New energy station primary frequency modulation device performance online evaluation method, system and storage medium

Abstract

The application provides a new energy station primary frequency modulation device performance online evaluation method, system and storage medium, the method comprises the following steps: collecting grid-connected point power grid frequency and active power, automatically identifying and screening effective frequency modulation events; calculating multi-dimensional performance indicators; extracting historical performance indicator data with a sliding window, calculating sample mean, standard deviation and other statistics; adaptively constructing a confidence interval according to the sample size and performance indicator distribution characteristics; comparing the confidence interval with the standard threshold, outputting a graded alarm and comprehensively obtaining the current performance evaluation result; analyzing the evolution trend of the confidence interval center value and width, and issuing a degradation or stability warning. The application realizes online non-stop evaluation by using actual power grid disturbance, replaces single determination with statistical inference, and improves the evaluation reliability and predictability.

Description

Technical Field

[0001] This invention relates to the performance evaluation of automatic generation control and grid-connected equipment in power systems, and in particular to an online performance evaluation method, system, and storage medium for primary frequency regulation devices in new energy power plants. Background Technology

[0002] The primary frequency regulation unit is the core control equipment of the grid-connected power generation unit in a new energy power plant. Its function is to participate in the primary regulation of the grid frequency by rapidly adjusting the active power output when the grid frequency deviates from the rated value. It is the first line of defense to ensure the frequency stability of the power system. According to relevant grid connection technical regulations and management rules, the primary frequency regulation unit needs to undergo performance verification regularly to ensure that its performance indicators continuously meet the grid connection requirements.

[0003] Currently, the performance verification of primary frequency regulation devices in new energy power plants mainly adopts the following methods:

[0004] Offline periodic testing method: This method involves applying a simulated frequency step disturbance to the primary frequency regulation device using a frequency signal generator, collecting the device's power response signal, and calculating indicators such as response time, regulation speed, and regulation accuracy to determine whether it meets the standard requirements. This method requires the device to be shut down or taken out of operation, has a long testing cycle, is costly, and cannot reflect the device's true performance in actual power grid operation.

[0005] Online single-event analysis: This method analyzes the response behavior of primary frequency regulation devices by utilizing actual frequency disturbance events that occur in the power grid. Although it can reflect the performance of the device under actual operating conditions, the analysis results of a single event are greatly affected by random factors such as the disturbance amplitude, rate of change, and power grid operation mode, resulting in high randomness and a lack of statistical representativeness.

[0006] Periodic sampling inspection method: In accordance with the requirements of the power grid dispatching department, in-service primary frequency regulation equipment is periodically sampled and tested. This method has a low testing frequency and a long time interval between two tests, making it difficult to detect gradual degradation of equipment performance. There is a risk that performance may have exceeded the standard between two tests without being detected.

[0007] Fixed threshold alarm method: During device operation, its single frequency modulation response index is monitored in real time, and an alarm is issued when the single response index exceeds a preset threshold. This method uses a single event as the basis for judgment and does not consider the influence of measurement noise or random disturbances, resulting in high false alarm and false negative rates, and it cannot distinguish between occasional out-of-tolerance and systemic performance degradation.

[0008] It is evident that existing technologies lack a method to collect long-term online operation data of the grid connection point of the primary frequency regulation device in a new energy power station in real time, and to evaluate whether the performance of the primary frequency regulation device continuously meets the grid connection requirements through statistical inference. Summary of the Invention

[0009] Purpose of the invention: The purpose of this invention is to provide a method, system, and storage medium for online performance evaluation of primary frequency regulation devices in new energy power plants. This method utilizes actual frequency disturbance events in the power grid and replaces single threshold judgment with statistical inference from multiple event samples within a sliding window, thereby enabling online, non-stop quantitative grading evaluation and degradation early warning of primary frequency regulation device performance.

[0010] Technical solution: The online performance evaluation method for the primary frequency regulation device of a new energy power station according to the present invention includes the following steps:

[0011] Collect grid frequency and active power data at the grid connection points of new energy power plants;

[0012] Based on preset conditions, valid verification events are identified from the power grid frequency data, and the power grid frequency curve and active power curve during each valid verification event are extracted as valid event samples.

[0013] Based on valid event samples, multiple performance evaluation metrics are calculated for each valid verification event to obtain performance metric data;

[0014] Use a sliding window to extract performance index data within the most recent preset time window, and calculate the statistics of each performance index within the window;

[0015] Based on the statistics and distribution characteristics of performance indicators, confidence intervals for each performance indicator under a pre-set confidence level are constructed.

[0016] The confidence intervals of each performance index are compared with the preset grid connection standard thresholds to obtain the grading results and grading alarms of each performance index. The current performance evaluation result of the primary frequency regulation device is obtained by combining the grading results of all performance indexes.

[0017] Trend analysis is performed on the time series of the center value and width of the confidence intervals for each performance indicator. When it is determined that the center value of the confidence interval has a continuous monotonic trend or the width of the confidence interval continues to increase, a corresponding early warning message is issued.

[0018] This method continuously collects relevant data from grid-connected points, uses actual frequency disturbance events in the power grid as excitation sources, automatically identifies valid frequency regulation events, and performs statistical evaluation. It achieves online performance evaluation without shutdown or external test signals. A sliding window approach is used to aggregate multiple historical events to construct confidence intervals, ensuring statistical representativeness and quantitative reliability of the evaluation conclusions, overcoming the randomness of single-event analysis and the false alarms and missed alarms of fixed threshold alarms. By comparing confidence intervals with standard thresholds for hierarchical judgment, it can finely distinguish between excellent, qualified, critical, and unqualified performance states, providing differentiated decision-making basis for operators. Simultaneously, trend analysis of the long-term evolution of the confidence interval's center value and width can capture early signals of degradation or stability decline before device performance fails, extending the capability from qualification judgment to predictive maintenance. This helps reduce unplanned outages and improves the safety and economy of grid-connected operation.

[0019] Preferably, valid verification events identified from power grid frequency data based on preset conditions include:

[0020] When the absolute value of the grid frequency deviating from the rated frequency exceeds the preset dead zone, the start time is marked; when the grid frequency falls back into the dead zone and remains stable for a preset time, the end time is marked, and the time between the start and end times is taken as a frequency regulation action event.

[0021] A frequency modulation action event that simultaneously meets the following preset conditions will be considered a valid verification event: the duration of the event is longer than the preset minimum effective duration; the maximum frequency offset during the event is greater than the preset minimum effective disturbance amplitude; the active power output of the device does not reach the upper or lower limit values; and no communication interruption or data loss occurs during the event.

[0022] By using frequency deviation exceeding the dead zone as the starting point and settling back to stability as the event boundary criterion, the frequency regulation actions triggered by actual frequency disturbances in the power grid can be accurately captured. On this basis, further validity screening is carried out using multi-dimensional conditions such as event duration, disturbance depth (maximum frequency offset), power limiting, and data integrity. Invalid events with too short duration, too small disturbance, limited power, or abnormal data can be eliminated, ensuring the quality and consistency of the samples entering the statistical analysis. This provides a high-quality data foundation for the accurate construction of subsequent confidence intervals, thereby improving the reliability of the evaluation conclusions.

[0023] Preferably, the performance evaluation indicators include at least one of response time, adjustment rate, adjustment accuracy, power response index, frequency measurement deviation, adjustment time and reverse adjustment amplitude;

[0024] Response time is the time difference from the moment the frequency exceeds the limit to the moment when the active power of the device begins to change;

[0025] The adjustment rate is the average value of the active power change rate during the period when the active power change rate is the largest within the power adjustment phase.

[0026] The adjustment accuracy is the absolute value of the deviation between the actual output power and the target power after the active power has stabilized;

[0027] The power response index is the ratio of the integral value of the actual contributed active power to the theoretically required contribution value.

[0028] The frequency measurement deviation is the absolute value of the deviation between the device's frequency measurement module and the measured value from the standard frequency source.

[0029] The settling time is the time required for the power to start changing and for the system to enter a stable error band.

[0030] The reverse adjustment amplitude is the amplitude by which the power response exceeds the target value and then reverts back.

[0031] Extracting these multi-dimensional indicators allows for a comprehensive characterization of the performance of a primary frequency modulation device from multiple perspectives, avoiding the one-sidedness of evaluation based on a single indicator. Furthermore, the multi-indicator vector provides rich and quantifiable performance characteristics for subsequent statistical analysis and confidence level assessment, resulting in a more comprehensive and objective evaluation.

[0032] As a preferred approach, the statistics for each performance indicator within the window include the sample mean, sample standard deviation, and sample size.

[0033] As a preferred method, when constructing the confidence interval, one of the following methods should be selected based on the sample size n of the current window and the distribution characteristics of the performance index:

[0034] When n≥30 and the JB statistic of the performance index meets the preset requirements, the normal distribution approximation method is used to construct the confidence interval: ,in, The sample mean. The standard deviation of the sample is 1. The upper α / 2 quantile of the standard normal distribution, where α is the pre-set significance level;

[0035] When n < 30 and the JB statistic of the performance index meets the preset requirements, the t-distribution method is used to construct the confidence interval: ,in The upper α / 2 quantile of a t-distribution with n-1 degrees of freedom;

[0036] When the distribution of performance metrics is unknown or deviates from the normal distribution, the Bootstrap method is used to construct it: repeatedly sample with replacement from the n samples in the current window, calculate the sample mean each time, and obtain multiple Bootstrap means. Take the α / 2 quantile and (1-α / 2) quantile of the multiple means as the lower limit and upper limit of the confidence interval, respectively.

[0037] Different methods are used to construct confidence intervals based on the size of the sample data and the actual distribution, enabling the acquisition of confidence intervals with reasonable coverage probabilities under various data conditions. For large samples, the normal approximation is simple and efficient; for small samples, the t-distribution is used, employing a more conservative and wider interval to compensate for the estimation unreliability caused by insufficient sample size, ensuring that the interval estimate reaches the preset nominal confidence level and avoiding overly optimistic judgments. The Bootstrap method makes no assumptions about the population distribution and provides robust interval estimates for any distribution. This adaptive mechanism guarantees the statistical effectiveness and robustness of the evaluation system under complex and variable field data conditions.

[0038] Preferably, the confidence intervals of each performance indicator are compared with preset grid connection standard thresholds to obtain the grading results and grading alarms for each performance indicator, specifically including:

[0039] For indicators where smaller values ​​are considered better, if the upper limit of the confidence interval is less than the first preset percentage P1 of the standard threshold L, it is judged as Level I Excellent; if the upper limit of the confidence interval is greater than or equal to the percentage of P1 of L and less than L, it is judged as Level II Qualified and a warning alarm is issued; if the lower limit of the confidence interval is less than L and the upper limit of the confidence interval is greater than L, it is judged as Level III Critical and a warning alarm is issued; if the lower limit of the confidence interval is greater than or equal to L, it is judged as Level IV Unqualified and a severe alarm is issued; where P1 is less than 1.

[0040] For indicators where a larger value is better, if the lower limit of the confidence interval is greater than the second preset percentage P2 of L, it is judged as Level I Excellent; if the lower limit of the confidence interval is greater than or equal to L and less than or equal to the percentage P2 of L, it is judged as Level II Qualified and a warning alarm is issued; if the lower limit is less than L and the upper limit is greater than or equal to L, it is judged as Level III Critical and a warning alarm is issued; if the upper limit is less than L, it is judged as Level IV Unqualified and a severe alarm is issued; where P2 is greater than 1.

[0041] By comparing the upper or lower limit of the confidence interval with a standard threshold and its preset percentage value, the performance status can be subdivided into four levels: excellent, acceptable but with insufficient margin, critical uncertainty, and obviously unacceptable, and corresponding alarms of different severity are issued. This grading mechanism provides richer information on device health status than traditional fixed-threshold single alarms, and can accurately guide the formulation of differentiated operation and maintenance strategies. For example, it can strengthen monitoring of critical states, arrange maintenance in advance for states with insufficient margin, effectively reduce the interference of false alarms and missed alarms on operation and maintenance decisions, and improve the utilization efficiency of operation and maintenance resources.

[0042] As a preferred option, the current performance evaluation result of the primary frequency modulation device is obtained by taking the lowest grade from the classification results of all performance indicators.

[0043] As a preferred approach, trend analysis includes:

[0044] Linear regression fitting is performed on the sequence of center values ​​of confidence intervals for each performance index. If the regression slope maintains the same sign within M consecutive windows and the absolute value exceeds the preset sensitivity threshold, it is determined that there is a continuous monotonic trend, where M≥3.

[0045] When the central value of the confidence interval of a performance indicator that is better the smaller the value, the more monotonically increasing it is, or when the central value of the confidence interval of a performance indicator that is better the larger the value, the more monotonically decreasing it is, a performance degradation warning is issued.

[0046] When the confidence interval width shows a continuous increasing trend, a warning of declining stability is issued.

[0047] By using linear regression to monitor the long-term evolution of the confidence interval's center value and width, trends of slow degradation or deteriorating output stability can be detected before device performance indicators exceed standard thresholds, allowing for early warnings. The center value trend reflects a directional shift in the device's average performance, while the width trend reflects changes in output consistency. This trend analysis provides maintenance personnel with early signals of potential device degradation, enabling proactive planning of maintenance activities. It transforms traditional planned maintenance or post-failure maintenance into condition-based predictive maintenance, helping to avoid unplanned downtime due to sudden performance degradation and improving equipment utilization and grid security.

[0048] The online automatic performance verification system for the primary frequency regulation device of a new energy power station according to the present invention includes:

[0049] The data acquisition module is used to collect grid frequency and active power data at the grid connection point of new energy power plants;

[0050] The event identification and filtering module is used to identify primary frequency regulation action events based on power grid frequency data and filter valid verification events;

[0051] The metrics calculation module is used to calculate multiple performance evaluation metrics for each valid verification event.

[0052] The historical database module is used to store performance metric data for each valid verification event.

[0053] The sliding window sampling module is used to extract the performance index data of the current window from the historical database module with a preset time window and sliding step size.

[0054] The statistics calculation module is used to calculate the statistics of each performance indicator within the current window;

[0055] The confidence interval construction module is used to construct confidence intervals for each performance indicator at a preset confidence level based on the statistics and distribution characteristics of the performance indicators.

[0056] The judgment and alarm module is used to compare the confidence interval with the grid connection standard threshold and output the grading result and grading alarm information according to the preset grading rules.

[0057] The trend analysis module is used to analyze the time evolution trend of the confidence interval center value and interval width, and output performance degradation warning and stability decline warning.

[0058] The computer storage medium of the present invention stores computer instructions, which, when invoked, are used to execute any of the methods described above.

[0059] Beneficial Effects: By collecting long-term online operation data from grid-connected power plants, this method automatically identifies valid verification events and extracts multi-dimensional performance indicators. Recent historical events are aggregated using a sliding window approach to form an analysis sample. Based on this, the statistical measures of the sample are calculated, and confidence intervals for each performance indicator are constructed at a pre-set confidence level according to the sample size and indicator distribution characteristics. These confidence intervals are then compared with grid-connected standard thresholds to provide tiered alarms and comprehensively derive the current performance evaluation results of the device. Simultaneously, the time series of the center value and width of each window's confidence interval are monitored for trends, and corresponding warnings are issued based on the monitoring results. This method requires no downtime, utilizes natural grid disturbances for verification, overcomes the randomness of single-event analysis, and replaces single-threshold judgments with statistical inference. It provides a quantitative and tiered assessment of the device's current performance status and possesses the predictive and early warning capability to capture degradation trends before performance exceeds standards, which is of positive significance for improving the safety and economy of grid-connected operation of new energy power plants. Attached Figure Description

[0060] Figure 1 This is an overall flowchart of the method of the present invention;

[0061] Figure 2 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0062] As shown in the figure, the online performance evaluation method for the primary frequency regulation device of a new energy power station according to the present invention includes the following process:

[0063] Data acquisition and primary frequency modulation event identification:

[0064] The analog data of the grid connection point of the new energy power station is continuously collected by the data acquisition device at a sampling frequency of not less than 100Hz, including the measured value of grid frequency f(t) and active power output value P(t).

[0065] Based on the collected power grid frequency data and the manually set power grid rated frequency f NParameters such as the frequency dead zone setting value D are used to automatically identify primary frequency modulation (FM) action events. The start time t of the primary FM action event is marked when the following conditions are met. start The absolute value of the deviation of the grid frequency f(t) from the rated frequency f_N exceeds the preset primary frequency regulation dead zone D. When the grid frequency f(t) falls back to the range [f_N-D, f_N+D] and remains stable for a preset time T_stable, the end time t_end of this event is marked. The event at the start and end times constitutes a valid primary frequency regulation action time. The frequency curve f(t) and active power curve P(t) from t_start to t_end can be extracted as a sample of valid primary frequency regulation action events, or they can be extracted after valid verification events are identified to save workload.

[0066] For each identified frequency modulation action event, it is further determined whether it is a valid verification event. The determination conditions include (all conditions must be met): the duration of the event is longer than the preset minimum valid duration; the maximum frequency offset during the event is greater than the preset minimum valid disturbance amplitude; the active power output of the device does not reach the upper or lower limit amplitude; and no communication interruption or data loss occurs during the event.

[0067] Validally verified event information is stored in the historical event database.

[0068] Performance index calculation:

[0069] For each valid verification event in the historical event database, calculate the multi-dimensional performance evaluation indicators of the frequency modulation device, including but not limited to:

[0070] Response time From the moment when the frequency exceeds the limit When the active power of the device begins to change The time difference, i.e. ;

[0071] Adjustment rate The change in active power per unit time during the power regulation phase, i.e. , This refers to the active power conversion quantity. This represents the time change corresponding to the change in active power.

[0072] Adjustment accuracy After the active power stabilizes, the actual output power With target power The absolute value of the deviation, i.e. ;

[0073] Power Response Index The ratio of the actual contributed active power integral value to the theoretically required contribution value, i.e. , dimensionless, of which This represents the actual active power. Theoretical active power;

[0074] Frequency measurement deviation : The absolute value of the deviation between the measured value of the frequency measurement module inside the primary frequency modulation device and the measured value of the standard frequency source;

[0075] Adjusting time From the start of power variation to the point where power enters a stable error band (typically...) The time required to reach the rated power;

[0076] Reverse adjustment amplitude The magnitude of the pullback after exceeding the target value during the power response process.

[0077] The performance metric vector calculated for each valid verification event is stored in the historical performance database.

[0078] Sliding window sampling and statistic calculation:

[0079] Set the time window length (Value range is 7 to 180 days, adjustable) and sliding step size (Values ​​range from 1 to 7 days, adjustable), extracting the most recent data from the historical performance database using a sliding window method. Performance metrics data for all valid verification events within a given time period.

[0080] For each performance metric within the current sliding window, calculate its statistics, including:

[0081] Sample mean : ;

[0082] Sample standard deviation : ;

[0083] Sample size That is, the total number of valid events within the current window. For the first The performance metric value for this event.

[0084] Confidence interval construction:

[0085] Based on the statistics of the current sliding window, construct confidence levels for each performance metric. The confidence interval, where significance level ( Typically, a value of 0.05 or 0.01 is used, corresponding to 95% and 99% confidence levels, respectively.

[0086] Construction method based on sample size Choose one of the following three methods based on the distribution characteristics of performance indicators:

[0087] Method A (Large Sample Normal Approximation): When When the performance index follows a normal distribution or approximately a normal distribution, using the normal distribution approximation, the confidence interval for the overall mean of the performance index is:

[0088] ,

[0089] in, The upper part of the standard normal distribution quantiles.

[0090] Method B (small sample t-distribution method): when When the performance index follows a normal or approximately normal distribution, the confidence interval can be constructed using the t-distribution.

[0091] ,

[0092] in For degrees of freedom The upper part of the t-distribution quantiles.

[0093] Method C (Bootstrap Method): When the distribution of performance metrics is unknown or significantly deviates from a normal distribution, the Bootstrap method is used to construct confidence intervals: starting from the current window... Repeated sampling with replacement was performed on each sample. Second-rate( , (where the integer is positive), the sample mean is calculated each time, resulting in... The average of the Bootstrap values ​​is taken. quantiles and Quantiles serve as the lower and upper limits of the confidence interval.

[0094] The method for determining whether the performance index distribution follows a normal or approximately normal distribution is as follows:

[0095] Suppose there are n performance index samples x1, x2, ..., x3 within the current sliding window. 2, ...,x n Its mean is ;

[0096] Calculate its skewness: ;

[0097] Then calculate its kurtosis: ;

[0098] Finally, calculate the JB statistic: ;

[0099] When JB≤5.991, it follows a normal or approximately normal distribution; when JB>5.991, it does not follow an approximately normal distribution.

[0100] Step S5: Capability Assessment and Graded Alarm

[0101] The confidence interval of each performance indicator constructed in step S4 is compared with the preset grid connection standard threshold. The comparison is performed, and the classification judgment and classification alarm are output based on the comparison results.

[0102] For metrics where smaller values ​​are always better (including response time) Adjustment accuracy Frequency measurement deviation Adjusting time and reverse adjustment amplitude ):

[0103] If the upper limit of the confidence interval is less than the standard threshold First preset percentage ( Its value ranges from 60% to 90%, with 80% being the normal choice, meaning the upper limit of the confidence interval is less than... The device is rated as Level I (Excellent), indicating that its capabilities in this dimension are sufficient and stable.

[0104] If the upper limit of the confidence interval is greater than or equal to and less than The device is classified as Level II (qualified), indicating that its capability in this dimension is qualified but its margin is insufficient, and a warning level alarm is issued.

[0105] If the lower limit of the confidence interval is less than And the upper limit is greater than The device is classified as Level III (critical), indicating that there is uncertainty in its performance in this dimension, and a warning-level alarm is issued.

[0106] If the lower limit of the confidence interval is greater than or equal to The device is classified as Level IV (unqualified), indicating that its capacity in this dimension no longer meets the grid connection requirements, and a severe alarm is issued.

[0107] For indicators where a larger value is always better (including the power response index) and adjustment rate ):

[0108] If the lower limit of the confidence interval is greater than Second preset percentage ( Its value ranges from 110% to 140%, with 120% being the normal choice), meaning the lower limit of the confidence interval is greater than... It is rated as Grade I (Excellent).

[0109] If the lower limit of the confidence interval is greater than or equal to and less than or equal to The system is classified as Level II (qualified), and a warning level alert is issued.

[0110] If the lower limit of the confidence interval is less than And the upper limit is greater than or equal to The system is classified as Level III (critical) and a warning alarm is issued.

[0111] If the upper limit of the confidence interval is less than The result is classified as Level IV (unacceptable), and a critical alarm is issued.

[0112] Based on the combined results of all performance indicators, the lowest-level method can be used to output the current performance evaluation result of the primary frequency modulation device.

[0113] Trend analysis and predictive maintenance alerts:

[0114] Record the center value and width of the confidence interval for each performance index at each sliding window moment to form a time evolution sequence of the confidence interval.

[0115] When the center value of the confidence interval of a certain performance index shows a continuous monotonically increasing trend (for indicators where smaller values ​​are better) or a continuous monotonically decreasing trend (for indicators where larger values ​​are better), a performance degradation warning is issued, indicating that the capability of the primary frequency regulation device may be deteriorating, and it is recommended to arrange parameter optimization or maintenance.

[0116] The method for determining the performance degradation trend is as follows: perform linear regression fitting on the confidence interval center value sequence; if the regression slope is within a continuous range... windows ( If the values ​​within a range remain the same and their absolute values ​​exceed a preset sensitivity threshold, then it is determined that there is a continuous monotonic trend.

[0117] When the confidence interval width of a certain performance indicator shows a continuous increasing trend, a stability decline warning is issued, indicating that the inverter output characteristics of the station where the primary frequency regulation device is located are fluctuating more sharply, and there may be abnormalities such as aging of internal components or parameter drift.

[0118] The online performance evaluation system for primary frequency regulation devices in new energy power plants according to the present invention includes: a data acquisition module, an event identification and screening module, an index calculation module, a historical database module, a sliding window sampling module, a statistical calculation module, a confidence interval construction module, a judgment and alarm module, and a trend analysis module.

[0119] The data acquisition module is used to continuously acquire the grid frequency at the grid connection point of new energy power plants at a sampling frequency of not less than 100Hz. and active power data.

[0120] The event identification and filtering module is connected to the data acquisition module. It automatically identifies primary frequency regulation action events based on power grid frequency data and filters valid verification events according to preset validity conditions.

[0121] The indicator calculation module is connected to the event identification and filtering module to calculate the response time for each valid verification event. Adjustment rate Adjustment accuracy Power response index Frequency measurement deviation Adjusting time and reverse adjustment amplitude Performance metrics are measured and the results are stored in the historical database module.

[0122] The sliding window sampling module extracts data from the historical database module within a preset time window. and sliding step size Extract performance metrics data from the current window.

[0123] The statistics calculation module calculates the sample mean of each indicator. Sample standard deviation and sample size .

[0124] The confidence interval construction module is based on the sample size. Based on the adaptive selection of the distribution characteristics of the indicators, the normal approximation method, the t-distribution method, or the bootstrap method are used to construct the confidence interval of each performance indicator under the preset confidence level.

[0125] The judgment and alarm module will compare the confidence interval with the grid connection standard threshold. The system compares the results and outputs the classification judgment results according to the preset judgment rules of levels I to IV, and issues corresponding attention level, warning level or serious level alarms. At the same time, it obtains the current comprehensive performance evaluation result of the device through the weighted comprehensive scoring method or the lowest level method.

[0126] The trend analysis module analyzes the time evolution trend of the confidence interval center value and width, and outputs performance degradation warnings or stability decline warnings based on the monitoring results.

[0127] In addition, the system may include an output and visualization module for outputting verification reports and alarm information, and for visualizing confidence intervals and trend curves.

Claims

1. A method for online performance evaluation of primary frequency regulation devices in new energy power plants, characterized in that, Includes the following steps: Collect grid frequency and active power data at the grid connection points of new energy power plants; Based on preset conditions, valid verification events are identified from the power grid frequency data, and the power grid frequency curve and active power curve during each valid verification event are extracted as valid event samples. Based on valid event samples, multiple performance evaluation metrics are calculated for each valid verification event to obtain performance metric data; Use a sliding window to extract performance index data within the most recent preset time window, and calculate the statistics of each performance index within the window; Based on the statistics and distribution characteristics of performance indicators, confidence intervals for each performance indicator under a pre-set confidence level are constructed. The confidence intervals of each performance index are compared with the preset grid connection standard thresholds to obtain the grading results and grading alarms of each performance index. The current performance evaluation result of the primary frequency regulation device is obtained by combining the grading results of all performance indexes. Trend analysis is performed on the time series of the center value and width of the confidence intervals for each performance indicator. When it is determined that the center value of the confidence interval has a continuous monotonic trend or the width of the confidence interval continues to increase, a corresponding early warning message is issued.

2. The method according to claim 1, characterized in that, Valid verification events identified from power grid frequency data based on preset conditions include: When the absolute value of the grid frequency deviating from the rated frequency exceeds the preset dead zone, the start time is marked; when the grid frequency falls back into the dead zone and remains stable for a preset time, the end time is marked, and the time between the start and end times is taken as a frequency regulation action event. A frequency modulation action event that simultaneously meets the following preset conditions will be considered a valid verification event: the duration of the event is longer than the preset minimum effective duration; the maximum frequency offset during the event is greater than the preset minimum effective disturbance amplitude; the active power output of the device does not reach the upper or lower limit values; and no communication interruption or data loss occurs during the event.

3. The method according to claim 1, characterized in that, The performance evaluation indicators include at least one of the following: response time, adjustment rate, adjustment accuracy, power response index, frequency measurement deviation, adjustment time, and reverse adjustment amplitude. Response time is the time difference from the moment the frequency exceeds the limit to the moment when the active power of the device begins to change; The adjustment rate is the change in active power per unit time during the power adjustment phase. The adjustment accuracy is the absolute value of the deviation between the actual output power and the target power after the active power has stabilized; The power response index is the ratio of the integral value of the actual active power contribution to the theoretical contribution value. The frequency measurement deviation is the absolute value of the deviation between the device's frequency measurement module and the measurement value from the standard frequency source. The settling time is the time required for the power to start changing and for the system to enter a stable error band. The reverse adjustment amplitude is the amplitude by which the power response exceeds the target value and then reverts back.

4. The method according to claim 1, characterized in that, The statistics for each performance indicator within the window include the sample mean, sample standard deviation, and sample size.

5. The method according to claim 1, characterized in that, When constructing confidence intervals, choose one of the following methods based on the current window's sample size n and the performance index distribution characteristics: When n≥30 and the JB statistic of the performance index meets the preset requirements, the normal distribution approximation method is used to construct the confidence interval: ,in, The sample mean. The standard deviation of the sample is 1. The upper α / 2 quantile of the standard normal distribution, where α is the pre-set significance level; When n < 30 and the JB statistic of the performance index meets the preset requirements, the t-distribution method is used to construct the confidence interval: ,in The upper α / 2 quantile of a t-distribution with n-1 degrees of freedom; When the distribution of performance metrics is unknown or deviates from the normal distribution, the Bootstrap method is used to construct it: repeatedly sample with replacement from the n samples in the current window, calculate the sample mean each time, and obtain multiple Bootstrap means. Take the α / 2 quantile and (1-α / 2) quantile of the multiple means as the lower limit and upper limit of the confidence interval, respectively.

6. The method according to claim 1, characterized in that, The confidence intervals of each performance indicator are compared with the preset grid connection standard thresholds to obtain the classification results and classification alarms for each performance indicator, specifically including: For indicators where smaller values ​​are considered better, if the upper limit of the confidence interval is less than the first preset percentage P1 of the standard threshold L, it is judged as Level I Excellent; if the upper limit of the confidence interval is greater than or equal to the percentage of P1 of L and less than L, it is judged as Level II Qualified and a warning alarm is issued; if the lower limit of the confidence interval is less than L and the upper limit of the confidence interval is greater than L, it is judged as Level III Critical and a warning alarm is issued; if the lower limit of the confidence interval is greater than or equal to L, it is judged as Level IV Unqualified and a severe alarm is issued; where P1 is less than 1. For indicators where a larger value is better, if the lower limit of the confidence interval is greater than the second preset percentage P2 of L, it is judged as Level I Excellent; if the lower limit of the confidence interval is greater than or equal to L and less than or equal to the percentage P2 of L, it is judged as Level II Qualified and a warning alarm is issued; if the lower limit is less than L and the upper limit is greater than or equal to L, it is judged as Level III Critical and a warning alarm is issued; if the upper limit is less than L, it is judged as Level IV Unqualified and a severe alarm is issued; where P2 is greater than 1.

7. The method according to claim 1, characterized in that, The current performance evaluation results of the primary frequency modulation device are derived by taking the lowest grade from the classification results of all performance indicators.

8. The method according to claim 1, characterized in that, Trend analysis includes: Linear regression fitting is performed on the sequence of center values ​​of confidence intervals for each performance index. If the regression slope maintains the same sign within M consecutive windows and the absolute value exceeds the preset sensitivity threshold, it is determined that there is a continuous monotonic trend, where M≥3. When the central value of the confidence interval of a performance indicator that is better the smaller the value, the more monotonically increasing it is, or when the central value of the confidence interval of a performance indicator that is better the larger the value, the more monotonically decreasing it is, a performance degradation warning is issued. When the confidence interval width shows a continuous increasing trend, a warning of declining stability is issued.

9. An online automatic performance verification system for primary frequency regulation devices in new energy power plants, characterized in that, include: The data acquisition module is used to collect grid frequency and active power data at the grid connection point of new energy power plants; The event identification and filtering module is used to identify primary frequency regulation action events based on power grid frequency data and filter valid verification events; The metrics calculation module is used to calculate multiple performance evaluation metrics for each valid verification event. The historical database module is used to store performance metric data for each valid verification event. The sliding window sampling module is used to extract the performance index data of the current window from the historical database module with a preset time window and sliding step size. The statistics calculation module is used to calculate the statistics of each performance indicator within the current window; The confidence interval construction module is used to construct confidence intervals for each performance indicator at a preset confidence level based on the statistics and distribution characteristics of the performance indicators. The judgment and alarm module is used to compare the confidence interval with the grid connection standard threshold and output the grading result and grading alarm information according to the preset grading rules. The trend analysis module is used to analyze the time evolution trend of the confidence interval center value and interval width, and output performance degradation warning and stability decline warning.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the method as described in any one of claims 1-8.

Images