to claim 3, characterized in that, The condition number index specifically includes: Let the information matrix be eigenvalues The condition number index is expressed using Formula Six, as follows: (6) The larger the condition number index, the more severe the collinearity among the parameters and the worse the identifiability.

6. The method according to claim 1, characterized in that, The normalization process based on the candidate set to form a comprehensive identifiability score specifically includes: Information content indicators Weakest direction identifiability index Condition number index The candidate set is normalized, as shown in Formula 7, as follows: (7) in To prevent small amounts from being divided by zero, These are the minimum and maximum values ​​of the corresponding indicators among all candidate combinations; combination The overall recognizability score is represented by Formula 8, as follows: (8) in These are weighting coefficients, set according to project requirements or determined through experience-based parameter tuning; The larger the value, the stronger the ability of the scene-measuring point-time window combination to identify the target parameters.

7. The method according to claim 1, characterized in that, The process involves introducing decision variables, constructing an optimization model based on constraints, and solving the optimization problem to obtain a set of optimal scenario-measurement point-time window combinations that are highly identifiable and meet resource constraints. Specifically, this set includes: Assume there is a total Each candidate scenario-measurement point-time window combination is used to introduce decision variables. It is represented by Formula Nine, as follows: (9) Given the maximum number of possible combinations Maximum number of available measurement points Under equal constraints, an optimization model is constructed, expressed by Formula 10, as follows: (10) For combination The overall recognizability score; The constraint conditions are expressed by Formula 11, as follows: (11) in, Indicates the first The number of measuring points included in each combination. This indicates the maximum number of available measurement points.

8. The method according to claim 1, characterized in that, The construction of the comprehensive error evaluation function specifically includes: Suppose that, under the selected combination, the measured trajectory feature vector is... The simulated trajectory feature vector is A comprehensive error evaluation function is constructed, expressed by Formula Twelve, as follows: (12) in, This is a weight matrix used to adjust the relative importance of different scenarios and measurement types in error evaluation.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-8.

Images