Inertia and damping online evaluation method, device and system of power system

Abstract

The invention discloses an inertia and damping online evaluation method, device and system of a power system, and belongs to the technical field of power system control. The online evaluation method comprises the following steps: slicing an observation point set in a preset time period according to a preset time interval, and inputting each slice observation set into a neural network embedded with physical information so as to output an evaluation result of slice inertia and damping; and the evaluation results of the slices are fused to obtain the evaluation results of the target inertia and damping. According to the invention, the measurement data is divided into a plurality of slices, then the slices are evaluated by using the corresponding neural networks, and finally the evaluation result is obtained through summarizing. According to the method, the problem of random errors caused by quasi-steady-state disturbance and damage measurement is solved, and the accuracy and robustness of evaluation are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of power system control, and more particularly relates to an inertia and damping online evaluation method, device and system for a power system. BACKGROUND

[0002] The rapid development and large-scale grid connection of new energy sources have caused the proportion of traditional synchronous machines in the power grid to decrease continuously, and the system presents low inertia and weak damping characteristics. As an important factor for maintaining the stability of the power grid, the decrease in inertia and damping levels will inevitably threaten the safe operation of the system. In order to discover and troubleshoot the risk of low-frequency oscillation in the system in a timely manner and develop corresponding stability control measures in advance, it is necessary for the power grid operators to monitor the inertia and damping levels of the system in real time and accurately.

[0003] The previous research on inertia and damping evaluation based on quasi-steady-state data includes the power spectral density method and the method based on variable covariance. However, after the quasi-steady state is accurately modeled, the power spectral density method can no longer observe the oscillation mode of the inertia center, so it cannot estimate the system inertia. The evaluation method based on covariance requires that the Jacobian matrix be known or that there be a linear relationship between active power and voltage phase angle, which hinders the application of such methods in actual power grids. At the same time, the fluctuation amplitude is small under the quasi-steady state, and the effective inertia response is easily submerged in the measurement noise, and the measurement data transmitted through the wide area will inevitably have bad or missing data, which will cause the accuracy of the evaluation to decrease. Therefore, it is urgent to propose a method that can effectively handle the random disturbance and damaged measurement under the quasi-steady state, but is more simple and practical. SUMMARY

[0004] In view of the above defects or improvement needs of the prior art, the present application provides an inertia and damping online evaluation method, device and system for a power system, which aims to solve the technical problem of poor accuracy of inertia and damping evaluation due to random errors caused by quasi-steady-state disturbance and damaged measurement.

[0005] To achieve the above-mentioned purpose, according to one aspect of the present application, an inertia and damping online evaluation method for a power system is provided, comprising: S1: When the quasi-steady-state measurement of the power system is damaged, the operating state parameters of the power system collected by the phasor measurement unit at each sampling time in a preset time period and the time stamp corresponding to the sampling time are taken as observation points, and then the observation point set corresponding to the preset time period is obtained; S2: The observation point set corresponding to the preset time period is sliced at a preset time interval to obtain a plurality of sliced observation sets; S3: input each of the slice observation sets into a preset neural network embedded with physical information, so that the preset neural network outputs a slice inertia evaluation result and a slice damping evaluation result of each of the slice observation sets; the physical information includes a rotor motion equation corresponding to a to-be-measured inertia and a to-be-measured damping; S4: fuse the slice inertia evaluation result and the slice damping evaluation result corresponding to each of the slice observation sets respectively, to obtain a target inertia evaluation result and a target damping evaluation result correspondingly.

[0006] Further, the S2 comprises: segmenting the observation point set corresponding to the preset time period at a preset time interval to obtain a plurality of slices; and randomly sampling in the time domain of each of the slices to obtain at least one slice observation set corresponding to the slice.

[0007] Further, the S4 comprises: S41: deleting an inertia outlier and a damping outlier from the slice inertia evaluation result and the slice damping evaluation result corresponding to all the slice observation sets respectively; S42: performing weighted fusion on each of the remaining slice inertia evaluation results to obtain the target inertia evaluation result; and performing weighted fusion on each of the remaining slice damping evaluation results to obtain the target damping evaluation result.

[0008] Further, the S41 comprises: using a formula finding the inertia outlier and the damping outlier from the slice inertia evaluation result and the slice damping evaluation result corresponding to all the slice observation sets, and deleting them respectively; wherein, is a training loss of the i th preset neural network, the preset neural network sequence number with the minimum training loss is m, and the training loss of the preset neural network is , represents a set threshold value of abnormal loss screening.

[0009] Further, the S41 comprises: using a formula finding the inertia outlier and the damping outlier from the slice inertia evaluation result and the slice damping evaluation result corresponding to all the slice observation sets, and deleting them respectively; wherein, and respectively represent an i th slice inertia evaluation result and an i th slice damping evaluation result, , respectively represent the mean values of the inertia evaluation result and the damping evaluation result; , represent the standard deviation of the inertia evaluation result and the damping evaluation result respectively; , represent the set threshold of the inertia abnormal value and the damping abnormal value respectively.

[0010] Further, the S42 comprises: using the formula to perform weighted fusion on the remaining inertia evaluation results of each slice to obtain the target inertia evaluation result ; using the formula to perform weighted fusion on the remaining damping evaluation results of each slice to obtain the target damping evaluation result ; wherein, is the training loss of the i-th preset neural network, and represent the i-th slice inertia evaluation result and the i-th slice damping evaluation result respectively, , are the number of the remaining slice inertia evaluation results and the slice damping evaluation results respectively.

[0011] Further, the operating state parameters comprise: frequency f and power imbalance P , and the observation points are represented as f , P , t}, t represent the sampling time.

[0012] According to another aspect of the present application, there is provided an inertia and damping online evaluation device of a power system, used for executing the inertia and damping online evaluation method of the power system, comprising: a sampling module, configured to, when inertia and damping steady-state measurement of the power system is damaged, collect operating state parameters of the power system and time stamps corresponding to the sampling time of the sampling module at each sampling time in a preset time period as observation points, and then obtain an observation point set corresponding to the preset time period; a slicing module, configured to slice the observation point set corresponding to the preset time period at a preset time interval to obtain a plurality of corresponding slice observation sets; an evaluation module, configured to input each slice observation set into a preset neural network embedded with physical information, so that the preset neural network outputs a slice inertia evaluation result and a slice damping evaluation result of each slice observation set; the physical information includes a rotor motion equation corresponding to the to-be-measured inertia and the to-be-measured damping; a fusion module, configured to fuse the slice inertia evaluation result and the slice damping evaluation result corresponding to each slice observation set respectively, and correspondingly obtain a target inertia evaluation result and a target damping evaluation result.

[0013] According to another aspect of the present application, there is provided an inertia and damping online evaluation system of a power system, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the inertia and damping online evaluation method of the power system when executing the computer program.

[0014] According to another aspect of the present application, there is provided a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the inertia and damping online evaluation method of the power system when executed by a processor.

[0015] Overall, compared with the prior art, the above technical solutions conceived by the present application can achieve the following beneficial effects: (1) The present application provides an inertia and damping online evaluation method of a power system, which divides the measurement data into multiple slices, evaluates each slice with a separate neural network, and finally obtains statistical values by summarizing the evaluation results of all slices. This method solves the problem of random errors caused by quasi-steady-state disturbances and damaged measurements, greatly improving the accuracy and robustness of the evaluation.

[0016] (2) The present application considers that due to limited training data and potential risk of damaged measurement data, the accuracy of the evaluation results based on short-time slices cannot be guaranteed, and therefore multiple time slice evaluation results need to be summarized and statistically analyzed to improve the evaluation accuracy. This process consists of two steps: removing outliers and weighted averaging. The inertia and damping evaluation of each time slice corresponds to a training process aimed at minimizing the loss function. Smaller training loss means more favorable evaluation results. Therefore, the training loss of each evaluation is used as a confidence indicator for subsequent statistical processing. (3) The present application removes outliers in the step of removing outliers from both the training loss and the evaluation value itself, which can remove outliers from multiple aspects and thus improve the accuracy of the entire online evaluation algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 Flowchart of the inertia and damping online evaluation method of a power system provided for Embodiment 1 of the present application; Figure 2 Schematic diagram of the inertia and damping online evaluation method of a power system provided for Embodiment 1 of the present application; Figure 3 Error distribution schematic diagram of 50 short-time slices of 10 generators provided for Embodiment 1 of the present application; Figure 4 Relative error schematic diagram of inertia constant evaluation of each region under different wind power penetration rates provided for Embodiment 1 of the present application; Figure 5 The inertia constant evaluation error distribution diagram of all slices of a single area before and after outlier rejection provided for embodiment 1 of the present application is shown in the figure; Figure 6 The inertia constant and damping coefficient evaluation error diagram of sp-PINN provided for embodiment 1 of the present application when the measurement is damaged is shown in the figure. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0019] Embodiment 1 As shown in Figure 1 and Figure 2 , the present embodiment provides an inertia and damping online evaluation method of a power system, comprising: S1: when the quasi-steady state measurement of the power system is damaged, the operating state parameters of the power system collected by the phasor measurement unit at each sampling time in a preset time period and the time stamp corresponding to the sampling time are taken as observation points, and then the observation point set corresponding to the preset time period is obtained; S2: the observation point set corresponding to the preset time period is sliced at a preset time interval to obtain a plurality of sliced observation sets corresponding thereto; S3: each sliced observation set is input into a preset neural network embedded with physical information, so that the preset neural network outputs the sliced inertia evaluation result and the sliced damping evaluation result of each sliced observation set; the physical information includes the rotor motion equation corresponding to the to-be-measured inertia and the to-be-measured damping; S4: the sliced inertia evaluation result and the sliced damping evaluation result corresponding to each sliced observation set are fused respectively, and the target inertia evaluation result and the target damping evaluation result are obtained correspondingly.

[0020] Regarding S1, the operating state parameters are collected by the phasor measurement unit PMU. As an optional embodiment, the operating state parameters can include: frequency f and power imbalance P。 The frequency and power imbalance of a single generator (or an aggregated generator representing the entire area) and the corresponding time stamp are collected by the PMU deployed in the power system to form an observation point f , P , t}. The sampling observation points obtained in a time window are collected under quasi-steady state and used as input data of the evaluation method. In the given embodiment, the collection frequency of the PMU is set to 100 Hz, and the collection period is 100 s of time window data under quasi-steady state.

[0021] Regarding S2, time window slicing. The collected long time window observation dataset is split into smaller time intervals, which is called slicing, and the observation points in each slice form an observation set. As an optional implementation, S2 includes: splitting the observation set corresponding to a preset time period into multiple slices with a preset time interval; and performing random sampling in the time domain of each slice to obtain at least one slice observation set corresponding to the slice. In order to enhance sample diversity, random sampling is performed in the time domain of each slice to generate a collocation set. The observation set and the collocation set together constitute the training set of the PINN.

[0022] In the given embodiment, the length of each short time slice is set to 2s, i.e., 100s of total data is divided into 50 short slices. The number of collocation sets is set to twice the number of observation sets, so that for each short slice, the training set contains a total of 200 groups of observation points and 400 groups of collocation points.

[0023] Regarding S3, parallel evaluation. An independent physics-informed neural network (PINN) is arranged for each slice to perform evaluation. The rotor motion equation (or the equivalent aggregated rotor motion equation of the region) containing unknown inertia and damping is embedded in the neural network as known physical information. Thanks to the powerful generalization characteristics, these PINNs can use the same hyperparameters. At the same time, since the evaluation of different time slices is independent, they can be arranged to run in parallel to reduce time consumption.

[0024] In the given embodiment, the hyperparameters of the PINN are set as follows:

[0025] Regarding S4, statistical analysis. Once all the evaluations are completed, the evaluation results of different time slices are combined together. Finally, statistical processing including removing outliers and weighted average is performed to produce the final evaluation results.

[0026] It's important to note that the goal of this step is to segment the long-term frequency dynamics into short-term frequency dynamics so that PINN can better fit them. Long-term frequency dynamics are crucial for inertia and damping assessments because the accuracy of assessments based on a single short slice cannot be guaranteed, and assessments may even fail. For example, PINN might converge to a local optimum during training, or the sampled data might be corrupted by measurement noise or anomalies. Therefore, long-term dynamics containing more observational data are needed to ensure the accuracy and robustness of the assessment. When determining the length of each short slice, two main factors are considered. First, the time slice needs sufficient data to ensure normal PINN training, so it cannot be too short. Second, the slice cannot be too long, exceeding PINN's fitting capability. Therefore, a trial-and-error approach can be used to select the optimal slice length.

[0027] Furthermore, when long-term dynamics are divided into multiple short-time slices, it means that multiple evaluations need to be performed, and the accumulated time consumption increases proportionally with the number of slices. To avoid this problem, a parallel evaluation strategy is needed. Each slice uses a PINN with the same structure to estimate inertia and damping separately. In this way, the training time for long-term dynamics is actually equivalent to the evaluation time of a single slice. In addition, compared to using PINN to evaluate the entire long-term dynamics, each PINN deals with less data, which further reduces time consumption and improves evaluation efficiency.

[0028] Specifically, due to limited training data and the potential risk of measurement data corruption, the accuracy of evaluation results based on short time slices cannot be guaranteed. Therefore, it is necessary to aggregate the evaluation results from multiple time slices for statistical analysis to improve evaluation accuracy. This process consists of two steps: outlier removal and weighted averaging. The inertia and damping evaluations for each time slice correspond to a training process aimed at minimizing the loss function. A smaller training loss implies a more favorable evaluation result. Therefore, the training loss of each evaluation is used as a confidence metric for subsequent statistical processing. As an optional implementation, S4 includes: S41: deleting inertia outliers and damping outliers from the slice inertia evaluation results and slice damping evaluation results corresponding to all slice observation sets, respectively; S42: performing weighted fusion on the remaining slice inertia evaluation results to obtain the target inertia evaluation result; performing weighted fusion on the remaining slice damping evaluation results to obtain the target damping evaluation result.

[0029] As an optional implementation, S41 includes: using the formula From the slice inertia assessment results and slice damping assessment results corresponding to all slice observation sets, identify inertia outliers and damping outliers, and delete them respectively; among them, Let m be the training loss of the i-th preset neural network. The preset neural network with the smallest training loss is indexed as m, and its training loss is denoted as m. , This indicates the set threshold for filtering abnormal losses.

[0030] As an optional implementation, S41 includes: using the formula From the slice inertia assessment results and slice damping assessment results corresponding to all slice observation sets, identify inertia outliers and damping outliers, and delete them respectively; among them, and These represent the inertia evaluation result and the damping evaluation result of the i-th slice, respectively. , These represent the mean values ​​of the inertia assessment results and the damping assessment results, respectively. , The standard deviations of the inertia assessment results and the damping assessment results are represented by their respective values. , These represent the set thresholds for inertia anomalies and damping anomalies, respectively, and can typically be set to 2 or 3.

[0031] It should be noted that the outlier removal step filters out outliers from both the training loss and the evaluation value itself. In other words, it can include both methods mentioned above, meaning two outlier removal operations are performed simultaneously.

[0032] As an optional implementation, the weights of the weighted average are set to the reciprocal of the training loss; that is, the smaller the training loss, the larger its corresponding weight. S42 includes: using the formula The inertia assessment results of the remaining individual slices are weighted and fused to obtain the target inertia assessment result. Using formulas The remaining damping evaluation results of each slice are weighted and fused to obtain the target damping evaluation result. ;in, Let i be the training loss of the i-th preset neural network. and These represent the inertia evaluation result and the damping evaluation result of the i-th slice, respectively. , These represent the number of remaining slice inertia assessment results and slice damping assessment results, respectively.

[0033] like Figure 3As shown in the violin plot, the error distribution of 50 short-time slices for 10 generators is illustrated. The results show that the evaluation error of the 50 slices exhibits randomness; while the evaluation error is small in some slices, it exceeds 6% in others. This demonstrates that the results of evaluation using only a single short-time slice are unreliable. However, it is noteworthy that the evaluation error for each generator nearly follows a normal distribution with a mean of approximately 0. This indicates that the statistical values ​​from multiple slices can eliminate random errors caused by short time intervals or PINN training.

[0034] like Figure 4 The figure shows a schematic diagram illustrating the relative error of inertia constant assessment in three different regions under different wind power penetration rates. The wind power penetration rates in the three regions are 9.37%, 32.18%, and 49.87%, respectively. The assessment results are compared with the commonly used ARMAX (Autoregressive moving average with exogenous inputs) method in the field. As shown in the figure, the inertia assessment error of the sp-PINN method is less than 1% for different assessment objects under different renewable energy penetration levels, and is significantly better than the compared ARMAX method. This result verifies the assessment accuracy of the proposed method and its adaptability to different renewable energy penetration levels.

[0035] like Figure 5 The figure shows the distribution of inertia constant evaluation errors for all slices in a single region before and after outlier removal under measurement constraints. The measurement impairment scenario was defined as 5% of the measurement data exhibiting anomalies, including 4% missing data and 1% data jumping to a maximum / minimum value. As can be seen from the figure, before outlier removal, the evaluation error for some slices exceeded 50%, with the maximum exceeding 100%, indicating that the evaluation of these slices had actually failed due to noise / data anomalies. Fortunately, such evaluation failures were isolated incidents; the evaluation error for the vast majority of slices remained below 15%, and was almost uniformly distributed around 0%. However, the introduction of noise / data anomalies did significantly weaken the algorithm's evaluation accuracy, thus demonstrating the superiority of the proposed long-term multi-slice parallel evaluation. After removing outliers from the evaluation results of all slices, the distribution range of the evaluation error was reduced.

[0036] like Figure 6 The diagram illustrates the error assessment of the inertial constant and damping coefficient in three different regions using sp-PINN when measurements are impaired. Two scenarios were set up for measurement impairment: one with only 10% noise injection, and the other with an additional 5% data anomaly superimposed on top of the noise. The results are presented in... Figure 5The final result was obtained by weighted averaging after outlier removal. As shown in the figure, even with only measurement noise, the maximum evaluation error of the proposed algorithm does not exceed 3%. Even with the addition of data anomalies, the overall evaluation error remains below 6%. This demonstrates the accuracy of the proposed evaluation method and its robustness to damaged measurements.

[0037] Example 2 This embodiment provides an online evaluation device for the inertia and damping of a power system, used to perform the above-described online evaluation method for the inertia and damping of a power system, including: The sampling module is used to take the operating state parameters of the power system and the corresponding timestamps collected by the phasor measurement unit at each sampling time within a preset time period as observation points when the inertial steady-state measurement of the power system is damaged, and then obtain the set of observation points corresponding to the preset time period. The slicing module is used to slice the set of observation points corresponding to a preset time period at a preset time interval to obtain multiple sliced ​​observation sets. The evaluation module is used to input each slice observation set into the corresponding preset neural network embedded with physical information, so that the preset neural network outputs the slice inertia evaluation results and slice damping evaluation results for each slice observation set; the physical information includes the rotor motion equations corresponding to the inertia and damping to be measured. The fusion module is used to fuse the slice inertia assessment results and slice damping assessment results corresponding to each slice observation set to obtain the target inertia assessment result and target damping assessment result.

[0038] Example 3 The present invention also relates to an online evaluation system for the inertia and damping of a power system, comprising a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described online evaluation method for the inertia and damping of a power system.

[0039] The online evaluation system for the inertia and damping of this power system can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The memory can be used to store computer programs and / or modules. The processor implements the various functions of the online evaluation system for the inertia and damping of the power system by running or executing the computer programs and / or modules stored in the memory, and by accessing data stored in the memory.

[0040] Example 4 The present invention also relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described online evaluation method for the inertia and damping of a power system.

[0041] Specifically, the memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0042] Example 5 This invention provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the method described in the above embodiments of this invention.

[0043] The technical features of the embodiments described above can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. It should be noted that the terms "in one embodiment," "for example," and "again" in this invention are intended to illustrate the invention and are not intended to limit the invention.

[0044] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for online evaluation of the inertia and damping of a power system, characterized in that, include: S1: When the inertial steady-state measurement of the power system is damaged, the operating state parameters of the power system collected by the phasor measurement unit at each sampling time within a preset time period and the timestamps corresponding to the sampling times are used as observation points, thereby obtaining the set of observation points corresponding to the preset time period. S2: Slice the set of observation points corresponding to the preset time period at a preset time interval to obtain multiple sliced ​​observation sets; S3: Input each of the slice observation sets into the corresponding preset neural network embedded with physical information, so that the preset neural network outputs the slice inertia evaluation result and slice damping evaluation result of each of the slice observation sets; The physical information includes the rotor motion equations corresponding to the inertia and damping to be measured. S4: The slice inertia evaluation results and slice damping evaluation results corresponding to each slice observation set are fused to obtain the target inertia evaluation results and target damping evaluation results.

2. The online evaluation method for the inertia and damping of a power system as described in claim 1, characterized in that, S2 includes: dividing the set of observation points corresponding to the preset time period into multiple slices by a preset time interval; and randomly sampling within the time domain of each slice to obtain at least one slice observation set corresponding to the slice.

3. The online evaluation method for the inertia and damping of a power system as described in claim 1, characterized in that, S4 includes: S41: Delete the inertia outlier and the damping outlier from the slice inertia evaluation results and the slice damping evaluation results corresponding to all the slice observation sets, respectively; S42: The remaining slice inertia evaluation results are weighted and fused to obtain the target inertia evaluation result; The remaining damping evaluation results of each slice are weighted and fused to obtain the target damping evaluation result.

4. The online evaluation method for the inertia and damping of a power system as described in claim 3, characterized in that, S41 includes: using the formula From all the slice inertia assessment results and slice damping assessment results corresponding to the slice observation sets, identify the inertia outliers and the damping outliers, and delete them respectively; wherein, Let m be the training loss of the i-th preset neural network. The preset neural network with the smallest training loss is indexed as m, and its training loss is denoted as m. , This indicates the set threshold for filtering abnormal losses.

5. The online evaluation method for the inertia and damping of a power system as described in claim 3, characterized in that, S41 includes: using the formula From all the slice inertia assessment results and slice damping assessment results corresponding to the slice observation sets, identify the inertia outliers and the damping outliers, and delete them respectively; wherein, and These represent the inertia evaluation result and the damping evaluation result of the i-th slice, respectively. , These represent the mean values ​​of the inertia assessment results and the damping assessment results, respectively. , The standard deviations of the inertia assessment results and the damping assessment results are represented by their respective values. , These represent the set threshold values ​​for inertia anomalies and damping anomalies, respectively.

6. The online evaluation method for the inertia and damping of a power system as described in claim 3, characterized in that, S42 includes: Using formula The remaining slice inertia assessment results are weighted and fused to obtain the target inertia assessment result. ; Using formula The remaining damping evaluation results of each slice are weighted and fused to obtain the target damping evaluation result. ; in, Let i be the training loss of the i-th preset neural network. and These represent the inertia evaluation result and the damping evaluation result of the i-th slice, respectively. , These represent the number of remaining slice inertia assessment results and slice damping assessment results, respectively.

7. The online evaluation method for the inertia and damping of a power system as described in any one of claims 1-6, characterized in that, The operating status parameters include: frequency f and power imbalance P The observation point is represented as { f , P , t }, t Indicates the sampling time.

8. An online evaluation device for the inertia and damping of a power system, characterized in that, The method for performing online evaluation of the inertia and damping of a power system according to any one of claims 1-7 includes: The sampling module is used to take the operating status parameters of the power system collected by the phasor measurement unit at each sampling time within a preset time period and the timestamp corresponding to the sampling time as observation points when the inertial steady-state measurement of the power system is damaged, and then obtain the set of observation points corresponding to the preset time period. The slicing module is used to slice the set of observation points corresponding to the preset time period at preset time intervals to obtain multiple sliced ​​observation sets. An evaluation module is used to input each of the slice observation sets into a preset neural network with embedded physical information, so that the preset neural network outputs the slice inertia evaluation results and slice damping evaluation results for each of the slice observation sets; the physical information includes the rotor motion equations corresponding to the inertia to be measured and the damping to be measured. The fusion module is used to fuse the slice inertia evaluation results and the slice damping evaluation results corresponding to each slice observation set to obtain the target inertia evaluation results and the target damping evaluation results.

9. An online evaluation system for the inertia and damping of a power system, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

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 to 7.

Images