Generator set AGC regulation performance evaluation method and system

Abstract

The invention relates to the technical field of industrial data analysis, and discloses a generator set AGC regulation performance evaluation method and system, and the method comprises the steps: obtaining an automatic power generation control instruction sequence and an actual power sequence of a generator set, carrying out the processing, obtaining a regulation time lag, and carrying out the phase alignment processing of the actual power sequence; generating a reference response curve and a dynamic residual sequence based on a phase alignment result and a preset generator set physical inertia time constant; carrying out feature analysis on the dynamic residual sequence, identifying an effective adjustment interval and calculating a symbol flipping frequency; correcting the dynamic residual error sequence based on the symbol flipping frequency and a preset compensation factor to obtain a corrected residual error sequence; and carrying out statistics on characteristic parameters of the corrected residual error sequence, carrying out calculation to obtain an adjustment efficiency evaluation index of the generator set in a preset sampling period, and evaluating the AGC adjustment performance of the generator set according to the adjustment efficiency evaluation index. The power grid frequency modulation stability and the power supply reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of industrial data analysis technology, specifically to a method and system for evaluating the AGC regulation performance of generator sets. Background Technology

[0002] With the high proportion of renewable energy connected to the grid, the frequency fluctuation characteristics of the power system exhibit a multi-source, wide-frequency distribution. The evaluation accuracy of automatic generation control and regulation of generator units is directly related to the fairness of ancillary service market settlement. Existing technologies use the arithmetic mean of regulation error or the linear slope of response time to measure regulation quality. The actual power generation sequence includes background random fluctuations composed of primary frequency regulation actions, combustion disturbances, and measurement noise. When the grid command enters the small-amplitude, frequent regulation range, the background fluctuations and the controlled regulation response are highly coupled in the time domain, forming statistical aliasing. Traditional linear evaluation indicators cannot distinguish between the actual regulation contribution of the unit and the random power components, resulting in biased evaluation conclusions under conditions with high background noise.

[0003] Even if hardware constraints such as generator inertial hysteresis and rate limitations are identified at the physical level, it remains difficult to achieve fair performance measurement at the software logic level without a matching, refined evaluation algorithm. For example, Chinese invention patent CN110930060B discloses a generator AGC regulation performance evaluation method based on parameter calculation. This method identifies the start and end of the regulation sub-process by setting judgment factors and calculates response time and deviation indicators. However, this method is based on static parameter statistics with fixed threshold breakpoints. Because it does not explore the internal physical inertial characteristics of the generator's thermal system, it is ill-suited for complex broadband fluctuations. In environmental conditions, it is difficult to decouple the controlled trend from the background noise statistically. Due to the lack of a mechanism to identify the characteristics of the nonlinear dead zone and physical saturation of the actuator, it is easy to misjudge the one-way deviation caused by physical boundary constraints as a deterioration in control accuracy. The uncertainty of the statistical benchmark leads to evaluation distortion, which restricts the execution of refined scheduling instructions and performance accounting. It is difficult to separate the nonlinear dead zone, rate limit and physical saturation internal constraints of the actuator by simply relying on low-pass filtering. The inevitable delay caused by physical links is misjudged as deviation of the adjustment intention. The interference of the signal source is included in the performance accounting, so that the evaluation results cannot reflect the true response of the unit under the physical limit.

[0004] Therefore, how to construct a reference response envelope that conforms to the physical inertial characteristics of the unit, and how to achieve a fair measurement of the non-stationary environment regulation efficiency through residual higher-order moment characteristics and correlation stability indicators, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] In order to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for evaluating the AGC regulation performance of generator sets, so as to solve the technical problem of how to construct a reference response envelope that conforms to the physical inertial characteristics of the generator set.

[0006] This invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for evaluating the AGC (Automatic Generative Control) regulation performance of a generator set, comprising: Step 1: Obtain the automatic power generation control command sequence and actual power generation sequence of the generator set, process them to obtain the adjustment time delay, and perform phase alignment processing on the actual power generation sequence; Step 2: Based on the phase alignment results and the preset physical inertia time constant of the generator set, generate the reference response curve and dynamic residual sequence; Step 3: Perform feature analysis on the dynamic residual sequence to identify the effective adjustment interval and calculate the sign flip frequency; Step 4: Correct the dynamic residual sequence based on the symbol flipping frequency and the preset compensation factor to obtain the corrected residual sequence; Step 5: Statistically correct the characteristic parameters of the residual sequence, calculate the regulation performance evaluation index of the generator set within the preset sampling period, and evaluate the AGC regulation performance of the generator set based on the regulation performance evaluation index.

[0007] Preferably, in step 1, the process for determining the adjustment time delay is as follows: The actual power generation sequence is shifted point by point within the preset time offset range, and the cross-relationship value between the shifted sequence and the automatic generation control command sequence is calculated; the shift time corresponding to the maximum value of the cross-relationship value is determined as the adjustment time lag of the generator set. Step 1 is followed by the following steps: establishing multiple sliding statistical windows within a preset sampling period, and calculating the maximum cross-correlation coefficient between the automatic power generation control command sequence and the actual power generation sequence within each sliding statistical window; extracting the maximum cross-correlation coefficients of all sliding statistical windows, and calculating the coefficient of variation of the maximum cross-correlation coefficients among all sliding statistical windows to generate a correlation stability index; determining the confidence weights of the regulation efficiency evaluation index based on the correlation stability index, and using the confidence weights to perform product correction on the regulation efficiency evaluation index.

[0008] Preferably, step 2, generating the reference response curve, includes the following steps: acquiring real-time load data of the generator set; matching the corresponding physical inertia time constant of the generator set from a preset parameter mapping table according to the preset interval where the real-time load data is located; and using the matched physical inertia time constant to perform hysteresis filtering on the automatic generation control command sequence to generate the reference response curve.

[0009] Preferably, in step 3, the formula for calculating the symbol flip frequency is as follows:

[0010] in, This represents the total number of samples within the sliding statistical window. This serves as a sampling time identifier. For first-order difference sequences in The value of the moment. It is a symbolic function; The first-order difference sequence is extracted from the dynamic residual sequence, and the sign flip frequency is the result of the standardization of the sum of the absolute values ​​of the differences between the sign function values ​​of adjacent sampling points in the first-order difference sequence within the sliding statistical window.

[0011] Preferably, the formula for magnitude shrinkage mapping of the dynamic residual sequence in step 4 is as follows:

[0012] in, To correct the residual sequence in The value of the moment. For dynamic residual sequences in The value of the moment. For mechanical dead zone compensation factor, The sign flipping frequency.

[0013] Preferably, before step 5, the following steps are also included: identifying the steady-state sampling segment of the generator set under the numerical constant state of the automatic power generation control command sequence; calculating the root mean square value of the dynamic residual sequence within the steady-state sampling segment to obtain the steady-state noise basis index; using the steady-state noise basis index, performing mapping compensation on the boundary values ​​of the preset value range through a preset monotonically increasing function to generate an adaptive evaluation bandwidth that is adjusted in real time with environmental background fluctuations.

[0014] Preferably, in step 5, the process of calculating the regulation performance evaluation index further includes the following steps: calculating the autocorrelation function sequence of the dynamic residual sequence in the time domain; identifying whether the generator set has controlled oscillations based on the envelope decay rate of the autocorrelation function sequence; if controlled oscillations are identified, extracting the dominant frequency of the autocorrelation function sequence to calculate the regulation loss factor, and performing a punitive correction on the regulation performance evaluation index based on the regulation loss factor.

[0015] Preferably, step 5, which involves statistically analyzing the distribution probability characteristics of the corrected residual sequence, includes the following steps: obtaining the rated capacity of the generator set and setting a percentage error band centered at zero and with a width of 1% of the rated capacity as a preset value range; statistically analyzing the proportion of samples whose corrected residual sequence falls within the preset value range to obtain the distribution probability value.

[0016] Preferably, after step 5, the following steps are also included: calculating the third central moment of the corrected residual sequence within a preset sampling period to obtain the skewness characteristics; identifying whether the generator set is in the actuator saturation condition based on the correlation between the sign of the skewness characteristics and the change direction of the automatic generation control command sequence; if the identification result is that the generator set is in the actuator saturation condition, then reducing the weight of the component related to the central offset in the regulation efficiency evaluation index through the weight correction operator.

[0017] Secondly, the present invention also provides a generator set AGC regulation performance evaluation system, comprising: The data acquisition and time delay processing module is used to acquire the automatic power generation control command sequence and the actual power generation sequence of the generator set, process them to obtain the adjustment time delay, and perform phase alignment processing on the actual power generation sequence. The benchmark generation and residual calculation module is used to generate benchmark response curves and dynamic residual sequences based on phase alignment results and preset generator physical inertia time constants. The frequency calculation and interval identification module is used to perform feature analysis on the dynamic residual sequence, identify the effective adjustment interval, and calculate the symbol flip frequency. The residual correction processing module is used to correct the dynamic residual sequence based on the sign flipping frequency and a preset compensation factor to obtain the corrected residual sequence. The index evaluation and processing module is used to statistically correct the characteristic parameters of the residual sequence, calculate the regulation performance evaluation index of the generator set within the preset sampling period, and evaluate the AGC regulation performance of the generator set based on the regulation performance evaluation index.

[0018] Compared with the prior art, the present invention has the following beneficial technical effects: This invention provides a method for evaluating the AGC (Automatic Generation Control) performance of generator sets. By incorporating the physical inertia time constant of the generator set into the generation process of the reference response curve and dynamically matching corresponding parameters with real-time load data, the reference response curve can accurately match the actual physical operating characteristics of the generator set, constructing a reference response envelope that highly conforms to the inertial characteristics of the generator set. This overcomes the limitations of traditional reference construction methods that ignore differences in generator set inertia and have poor universality. Simultaneously, through a series of processes such as time-delay phase alignment, dead-zone compensation correction, and adaptive adjustment of operating conditions, the interference of factors such as time delay, nonlinear dead zone, and operating condition fluctuations on the comparison between the reference envelope and the actual response is effectively eliminated, ensuring that the evaluation indicators of regulation efficiency can truly reflect the AGC performance of the generator set. Compared with traditional evaluation methods, the reference response envelope constructed by this scheme is more targeted and accurate, and the evaluation results obtained based on this benchmark are more precise. It can provide a reliable basis for the optimization and adjustment of the generator set AGC regulation system, helping to improve the frequency regulation stability and power supply reliability of the power grid, and has significant technical advantages and practical value.

[0019] Furthermore, in the AGC regulation performance of the generator set, the controlled regulation response and random environmental fluctuations are statistically decoupled. A reference response envelope that conforms to the physical inertial characteristics of the generator set is constructed. The dynamic residual sequence of the actual generated power relative to this envelope is extracted. The regulation trend is separated from the aliased background noise. The probability density characteristics and kurtosis characteristics of the residual distribution are used to characterize the regulation consistency. Frequency fluctuations or measurement noise interference are suppressed. The evaluation index reflects the quality of the generator set in following the command and improves the objectivity and fairness of industrial data statistics in complex environments.

[0020] Furthermore, a correlation mapping between statistical distribution characteristics and physical boundary constraints is established. The skewness characteristics of the residual sequence are used to identify the saturation state or dead zone action of the actuator. When the residual distribution is asymmetric, the system determines whether the unit is in a state of physical output limit or nonlinear constraint based on the correlation between the skewness sign and the command direction. Based on the adaptive capability of the evaluation system, the system distinguishes between the unit's subjective adjustment failure and objective physical limitations, avoiding misjudging the unidirectional deviation of the actuator caused by physical walls as a deterioration in control accuracy. A source quality defense system based on correlation stability is constructed. The credibility of the data is quantified by the maximum cross-correlation envelope of the command sequence and response sequence in the statistical space. The dispersion of the correlation peak within the statistical sliding window is used to audit whether there are non-physical disturbances or signal drift in the data source. Logical self-consistency checks are used to automatically suppress interfered samples, ensuring that the final evaluation conclusion has the authenticity of industrial audit level.

[0021] Furthermore, this study identifies implicit controlled oscillations and their structural impact on regulation efficiency, mines the decay characteristics of the autocorrelation function of the residual sequence, identifies cyclical actions or regulation oscillations within the power sequence, regulates quality from the perspective of information flow orderliness, captures behaviors that sacrifice equipment lifespan for performance indicators, and corrects comprehensive efficiency by generating regulation loss factors to guide generator units to consider the operational health of actuators. It also eliminates evaluation standard drift caused by non-stationary background noise, introduces a real-time inversion mechanism based on steady-state background noise, adjusts the evaluation bandwidth according to the environmental noise floor energy density, identifies the fluctuation intensity within the unit's steady-state window, and performs nonlinear rescaling on the statistical evaluation confidence boundary to dynamically align the evaluation criteria with the intensity of environmental disturbances. This ensures that during periods of high renewable energy grid connection, the evaluation conclusions still focus on the unit's own regulation contribution, enhancing the technical resilience of statistical evaluation methods in non-stationary industrial environments. Attached Figure Description

[0022] Figure 1 This is a flowchart of the generator set AGC regulation performance evaluation method of the present invention; Figure 2 This is a schematic diagram of the three-layer logic architecture and data interaction of the generator set AGC regulation performance evaluation system of the present invention; Figure 3 This is a comparison chart of the anti-interference stability of the coefficient of variation of the evaluation index of this invention under different background noise intensities; Figure 4 This is a multidimensional influencing factor decomposition diagram of the accuracy of AGC regulation performance evaluation of the fishbone model of the present invention; Figure 5 This is a schematic diagram of the generator set AGC regulation performance evaluation system of the present invention; In the diagram: 1. Data acquisition and time delay processing module; 2. Benchmark generation and residual calculation module; 3. Frequency calculation and interval identification module; 4. Residual correction processing module; 5. Index evaluation processing module. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] The purpose of this invention is to provide a method and system for evaluating the AGC regulation performance of generator sets, so as to solve the technical problem of how to construct a reference response envelope that conforms to the physical inertial characteristics of the generator set.

[0026] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1 In one embodiment of the present invention, a method for evaluating the AGC regulation performance of a generator set is provided, comprising: Step 1: Obtain the automatic power generation control command sequence and actual power generation sequence of the generator set, process them to obtain the adjustment time delay, and perform phase alignment processing on the actual power generation sequence; Specifically, in step 1, the process for determining the adjustment time delay is as follows: The actual power generation sequence is shifted point by point within the preset time offset range, and the cross-relationship value between the shifted sequence and the automatic generation control command sequence is calculated; the shift time corresponding to the maximum value of the cross-relationship value is determined as the adjustment time lag of the generator set. Step 1 is followed by the following steps: establishing multiple sliding statistical windows within a preset sampling period, and calculating the maximum cross-correlation coefficient between the automatic power generation control command sequence and the actual power generation sequence within each sliding statistical window; extracting the maximum cross-correlation coefficients of all sliding statistical windows, and calculating the coefficient of variation of the maximum cross-correlation coefficients among all sliding statistical windows to generate a correlation stability index; determining the confidence weights of the regulation efficiency evaluation index based on the correlation stability index, and using the confidence weights to perform product correction on the regulation efficiency evaluation index.

[0027] Step 2: Based on the phase alignment results and the preset physical inertia time constant of the generator set, generate the reference response curve and dynamic residual sequence; Specifically, step 2, generating the reference response curve, includes the following steps: acquiring real-time load data of the generator set; matching the corresponding physical inertia time constant of the generator set from a preset parameter mapping table based on the preset interval of the real-time load data; and using the matched physical inertia time constant to perform hysteresis filtering on the automatic generation control command sequence to generate the reference response curve.

[0028] Step 3: Perform feature analysis on the dynamic residual sequence to identify the effective adjustment interval and calculate the sign flip frequency; Specifically, in step 3, the formula for calculating the symbol flip frequency is as follows:

[0029] in, This represents the total number of samples within the sliding statistical window. This serves as a sampling time identifier. For first-order difference sequences in The value of the moment. It is a symbolic function; The first-order difference sequence is extracted from the dynamic residual sequence, and the sign flip frequency is the result of the standardization of the sum of the absolute values ​​of the differences between the sign function values ​​of adjacent sampling points in the first-order difference sequence within the sliding statistical window.

[0030] Step 4: Correct the dynamic residual sequence based on the symbol flipping frequency and the preset compensation factor to obtain the corrected residual sequence; Specifically, the formula for magnitude shrinkage mapping of the dynamic residual sequence in step 4 is as follows:

[0031] in, To correct the residual sequence in The value of the moment. For dynamic residual sequences in The value of the moment. For mechanical dead zone compensation factor, The sign flipping frequency.

[0032] Step 5: Statistically correct the characteristic parameters of the residual sequence, calculate the regulation performance evaluation index of the generator set within the preset sampling period, and evaluate the AGC regulation performance of the generator set based on the regulation performance evaluation index.

[0033] Specifically, before step 5, the following steps are also included: identifying the steady-state sampling segment of the generator set under the numerical constant state of the automatic generation control command sequence; calculating the root mean square value of the dynamic residual sequence within the steady-state sampling segment to obtain the steady-state noise basis index; using the steady-state noise basis index, performing mapping compensation on the boundary values ​​of the preset value range through a preset monotonically increasing function to generate an adaptive evaluation bandwidth that is adjusted in real time with environmental background fluctuations.

[0034] Specifically, in step 5, the process of calculating the regulation performance evaluation index also includes the following steps: calculating the autocorrelation function sequence of the dynamic residual sequence in the time domain; identifying whether the generator set has controlled oscillations based on the envelope decay rate of the autocorrelation function sequence; if controlled oscillations are identified, extracting the dominant frequency of the autocorrelation function sequence to calculate the regulation loss factor, and performing a punitive correction on the regulation performance evaluation index based on the regulation loss factor.

[0035] Specifically, step 5 involves the following steps to statistically analyze the distribution probability characteristics of the corrected residual sequence: obtaining the rated capacity of the generator set and setting a percentage error band centered at zero and with a width of 1% of the rated capacity as a preset value range; statistically analyzing the proportion of samples whose corrected residual sequence falls within the preset value range to obtain the distribution probability value.

[0036] Specifically, after step 5, the following steps are also included: calculating the third central moment of the corrected residual sequence within a preset sampling period to obtain the skewness characteristics; identifying whether the generator set is in the actuator saturation condition based on the correlation between the sign of the skewness characteristics and the change direction of the automatic generation control command sequence; if the identification result is that the generator set is in the actuator saturation condition, then reducing the weight of the component related to the central offset in the regulation efficiency evaluation index through the weight correction operator.

[0037] This invention provides a generator set The performance evaluation method is used to quantify and statistically analyze the quality of generator set response to automatic generation control commands during industrial production. It obtains the sequence of automatic generation control commands from the generator set within a preset sampling period via a data interface. and actual power sequence Due to system time delays in the physical transmission and measurement stages between command issuance and power feedback, the system shifts the actual power sequence point by point within a preset time offset range. Calculate the shifted sequence and the automatic generation control command sequence. The cross-relationship values ​​between the two variables are used to determine the translation time when the cross-relationship value reaches its maximum value as the adjustment time delay. According to the adjustment time delay For the actual power sequence Phase alignment is performed to eliminate statistical phase deviations caused by communication delays. To address the physical inertia constraints of the generator set during regulation, the system acquires real-time load data of the generator set and matches the corresponding physical inertia time constant from a preset parameter mapping table based on the preset load range of the real-time load data. ; Utilizing this physical inertial time constant Automatic generation control command sequence Perform single-order lag filtering to generate a reference response curve. This curve characterizes the ideal response trajectory of the unit under physical limit constraints; the actual generated power sequence after phase alignment is calculated. Compared with the reference response curve The difference between them yields the dynamic residual sequence. The calculation formula is: To remove background random noise and identify effective adjustment actions, the system extracts dynamic residual sequences. First-order difference sequence And within a preset sliding statistical window, the absolute value of the difference between the sign function values ​​of adjacent sampling points is calculated; the sign flip frequency... Calculated according to the following rules: ,in, This represents the total number of samples within the sliding statistical window. This serves as a sampling time identifier. For first-order difference sequences in The value of the moment. It is a symbolic function.

[0038] Collect raw power data from the unit under steady-state operation for no less than 24 hours and calculate the first-order difference sequence. For first-order difference sequences The sign-flipping frequency statistical probability density was used, and the frequency value where the cumulative probability of the distribution curve reached 95% was selected as the preset fluctuation threshold to eliminate high-frequency jitter with extremely small amplitude caused by background thermal noise. Based on the static friction test data of the turbine high-pressure regulating valve, the minimum command change required for valve start-up was extracted, and the ratio of the change to the rated regulating rate of the unit was set as the preset regulating rate ratio. A first-order difference sequence was then used. The amplitude exceeds this ratio and the sign flip frequency When the frequency is below the preset fluctuation threshold, the sampling segment is determined to be the operating range after the unit overcomes mechanical hysteresis. At the physical level, random environmental noise is separated from the regulation response logic with a clear control intention. When the symbol flip frequency... The first-order difference sequence is below the preset fluctuation threshold. When the amplitude is greater than the preset adjustment rate ratio, the system identifies the current sampling period as the effective adjustment range, which corresponds to the physical process by which the generator set overcomes the nonlinear dead zone of the actuator; utilizing the preset mechanical dead zone compensation factor. With sign flip frequency Amplitude contraction mapping is performed on the dynamic residual sequence within the effective adjustment interval to generate a corrected residual sequence. The amplitude contraction mapping process follows these calculation rules: ,in, To correct the residual sequence in The value of the moment. For dynamic residual sequences in The value of the moment. The mechanical dead zone compensation factor is set within a certain range. to Between these, energy hysteresis is compensated for by the backlash of the actuator.

[0039] In obtaining the corrected residual sequence Then, the system calculates its probability distribution characteristics and kurtosis characteristics; it sets the value at zero and the width to the rated capacity of the generator set. The percentage error band is used as the preset value range to statistically correct the residual sequence. The probability distribution of samples falling within this interval is obtained by calculating the percentage of samples that fall within it; simultaneously, the corrected residual sequence is calculated. The fourth central moment is used to obtain the kurtosis characteristic. The regulation efficiency evaluation index of the generator set within the preset sampling period is obtained by weighting the distribution probability value and the kurtosis characteristic.

[0040] Mechanical dead zone compensation factor during amplitude contraction mapping The calibration process includes: the unit is in open-loop command test state, a step power command with an amplitude of 0.5% to 1.5% of the rated capacity is issued, the hysteresis energy distribution is monitored when the power feedback curve crosses the reverse gap, and the residual sequence is corrected using least squares fitting. The mean of the action range distribution approaches zero, and the fitting coefficient is used as the mechanical dead zone compensation factor under this load base point. Store in the parameter mapping table; in step S108, the evaluation index calculation process addresses the one-way bias caused by the physical output limit by using the corrected residual sequence. Skewness characteristics With automatic generation control command sequence By identifying the correlation of change direction, the unit is in the saturation condition of the actuator. The weight value of the center offset component of the evaluation index is reduced so that the evaluation results focus on the quality of the unit within the physical reach of the unit.

[0041] To improve the adaptability of the evaluation method under extreme working conditions, the system calculates the corrected residual sequence. The third-order central moment within a preset sampling period is used to extract skewness features; based on the sign of the skewness features and the automatic generation control command sequence... The system identifies whether the generator set is in actuator saturation condition by analyzing the correlation between the directions of change. If the generator set is determined to be in actuator saturation condition, the system reduces the weight of the component related to center offset in the regulation efficiency evaluation index to correct the statistical bias caused by the physical output limit. In addition, the system establishes multiple sliding statistical windows within a preset sampling period and calculates the automatic generation control command sequence within each window. With actual power sequence The system calculates the maximum cross-correlation coefficients and the coefficient of variation of each maximum within a sliding statistical window to generate a correlation stability index, which is used to determine the confidence weights of the regulation effectiveness evaluation index. For non-stationary background disturbances, the system identifies generator sets within an automatic generation control command sequence. Calculate the dynamic residual sequence within a steady-state sampling period under constant numerical conditions. The root mean square value is used to obtain the steady-state noise basis index, and a monotonically increasing function is used to perform mapping compensation on the boundary values ​​of the preset range to generate an adaptive evaluation bandwidth; at the same time, the dynamic residual sequence is calculated. The time-domain autocorrelation function sequence is used to identify controlled oscillations based on the envelope decay rate. If controlled oscillations are identified in the generator set, the dominant frequency of the autocorrelation function sequence is extracted to calculate the regulation loss factor.

[0042] Example 1 In a power system equipped with 600MW coal-fired generating units and a high proportion of renewable energy sources, the automatic generation control command sequence... The amplitude is the rated capacity of the unit. to Fluctuations due to the actual power sequence Due to background random disturbances caused by fluctuations in the coal feeding system and primary frequency regulation actions, the response trajectory and environmental random components are statistically mixed. As a result, within the command variation range of 6MW, the power deviation cannot directly correspond to the unit's regulation contribution. The system uses the plant monitoring interface to collect command sequences. With actual power sequence The cross-correlation function between the two is calculated within an offset range of 0s to 60s to determine the current adjustment time lag. The value is 12 seconds, and based on this, the actual power sequence is calculated. Perform phase alignment; based on the unit's current load condition of 480MW, obtain the physical inertia time constant from the preset parameter mapping table. For 50 seconds, the instruction sequence A single-order hysteresis filter is applied to generate a baseline response curve that conforms to the thermophysical characteristics of the 600MW unit. .

[0043] In obtaining dynamic residual sequences Then, the system establishes a sliding statistical window of length 300 seconds to calculate the first-order difference sequence. and its symbol flipping frequency When the unit is overcoming the valve nonlinear dead zone, the power evolution shows a consistent trend, and the sign flip frequency... Descending to At this point, the conditions for determining the effective adjustment range are met, and the system uses a preset value. Mechanical dead zone compensation factor For dynamic residual sequences Perform amplitude contraction mapping to calculate the corrected residual sequence. The specific rules are as follows: Corrected residual sequence The sample distribution probability within the preset value range of ±6MW is: This probability distribution value, combined with the corrected residual sequence The kurtosis characteristic is used to quantify the consistency of the unit's command response; at the same time, the residual sequence is corrected by calculation. The skewness characteristics were used to identify the actuator saturation state of the unit when it fluctuated around 595MW, and the evaluation indicators were corrected by multiplying them according to the confidence weights determined by the correlation stability index; the unit was in the automatic generation control command sequence. During a constant steady-state sampling period, the system calculates the dynamic residual sequence. The root mean square value is used to determine the steady-state noise floor index. The boundary values ​​of a preset range are adjusted in real-time using a monotonically increasing function to an adaptive evaluation bandwidth that varies with the ambient noise floor. The dynamic residual sequence... The decay rate of the autocorrelation function sequence envelope remains stable in the time domain.

[0044] Example 2 To verify the fairness of the evaluation of generator unit response to automatic power generation control commands, a verification environment was established on a 600MW coal-fired unit simulation platform. The response process was simulated using the governor logic model and the thermodynamic system transfer function. The experimental data originated from operating section data with a sampling frequency of 1Hz exported from an industrial real-time database. Addressing the technical challenges of evaluating the regulation performance under background random fluctuations, the actual power generation sequence was used... Superimposed with Gaussian white noise of 20dB signal-to-noise ratio, and the sampling window length The setting is based on the balance between the unit's regulating rate and the calculated load. If the value is too low, sampling noise cannot be filtered out. If the value is too high, it will weaken the sensitivity of the instruction following feature. This experiment will adjust the sampling window length. The value was set to 300; the experimental group used the regulation performance evaluation method provided by this invention, while the control group used the error arithmetic mean statistical method. After the experiment started, the system calculated the automatic power generation control command sequence. With actual power sequence The cross-correlation function between them determines the adjustment time delay. Phase alignment is performed within 15 seconds, and the physical inertia time constant is matched based on the real-time load of the unit under 500MW operating conditions. Generate the baseline response curve after 45 seconds. To extract dynamic residual sequences .

[0045] Table 1: Intermediate Calculation Process Data within a Typical Sampling Segment

[0046] See Table 1, in At time 101, the automatic generation control command sequence In a constant state, the actual power sequence The fluctuations originate from simulated background noise, and the calculated symbol flip frequency... for Exceeding the preset fluctuation threshold If the current sampling segment is determined to be an invalid adjustment interval, then the residual sequence is corrected. Directly inherit dynamic residual sequences The value; in At time 251, the unit is in the process of overcoming the valve nonlinear dead zone, and the sign flipping frequency is... Descending to The system uses preset values Mechanical dead zone compensation factor For dynamic residual sequences Amplitude contraction mapping was performed to correct the residual amplitude from -1.95MW to -1.65MW. The corrected value was used for subsequent probability distribution statistics to eliminate statistical bias caused by actuator stickiness; this was done to verify the mechanical dead zone compensation factor. The rationality of the selected values; the experimental setup includes a parameter set with gradients. Below At that time, insufficient residual correction resulted in the evaluation indicators still containing errors during the instruction change period. The non-adjustment error, when Higher than At that time, correct the residual sequence The probability distribution characteristics show non-physical over-concentration, as determined by experiments. These are the optimal parameters given the current actuator characteristics; the unit is... As the system approaches its rated output limit of 600MW, the power deviation exhibits a unidirectional bias due to physical constraints. The system calculates the corrected residual sequence. The skewness eigenvalue is It identifies the current state of actuator saturation and automatically reduces the weight of center offset in the evaluation index, avoiding misjudging physical limitations as deterioration of adjustment accuracy; experimental data comparison shows that the control group is affected by background noise interference. The indicators drifted randomly, while the experimental group... The probability distribution within the rated capacity error band is stable at Furthermore, as the white noise power in the simulated environment increased from 0.5MW to 2.5MW, the coefficient of variation of the evaluation indicators of the experimental group remained stable. Within.

[0047] Example 3 This embodiment combines Figures 2 to 4 This document describes a method for evaluating the AGC (Automatic Generative Control) regulation performance of a generator set. Figure 2As shown, the overall architecture is divided into three logical domains from bottom to top: the field device layer physical domain, the core computing and storage layer logical domain, and the display and interaction layer user domain. In the bottom field device layer physical domain, the generator set body is coupled to the actuator valve through mechanical connection. The data acquisition terminal RTU samples the actuator execution signal and uploads the collected data through the industrial Ethernet channel. The middle core computing and storage layer logical domain deploys a parameter mapping database and an industrial real-time database. The two interact with the performance evaluation calculation server through data reading and storage interfaces. The server integrates a data preprocessing component to perform phase alignment and filtering, a benchmark response generation component to perform physical modeling, a dead zone identification and energy compensation component, a statistical feature extraction component to handle distribution, kurtosis and skewness calculation, and an indicator adaptive correction component to perform noise reduction and anti-saturation processing. The top display and interaction layer user domain is configured with a performance monitoring engineer station. This workstation is connected to an indicator visualization screen and has the function of generating evaluation reports for the publication and display of the final evaluation results.

[0048] like Figure 3 As shown, within the three horizontal axis test intervals of low noise (0.5MW), medium noise (1.5MW), and high noise (2.5MW), the performance differences between the method of this invention and the traditional method are compared and analyzed. The vertical axis represents the magnitude of the coefficient of variation. The legend defines the coefficient of variation histogram of the method of this invention with a grid texture and the coefficient of variation histogram of the traditional method with a vertical stripe texture, respectively. The statistical results show that as the background noise intensity increases from 0.5MW to 2.5MW, the coefficient of variation of the traditional method rises and exceeds 0.12, while the coefficient of variation of the method of this invention remains below 0.05 throughout the entire interval. Figure 4 As shown, the various factors affecting the accuracy of AGC regulation performance evaluation are constructed into a fishbone diagram model. The head of the fish points to the core objective, namely the accuracy of AGC regulation performance evaluation. The two sides of the fish expand into six branches. Among them, the signal transmission and timing branch includes the calculation of regulation time delay and cross-correlation phase alignment; the physical inertial response branch covers the generation of the reference response curve and the physical inertial time constant; the environmental background interference branch involves statistical decoupling technology, adaptive evaluation bandwidth, and steady-state noise basis; the actuator characteristics branch corresponds to nonlinear dead zone identification, mechanical dead zone compensation factor, and sign flip frequency monitoring; the operating condition boundary constraint branch is related to the weight correction of the relevant indicators, actuator saturation determination, and skewness feature identification; and the abnormal regulation behavior branch integrates signal correlation auditing, regulation loss factor, and controlled oscillation identification.

[0049] Example 4 During cross-regional load tracking, thermal power units are subject to objective constraints such as boiler thermal inertia and the variation of turbine condensate status with output level, resulting in differences in physical response characteristics under different load conditions. To determine the physical inertia time constant... The system extracts command response profile data from the historical operating cycles of the generating units, and uses a least-squares fitting algorithm to identify the dynamic parameters of the actual power response process at each load level, in order to minimize the actual power sequence. Compared with the reference response curve The sum of squared residuals between the load intervals is used as the optimization objective to establish the relationship between the load interval and the physical inertia time constant. The parameter mapping table; taking a 600MW rated capacity coal-fired unit as an example, when the unit is at a 300MW output level, the fitted physical inertia time constant is determined. The time constant of physical inertia is set at 65.2s; when the output level is increased to 540MW, the time constant of physical inertia is... The time constant of the inertial link is shortened to 42.5s. This parameter mapping table is pre-stored in the system memory and is used to dynamically match the corresponding inertial link time constant based on the real-time load data of the generator set during the real-time evaluation process.

[0050] When the system is from the dynamic residual sequence When removing background interference, the identification logic of the steady-state sampling segment is determined by the fluctuation intensity of the command sequence; the system uses a detection window of 60 seconds within the evaluation period and periodically calculates the automatic generation control command sequence within this window. The sample standard deviation; if the sample standard deviation is lower than the rated capacity of the unit for 120 consecutive seconds. The system determines that the current sampling segment has entered a steady state; for a 600MW unit, this is defined as the sample standard deviation consistently falling below 0.3MW. Within this steady-state sampling segment, the system calculates the dynamic residual sequence. The root mean square value is defined as the steady-state noise basis index. The unit is MW; the system utilizes steady-state noise baseline performance. For the preset value range Perform nonlinear rescaling to calculate the adaptive evaluation bandwidth. The calculation formula follows: ,in, To adaptively evaluate bandwidth, For the preset value range, As a steady-state noise baseline index, It is an exponential function; the evaluation bandwidth is adjusted in real time according to the ambient noise energy density, ensuring that the statistical results still focus on the regulation contribution of the unit itself under the background of frequency fluctuations caused by the high proportion of new energy grid connection.

[0051] In addition, to address the statistical bias caused by the physical output limit, the system utilizes a corrected residual sequence. Skewness characteristics Perform operating condition identification and weight allocation; the system calculates and corrects the residual sequence. The third central moment within a preset sampling period is used to obtain skewness characteristics. When skewness characteristics The absolute value exceeds And skewness characteristics Symbols and automatic generation control command sequences When the changes are in the same direction, the system determines that the generator set is in the saturation condition of the actuator; under this condition, the system calls the weight correction operator to change the weight of the component related to the center offset in the regulation efficiency evaluation index from the initial value. Reduced to Furthermore, if the dynamic residual sequence The energy is dominant in the frequency range of 0.01Hz to 0.05Hz, and the envelope decay rate of the autocorrelation function sequence is lower than that of the other two frequencies. The system determined that a hidden controlled oscillation existed and calculated the value as follows: The adjustment loss factor is used to subtract and correct the final adjustment effectiveness evaluation index.

[0052] Example 5 In the on-site commissioning scenario for the first deployment of a 300MW subcritical unit, the system established the mechanical dead zone compensation factor through an offline calibration program. The initial reference value is used to control the unit to enter automatic power generation control. Before the closed-loop test, the amplitude of the generator set at the current stable load point of 240MW is the rated capacity of the unit. to The step test command is used to monitor the feedback stroke of the turbine's high-pressure regulating valve and the actual power generation sequence. The system identifies the power hysteresis characteristics caused by the backlash of the actuator by measuring the response difference. A calibration algorithm extracts the hysteresis time of power variation relative to command variation within the step response segment. Least squares fitting is then used to map this hysteresis time to the rated regulating rate of the unit, calculating the equivalent power deviation of the current valve mechanical dead zone. This deviation is used as input, and the system determines the mechanical dead zone compensation factor according to the linear mapping rule. for The value is then stored in the configuration parameter table of the memory as a calculation benchmark for performing amplitude contraction mapping on the effective adjustment range during subsequent evaluation.

[0053] When the system faces deployment environments with varying background noise levels, a dynamic calibration procedure with a preset fluctuation threshold is used to determine the symbol switching frequency. The determination boundary is defined as follows: during the steady-state operation phase after the unit is connected to the grid, the system collects a sequence of automatic generation control commands for 1800 seconds. The original power section data with a sample standard deviation below 0.1 MW, and according to the dynamic residual sequence The first-order difference procedure is used to calculate the symbol flip frequency sequence under purely random perturbations. The system performs probability distribution statistics on it, and then... The frequency value of the upper boundary of the confidence interval is determined as the preset fluctuation threshold under the current environment, and the calculated value is: Under these parameter settings, the system separates the random thermal noise from the sensors from the high-frequency random signals caused by combustion system disturbances from the controlled regulation response. Through the aforementioned pre-deployment calibration process, the system establishes a parameter system corresponding to the unit's physical characteristics and environmental noise intensity. The resulting regulation performance evaluation index maintains a fluctuation deviation within a certain range during the subsequent 72 hours of continuous operation. Within.

[0054] Example 6 Automatic generation control is implemented for 1000MW ultra-supercritical units. During the on-site verification of regulation efficiency, the system collected the actual power generation sequence of the unit during continuous operation at an 800MW load for 3600 seconds. The dynamic residual sequence is extracted according to the procedure determined by the specific implementation method. To quantify and identify the controlled oscillation characteristics during the power command following process, the system calculates the dynamic residual sequence. Autocorrelation function sequence ,in As the lag order, the system extracts The envelope decay rate within the current sampling period is calculated by fitting the envelope to the first three local maxima of the sequence using an exponential function. ;when The value is lower than the preset attenuation threshold Furthermore, when the dominant frequency of the autocorrelation function sequence is within the 0.02Hz to 0.08Hz frequency band, the system determines that the unit has implicit controlled oscillations and extracts the regulation loss factor corresponding to the dominant frequency. Adjust the loss factor The penalty component is included in the regulatory effectiveness evaluation index to correct the accuracy contribution caused by the reciprocating motion of the actuator.

[0055] When the system detects phase drift or random jump interference in the power signal collected by the remote terminal unit, the correlation stability index is used to determine the confidence weight of the regulation effectiveness evaluation index. The system establishes a sliding statistical window with a width of 300s and a sliding step size of 60s within a 3600s evaluation period, and calculates the automatic generation control command sequence within each sliding statistical window. With actual power sequence maximum cross-correlation coefficient , For window indexing; the system calculates the coefficient of variation of the maximum cross-correlation coefficients across all sliding statistical windows. Through calculation The coefficient of variation is obtained by the ratio of the standard deviation to the mean. Value exceeds the limit The system determines that there is non-physical distortion in the causal relationship of the signal within the current evaluation period, and uses a linear attenuation operator. Determine the confidence weights of the evaluation indicators ; this confidence weight The product is adjusted by multiplying the regulation efficiency evaluation index, and the statistical variability of the adjusted index stabilizes in the unit's cross-load variable operating condition range. the following.

[0056] Example 7 according to Figure 5 As shown, this embodiment also provides a generator set AGC regulation performance evaluation system, including: The data acquisition and time delay processing module 1 is used to acquire the automatic power generation control command sequence and the actual power generation sequence of the generator set, process them to obtain the adjustment time delay, and perform phase alignment processing on the actual power generation sequence. The reference generation and residual calculation module 2 is used to generate a reference response curve and a dynamic residual sequence based on the phase alignment result and the preset physical inertia time constant of the generator set. The frequency calculation and interval identification module 3 is used to perform feature analysis on the dynamic residual sequence, identify the effective adjustment interval, and calculate the symbol flip frequency. The residual correction processing module 4 is used to correct the dynamic residual sequence based on the sign flip frequency and the preset compensation factor to obtain the corrected residual sequence. The index evaluation and processing module 5 is used to statistically correct the characteristic parameters of the residual sequence, calculate the regulation performance evaluation index of the generator set within the preset sampling period, and evaluate the AGC regulation performance of the generator set based on the regulation performance evaluation index.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for evaluating the AGC (Automatic Generative Control) performance of a generator set, characterized in that, include: Step 1: Obtain the automatic power generation control command sequence and actual power generation sequence of the generator set, process them to obtain the adjustment time delay, and perform phase alignment processing on the actual power generation sequence; Step 2: Based on the phase alignment results and the preset physical inertia time constant of the generator set, generate the reference response curve and dynamic residual sequence; Step 3: Perform feature analysis on the dynamic residual sequence to identify the effective adjustment interval and calculate the sign flip frequency; Step 4: Correct the dynamic residual sequence based on the symbol flipping frequency and the preset compensation factor to obtain the corrected residual sequence; Step 5: Statistically correct the characteristic parameters of the residual sequence, calculate the regulation performance evaluation index of the generator set within the preset sampling period, and evaluate the AGC regulation performance of the generator set based on the regulation performance evaluation index.

2. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, In step 1, the process for determining the adjustment time delay is as follows: The actual power generation sequence is shifted point by point within a preset time offset range, and the correlation values ​​between the shifted sequence and the automatic generation control command sequence are calculated. The translation time corresponding to when the cross-correlation value reaches its maximum value is determined as the regulation time delay of the generator set; Step 1 is followed by the following steps: establishing multiple sliding statistical windows within a preset sampling period, and calculating the maximum cross-correlation coefficient between the automatic power generation control command sequence and the actual power generation sequence within each sliding statistical window; extracting the maximum cross-correlation coefficients of all sliding statistical windows, and calculating the coefficient of variation of the maximum cross-correlation coefficients among all sliding statistical windows to generate a correlation stability index; determining the confidence weights of the regulation efficiency evaluation index based on the correlation stability index, and using the confidence weights to perform product correction on the regulation efficiency evaluation index.

3. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, Step 2, generating the baseline response curve, includes the following steps: acquiring real-time load data of the generator set; matching the corresponding physical inertia time constant of the generator set from a preset parameter mapping table based on the preset interval of the real-time load data; and using the matched physical inertia time constant to perform hysteresis filtering on the automatic generation control command sequence to generate the baseline response curve.

4. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, In step 3, the formula for calculating the symbol flip frequency is as follows: in, This represents the total number of samples within the sliding statistical window. This serves as a sampling time identifier. For first-order difference sequences in The value of the moment. It is a symbolic function; The first-order difference sequence is extracted from the dynamic residual sequence, and the sign flip frequency is the result of the standardization of the sum of the absolute values ​​of the differences between the sign function values ​​of adjacent sampling points in the first-order difference sequence within the sliding statistical window.

5. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, The formula for magnitude shrinkage mapping of the dynamic residual sequence in step 4 is as follows: in, To correct the residual sequence in The value of the moment. For dynamic residual sequences in The value of the moment. For mechanical dead zone compensation factor, The sign flipping frequency.

6. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, Before step 5, the following steps are also included: identifying the steady-state sampling segment of the generator set in a numerically constant state of the automatic power generation control command sequence; calculating the root mean square value of the dynamic residual sequence within the steady-state sampling segment to obtain the steady-state noise basis index; using the steady-state noise basis index, performing mapping compensation on the boundary values ​​of the preset value range through a preset monotonically increasing function to generate an adaptive evaluation bandwidth that is adjusted in real time with environmental background fluctuations.

7. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, Step 5, in the process of calculating the regulation performance evaluation index, also includes the following steps: calculating the autocorrelation function sequence of the dynamic residual sequence in the time domain; identifying whether the generator set has controlled oscillations based on the envelope decay rate of the autocorrelation function sequence; if controlled oscillations are identified, extracting the dominant frequency of the autocorrelation function sequence to calculate the regulation loss factor, and performing a punitive correction on the regulation performance evaluation index based on the regulation loss factor.

8. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, Step 5 involves the following steps to statistically analyze the distribution probability characteristics of the corrected residual sequence: obtaining the rated capacity of the generator set and setting a percentage error band centered at zero and with a width of 1% of the rated capacity as a preset value interval; statistically analyzing the proportion of samples whose corrected residual sequence falls within the preset value interval to obtain the distribution probability value.

9. The method for evaluating the AGC regulation performance of a generator set according to claim 1, characterized in that, After step 5, the following steps are also included: calculating the third central moment of the corrected residual sequence within the preset sampling period to obtain the skewness characteristics; identifying whether the generator set is in the actuator saturation condition based on the correlation between the sign of the skewness characteristics and the change direction of the automatic generation control command sequence; if the identification result is that the generator set is in the actuator saturation condition, then reducing the weight of the component related to the central offset in the regulation efficiency evaluation index through the weight correction operator.

10. A generator set AGC regulation performance evaluation system, characterized in that, include: The data acquisition and time delay processing module is used to acquire the automatic power generation control command sequence and the actual power generation sequence of the generator set, process them to obtain the adjustment time delay, and perform phase alignment processing on the actual power generation sequence. The benchmark generation and residual calculation module is used to generate benchmark response curves and dynamic residual sequences based on phase alignment results and preset generator physical inertia time constants. The frequency calculation and interval identification module is used to perform feature analysis on the dynamic residual sequence, identify the effective adjustment interval, and calculate the symbol flip frequency. The residual correction processing module is used to correct the dynamic residual sequence based on the sign flipping frequency and a preset compensation factor to obtain the corrected residual sequence. The index evaluation and processing module is used to statistically correct the characteristic parameters of the residual sequence, calculate the regulation performance evaluation index of the generator set within the preset sampling period, and evaluate the AGC regulation performance of the generator set based on the regulation performance evaluation index.

Images