An emergency repair scheme generation method for a power generation terminal based on artificial intelligence

Abstract

The present application relates to the field of power generation equipment maintenance, especially to a power generation terminal emergency maintenance scheme generation method based on artificial intelligence, the present application obtains the load characteristics of the power grid accessed by the generator set, determines the load disturbance excitation value, carries out interference disturbance test on the generator set, obtains the interference disturbance characteristics of the remaining generator within the limit time, comprehensively evaluates the interference state of the target generator affected by the system interference, and then carries out specific dimension screening on the multi-dimensional operation data in the strong interference period, the present application builds the exclusive specific dimension of each generator, accurately monitors the running state, effectively isolates the influence of the interference between the power grid and the unit, reduces the influence of the external power grid load change and the complex mutual interference between multiple generator sets on the generator fault judgment when the generator works in the strong interference state, significantly reduces the detection error and the fault misjudgment rate, and shortens the fault to improve the accuracy of subsequent maintenance decision.

Description

Technical Field

[0001] This invention relates to the field of power generation equipment maintenance, and in particular to a method for generating emergency maintenance plans for power generation terminals based on artificial intelligence. Background Technology

[0002] When large generator units require maintenance or routine upkeep, power plant operators typically need to coordinate resources from multiple parties, such as hiring professional maintenance teams, leasing large hoisting and transportation equipment, and procuring and ensuring the timely supply of critical spare parts. However, large power plants are often located in resource-rich areas, such as remote mountainous regions, river valleys, mining areas, or the outskirts of industrial bases. These areas are often geographically remote, with widely distributed or complex equipment structures, posing challenges to accessibility and logistical efficiency. These factors collectively make the maintenance and repair process for critical equipment extremely cumbersome and time-consuming. To address these pain points, various online monitoring and fault diagnosis systems have emerged.

[0003] For example, Chinese Patent Publication No. CN119199526A discloses a green energy generator set monitoring system, comprising: a generator set monitoring module for real-time monitoring of the generator set and acquiring basic operating data information of the generator set; a water source monitoring module for acquiring current water supply information of the generator set; an anomaly analysis module for receiving the basic operating data information and water supply information, and performing potential anomaly analysis on the basic operating data information and water supply information to obtain an anomaly variation coefficient of the generator set; an anomaly judgment module for comparing the anomaly variation coefficient of the generator set with a pre-set threshold range, and judging whether there is an anomaly in the current generator set based on the comparison result; and an anomaly early warning module for generating a corresponding early warning signal based on the judgment result and executing the early warning strategy corresponding to the current early warning signal. This invention enables early judgment of generator set faults.

[0004] However, the following problems still exist in the existing technology:

[0005] Different generator sets are affected by different external loads during actual operation, resulting in grid load fluctuations that are difficult to predict accurately. Furthermore, during actual operation, multiple generator sets in the same power plant operate in parallel, and complex mutual interference occurs between the units due to power distribution, frequency coupling, and cross-influence of the excitation system. This leads to high-dimensional and nonlinear coupling characteristics in the operating parameters, resulting in different performance of different generators in various dimensions of operating parameters. Consequently, the detection data errors are large, and the generator fault misjudgment rate is high. Summary of the Invention

[0006] To address this, the present invention provides an artificial intelligence-based method for generating emergency maintenance plans for power generation terminals. This method overcomes the challenges posed by existing technologies where multiple generator units operate in parallel within the same power plant, resulting in complex mutual interference between units due to power distribution, frequency coupling, and cross-influence of the excitation system. This leads to high-dimensional, nonlinear coupling characteristics in the operating parameters, causing different generators to exhibit varying performance in different dimensions of operating parameters. Consequently, this results in larger errors in the detection data and a higher rate of misdiagnosis of generator faults.

[0007] To achieve the above objectives, the present invention provides a method for generating emergency maintenance plans for power generation terminals based on artificial intelligence, comprising:

[0008] The load characteristics of the power grid to which the generator set is connected are obtained to identify the interference excitation period, and the load disturbance excitation value is determined based on the end value of the interference excitation period;

[0009] Interference disturbance tests are conducted on generator sets, including limiting the operating parameters of individual generators one by one and obtaining the interference disturbance characteristics of the remaining generators within the limited time.

[0010] Based on the load disturbance excitation value and interference disturbance characteristics, interference characterization parameters for the engine set are calculated to evaluate the interference state of the target generator affected by system interference.

[0011] To address the interference state of the generator set, the generator set is monitored, including...

[0012] The test operation characteristics of each generator in several dimensions are obtained during the interference disturbance test. Based on the test operation characteristics, a set of specific dimensions for each generator is constructed. Based on the set of specific dimensions, the operation characteristics of the generator corresponding to the low interference dimension during daily operation are extracted. Based on the operation characteristics, it is determined whether there is an abnormality in the generator set.

[0013] Alternatively, extract all dimensions of the generator's operating characteristics to determine if the generator is malfunctioning;

[0014] The step of constructing a set of specific dimensions for each generator based on operational characteristics includes screening low-interference dimensions based on test operational characteristics and constructing a set of specific dimensions based on low-interference dimensions.

[0015] Furthermore, the load characteristics of the power grid to which the generator set is connected are obtained to identify the process during the interference excitation period, including,

[0016] Real-time monitoring of the dynamic sequence data of active power at the grid connection point, and calculation of the gradient of active power change;

[0017] The ratio of the active power change gradient to the standard change gradient is determined as the change gradient factor.

[0018] Real-time acquisition of power grid frequency offset is used to analyze the duration of frequency offset exceeding the offset threshold.

[0019] If the gradient change factor is greater than a first predetermined threshold and the frequency duration is greater than a second predetermined threshold, then the time domain segment corresponding to the frequency duration is determined, and the time domain segment is identified as the interference excitation period.

[0020] Furthermore, the process of determining the load disturbance excitation value based on the end-point value of the interference excitation period includes,

[0021] The active power value at the end of the interference excitation period is used as the terminal value;

[0022] The ratio of the terminal endpoint value to the standard terminal power value is used as the base value;

[0023] The load disturbance coefficient is determined based on the changing gradient factor and the duration of the interference excitation period;

[0024] The product of the base value and the load disturbance coefficient is determined as the load disturbance excitation value.

[0025] Furthermore, by individually restricting the operating parameters of each generator, the interference disturbance characteristics of the remaining generators within the restricted time are obtained, including...

[0026] Limit the operating parameters of each generator in the generator set, including frequency and voltage;

[0027] The instantaneous frequency deviation peak value of each remaining generator in the generator set and the steady-state recovery time of each generator voltage to the allowable tolerance are collected synchronously.

[0028] The maximum instantaneous frequency deviation peak value and the maximum steady-state recovery time value are defined as the interference disturbance characteristics.

[0029] Furthermore, the process of calculating the interference characterization parameters for the engine set based on the load disturbance excitation value and interference disturbance characteristics includes,

[0030] The ratio of the peak value of the instantaneous frequency deviation to the standard value of the instantaneous frequency deviation is determined as the instantaneous frequency parameter.

[0031] The ratio of the maximum steady-state recovery time to the standard steady-state recovery time is determined as the recovery time parameter;

[0032] The weighted summation of the load disturbance excitation value, instantaneous frequency parameter, and recovery time parameter is determined as the interference characterization parameter.

[0033] Furthermore, the interference state of the target generator under the influence of system interference is assessed, including,

[0034] If the interference characterization parameter is greater than the interference characterization threshold, a strong interference state label is set for the target generator;

[0035] If the interference characterization parameter is less than or equal to the interference characterization threshold, a weak interference state label is set for the target generator.

[0036] Furthermore, the test operation characteristics of each generator in several dimensions are obtained during the interference disturbance test. Based on these test operation characteristics, a set of specific dimensions for each generator is constructed, including:

[0037] The test operation characteristics of the target generator are collected in several dimensions, including stator winding temperature, bearing temperature, frequency, vibration, and active power.

[0038] Dimensions with a sliding window standard deviation lower than a predetermined fluctuation threshold are selected as low-interference dimensions;

[0039] Based on the selected low-interference dimensions, a set of specific dimensions for the target generator is constructed.

[0040] Furthermore, low-interference dimensions with a sliding window standard deviation below a predetermined fluctuation threshold are selected, including:

[0041] Obtain the sliding window standard deviation of the corresponding values ​​of stator winding temperature, bearing temperature, frequency, vibration, and active power respectively;

[0042] Each sliding window standard deviation is compared with its corresponding preset fluctuation threshold. If the sliding window standard deviation is lower than the fluctuation threshold, the dimension corresponding to the sliding window standard deviation is marked as a low interference dimension.

[0043] Furthermore, based on the specific dimension set, the corresponding dimension's operational characteristics during the generator's daily operation are extracted. Based on these operational characteristics, it is determined whether the generator set exhibits any abnormalities, including...

[0044] Based on the specific dimensional set of the target generator, the corresponding operational characteristics of the dimensions are collected;

[0045] If a parameter in a specific dimension exceeds the preset anomaly threshold for that dimension, the target generator is determined to be abnormal.

[0046] Furthermore, all dimensions of the generator's operational characteristics are extracted to determine whether the generator exhibits any abnormalities, including...

[0047] Collect the operating characteristics of the target generator, including stator winding temperature, bearing temperature, frequency, vibration, and active power.

[0048] The aforementioned operational features are input into a pre-trained anomaly detection model to determine if the target generator is abnormal.

[0049] Compared with existing technologies, this invention obtains the load characteristics of the power grid to which the generator set is connected, determines the load disturbance excitation value, conducts interference disturbance tests on the generator set, obtains the interference disturbance characteristics of the remaining generators within a limited time, comprehensively evaluates the interference state of the target generator under the influence of system interference, and then filters the multi-dimensional operating data of strong interference states. This invention considers screening out specific dimensions with low mutual interference to accurately monitor the engine operating status. By screening specific dimensions, this invention reduces the adverse effects of external power grid load changes and complex mutual interference between multiple generator sets on generator fault diagnosis when the generator is operating under strong interference states, reduces the fault misjudgment rate, improves the accuracy of fault detection, and is conducive to improving subsequent maintenance efficiency.

[0050] In particular, existing solutions often struggle to accurately distinguish between external disturbances such as grid load jumps and harmonic impacts, potentially misdiagnosing them as internal unit defects and leading to frequent misjudgments. The active power change gradient reflects the severity of sudden load changes and the intensity of energy impacts. In reality, external disturbances such as grid load jumps, the start-up / shutdown of large-capacity equipment, or fault clearing directly cause significant jumps in active power at the connection point within a short period. The active power change gradient can reflect the severity of these external disturbances. Frequency offset and its duration reflect the overall inertia support capacity and frequency regulation response level of the power system. External disturbances disrupt the system's power balance, causing the frequency to deviate from its rated value. The magnitude and duration of the frequency offset directly reflect the impact intensity of the disturbance on system stability and the system's ability to resist disturbances, serving as a characterization of whether the disturbance triggers systemic interference. This invention considers the active power change gradient and frequency offset to accurately select the interference excitation period, thereby providing a basis for distinguishing between grid disturbances and internal unit faults, reducing the fault misjudgment rate, improving the accuracy of fault detection, and improving subsequent maintenance efficiency.

[0051] In particular, when multiple generator units operate in parallel, a sudden drop in output from a single unit can trigger transient frequency shifts and oscillations in the group of units. Traditional monitoring methods struggle to effectively quantify the intensity of such interference between units. This invention sets up an interference disturbance test on the generator units, extracting the peak value of the instantaneous frequency deviation and the steady-state recovery time as key interference disturbance features. The peak value of the instantaneous frequency deviation reflects the instantaneous impact intensity caused by the sudden change in output on the unit; the steady-state recovery time quantifies the time required for the generator to recover from the disturbance to a stable operating state, reflecting the system's regulation capability. By considering both the peak value of the instantaneous frequency deviation and the steady-state recovery time, this invention transforms the complex inter-unit coupling impact into clear and measurable quantitative indicators. This avoids misjudging inter-unit coupling oscillations as single-unit internal faults, reduces the fault misjudgment rate, improves the accuracy of fault detection, and facilitates improved subsequent maintenance efficiency.

[0052] In particular, this invention selects five dimensions—stator winding temperature, bearing temperature, frequency, vibration, and active power—as candidate feature sets. During the identified strong interference periods, the sliding window standard deviation for each dimension is calculated to quantify its sliding window standard deviation under strong external and coupled disturbances. The calculated sliding window standard deviation is then compared with a preset fluctuation threshold for each dimension. This invention considers dimensions whose sliding window standard deviation is lower than their preset threshold under strong interference conditions—that is, dimensions with small sliding window standard deviations and weak influence from external and coupled disturbances—to construct a unique set of dimensions specific to the generator. In daily monitoring, features from this unique set of dimensions are detected, eliminating dimensions significantly affected by external factors. This improves the accuracy and real-time performance of anomaly detection, reduces the false fault rate, and enhances the accuracy of fault detection, thus improving subsequent maintenance efficiency. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the steps of an embodiment of the invention;

[0054] Figure 2 The logic block diagram is marked as the interference excitation period;

[0055] Figure 3 A logic block diagram for generating system interference state labels for the target generator;

[0056] Figure 4 The logic diagram is marked as low interference dimension. Detailed Implementation

[0057] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0058] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0059] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0060] Please see Figure 1As shown, this is a schematic diagram of the steps in an embodiment of the present invention. The method for generating an emergency maintenance plan for a power generation terminal based on artificial intelligence of the present invention includes:

[0061] Step S1: Obtain the load characteristics of the power grid to which the generator set is connected in order to identify the interference excitation period and determine the load disturbance excitation value based on the end value of the interference excitation period.

[0062] Step S2: Conduct interference disturbance tests on the generator set, including limiting the operating parameters of individual generators one by one, and obtaining the interference disturbance characteristics of the remaining generators within the limited time.

[0063] Step S3: Calculate the interference characterization parameters for the engine group based on the load disturbance excitation value and interference disturbance characteristics to evaluate the interference state of the target generator affected by system interference;

[0064] Step S4: Monitor the generator set for interference conditions, including...

[0065] The test operation characteristics of each generator in several dimensions are obtained during the interference disturbance test. Based on the test operation characteristics, a set of specific dimensions for each generator is constructed. Based on the set of specific dimensions, the operation characteristics of the generator corresponding to the low interference dimension during daily operation are extracted. Based on the operation characteristics, it is determined whether there is an abnormality in the generator set.

[0066] Alternatively, extract all dimensions of the generator's operating characteristics to determine if the generator is malfunctioning;

[0067] The step of constructing a set of specific dimensions for each generator based on operational characteristics includes screening low-interference dimensions based on test operational characteristics and constructing a set of specific dimensions based on low-interference dimensions.

[0068] Specifically, the load characteristics of the power grid to which the generator set is connected are obtained to identify the process during the interference excitation period, including:

[0069] Real-time monitoring of the dynamic sequence data of active power at the grid connection point, and calculation of the gradient of active power change;

[0070] The ratio of the active power change gradient to the standard change gradient is determined as the change gradient factor.

[0071] Real-time acquisition of power grid frequency offset is used to analyze the duration of frequency offset exceeding the offset threshold.

[0072] Specifically, the active power dynamic sequence data includes the active power at each time point. The gradient of active power change is the difference between the active power at a single time point and the active power at the previous time point. The interval between each time point can be set to 1 second.

[0073] Please see Figure 2 As shown, Figure 2 The logic block diagram is marked as the interference excitation period;

[0074] If the gradient change factor is greater than a first predetermined threshold and the frequency duration is greater than a second predetermined threshold, then the time domain segment corresponding to the frequency duration is determined, and the time domain segment is identified as the interference excitation period.

[0075] If the changing gradient factor is greater than a first predetermined threshold and the duration of the frequency is greater than a second predetermined threshold, this period is marked as the interference excitation period.

[0076] Specifically, the active power change gradient is collected in real time by a power transmitter or other equipment with power acquisition capability installed at the grid connection point of the generator set, and the sampling frequency should not be lower than 1Hz.

[0077] Specifically, the frequency offset of the power grid is collected in real time by a frequency monitoring device installed at the power grid connection point of the generator set.

[0078] Specifically, the standard variation gradient is derived from historical operating data of the power grid to which the target generator unit is connected. The average active power variation gradient of the power grid on a typical stable operating day is taken as the standard variation gradient of the power grid.

[0079] Specifically, the power grid frequency offset is the deviation between the actual operating frequency and the rated frequency of the power system. The offset threshold can be directly adopted from the allowable frequency deviation limit specified for the safe and stable operation of the power system. For example, the normal allowable deviation is usually specified as ±0.2Hz. Alternatively, it can be set according to specific power grid requirements and determined by those skilled in the art.

[0080] Specifically, the first predetermined threshold is used to determine whether the change in active power is so drastic that it may trigger significant interference. Based on the statistical analysis of historical interference triggering events such as heavy load switching and line faults, the typical lower limit value of the change gradient factor is determined when these events occur, and the average of the typical lower limit value is used as the first predetermined threshold.

[0081] Specifically, the second predetermined threshold is used to determine whether the duration of the frequency exceedance is long enough. This avoids frequent triggering of interference excitation period markers due to short-term instantaneous exceedances, and filters out disturbance events whose duration is sufficient to reflect the true power imbalance and system regulation capability. The value of the second predetermined threshold is not specifically limited according to the grid load and generator equipment conditions, and the reference value range is [0.5 seconds, 3 seconds].

[0082] Specifically, the process of determining the load disturbance excitation value based on the end-point value of the interference excitation period includes,

[0083] The active power value at the end of the interference excitation period is used as the terminal value;

[0084] The ratio of the terminal endpoint value to the standard terminal power value is used as the base value;

[0085] The load disturbance coefficient is determined based on the changing gradient factor and the duration of the interference excitation period;

[0086] The product of the base value and the load disturbance coefficient is determined as the load disturbance excitation value.

[0087] Specifically, the average active power at the end of several interference excitation periods under typical stable operating conditions of the power grid is used as the standard value of the terminal power.

[0088] Specifically, the ratio of the duration of the interference excitation period to a second predetermined threshold is used as the duration factor;

[0089] When the product of the gradient factor and the duration factor is less than 1.3, the load disturbance factor is 1.0;

[0090] When the product of the gradient factor and the duration factor is greater than or equal to 1.3 and less than 1.8, the load disturbance factor is 1.2;

[0091] When the product of the gradient factor and the duration factor is greater than or equal to 1.8 and less than 2.4, the load disturbance factor is 1.6;

[0092] When the product of the gradient factor and the duration factor is greater than 2.4, the load disturbance factor is 2.0;

[0093] Understandably, the gradient variation factor represents the deviation of the gradient variation from the normal value. Due to its calculation method, it is usually greater than 1 when there is a deviation. Similarly, the duration factor represents the duration of the interference excitation period exceeding the limit. It is usually greater than 1 when there is an exceedance. Based on the comprehensive consideration of the product of the gradient variation factor and the duration factor, a threshold value is set and a corresponding load disturbance coefficient is assigned to represent the disturbance intensity.

[0094] Understandably, the baseline value reflects the power level at the end of the disturbance, while the load disturbance factor is based on the overall intensity of the disturbance. By amplifying the baseline value through the load disturbance factor, the final load disturbance excitation value can better characterize the load impact intensity caused by the disturbance event on the generator set.

[0095] Specifically, the operating parameters of each generator in the generator set are restricted, including frequency and voltage.

[0096] The instantaneous frequency deviation peak value of each remaining generator in the generator set and the steady-state recovery time of each generator voltage to the allowable tolerance are collected synchronously.

[0097] The maximum instantaneous frequency deviation peak value and the maximum steady-state recovery time value are defined as the interference disturbance characteristics.

[0098] Specifically, the method of limiting the operating parameters of each generator in the generator set is not specifically limited. It could involve injecting a step disturbance signal into the automatic voltage regulator of the target generator, causing a step change in the excitation current of ±5% of its rated value. Then, under the condition that the excitation voltage does not exceed the generator's leading and lagging phase stability limits, a corresponding step change is caused, thereby generating a disturbance to the generator system. The generator frequency can be directly controlled. Of course, those skilled in the art can also choose other operations that can affect the operating parameters of the generators in the generator set, which will not be elaborated further here.

[0099] Specifically, the instantaneous frequency deviation peak value can be determined by capturing the frequency transient process using a high-precision frequency recording device connected to the bus and extracting the peak value. The steady-state recovery time can be calculated based on the data acquired by the high-precision frequency recorder, determining the time required for the frequency to recover from the maximum deviation point after the disturbance to the allowable tolerance.

[0100] It is understandable that the allowable tolerance can be set at ±0.2Hz according to national standards, and those skilled in the art can also adjust the allowable tolerance according to the specific usage scenarios of the generator set and the power grid.

[0101] Specifically, the process of calculating the interference characterization parameters for the engine assembly based on the load disturbance excitation value and interference disturbance characteristics includes,

[0102] The ratio of the peak value of the instantaneous frequency deviation to the standard value of the instantaneous frequency deviation is determined as the instantaneous frequency parameter.

[0103] The ratio of the maximum steady-state recovery time to the standard steady-state recovery time is determined as the recovery time parameter;

[0104] The weighted summation of the load disturbance excitation value, instantaneous frequency parameter, and recovery time parameter is determined as the interference characterization parameter.

[0105] Specifically, the standard value for steady-state recovery time is determined based on power system stability requirements and the performance of the unit's speed control system. Those skilled in the art can refer to relevant standards and historical test data or simulation results of the unit to set an acceptable maximum time for the frequency to recover from a specified deviation to within the allowable tolerance band and stabilize. In this embodiment of the invention, the reference range for the standard value of steady-state recovery time is [3 seconds, 10 seconds].

[0106] Specifically, the weighting weight of the load disturbance excitation value is 0.4, the weighting weight of the instantaneous frequency parameter is 0.3, and the weighting weight of the recovery time parameter is 0.3.

[0107] Please see Figure 3 As shown, Figure 3 A logic block diagram for generating system interference state labels for the target generator.

[0108] Specifically, assessing the interference state of the target generator affected by system interference includes,

[0109] If the interference characterization parameter is greater than the interference characterization threshold, a strong interference state label is set for the target generator;

[0110] If the interference characterization parameter is less than or equal to the interference characterization threshold, a weak interference state label is set for the target generator.

[0111] Specifically, the interferometric characterization threshold is determined based on the statistical values ​​of historical interferometric characterization parameters. The product of the average value of the historical interferometric characterization parameters over 180 days and the error amplification factor is used as the interferometric characterization threshold. The error amplification factor is selected within the interval [1.05, 1.15].

[0112] Specifically, the test operation characteristics of each generator in several dimensions are obtained during the interference disturbance test. Based on these test operation characteristics, a set of specific dimensions for each generator is constructed, including:

[0113] The test operation characteristics of the target generator are collected in several dimensions, including stator winding temperature, bearing temperature, frequency, vibration, and active power.

[0114] Dimensions with a sliding window standard deviation lower than a predetermined fluctuation threshold are selected as low-interference dimensions;

[0115] Based on the selected low-interference dimensions, a set of specific dimensions for the target generator is constructed.

[0116] Specifically, low-interference dimensions with a sliding window standard deviation below a predetermined fluctuation threshold are selected, including:

[0117] Obtain the sliding window standard deviation of the corresponding values ​​of stator winding temperature, bearing temperature, frequency, vibration, and active power respectively;

[0118] Each sliding window standard deviation is compared with its corresponding preset fluctuation threshold. If the sliding window standard deviation is lower than the fluctuation threshold, the dimension corresponding to the sliding window standard deviation is marked as a low interference dimension.

[0119] Please see Figure 4 As shown, Figure 4 The logic block diagram is marked as low interference dimension;

[0120] The sliding window standard deviation of each dimension is compared with its corresponding preset fluctuation threshold one by one. If the sliding window standard deviation of a dimension is lower than the fluctuation threshold, the dimension is marked as a low interference dimension.

[0121] Specifically, during periods of strong interference, a sliding time window of 5 seconds with a fixed length is used to slide and extract subsequences of sampled values ​​for each dimension of the test operation characteristics. The standard deviation of all sampled values ​​in the subsequence is then calculated to determine the standard deviation of the sliding window.

[0122] Specifically, no specific limit is made on the sampling interval for each dimension within the sliding time window. Those skilled in the art can select an appropriate sampling interval based on the specific circumstances of the unit equipment, which will not be elaborated here.

[0123] Specifically, the preset fluctuation threshold is obtained by performing normal distribution statistics on historical data collected from the target generator set under rated operating conditions, typical cooling conditions, and standard grid disturbance spectrum for at least 30 consecutive days, based on each dimension of the data. The right boundary value of the 95% confidence interval in each dimension of the data is used as the preset fluctuation threshold.

[0124] Specifically, based on a specific set of dimensions, operational characteristics of the generator in corresponding dimensions during daily operation are extracted. Based on these operational characteristics, it is determined whether the generator set exhibits any abnormalities, including...

[0125] Based on the specific dimensional set of the target generator, the corresponding operational characteristics of the dimensions are collected;

[0126] If a parameter in a specific dimension exceeds the preset anomaly threshold for that dimension, the target generator is determined to be abnormal.

[0127] Specifically, the generator's operational characteristics across all dimensions are extracted to determine if any abnormalities exist, including...

[0128] Collect the operating characteristics of the target generator, including stator winding temperature, bearing temperature, frequency, vibration, and active power.

[0129] The aforementioned operational features are input into a pre-trained anomaly detection model to determine if the target generator is abnormal.

[0130] Specifically, the preset anomaly thresholds for each dimension are determined based on historical data of specific dimensions collected from the target generator under normal operating conditions for at least 30 consecutive days. The normal distribution characteristics of the historical data sequence for each dimension are calculated and statistically analyzed according to a fixed time window, and the upper limit satisfying the 95% confidence interval is taken as the preset anomaly threshold.

[0131] Specifically, the construction of the anomaly detection model is not limited. It can be understood that the training process of the anomaly detection model is essentially to use the multi-dimensional operating feature data of the target generator under known normal and known abnormal states to train a discriminative model. Those skilled in the art can use a deep neural network structure to treat the problem as a binary classification task: the input is the multi-dimensional features of the generator at a certain moment or within a certain time window, and the output is the output result of whether the state at that moment / window belongs to "normal" (0) or "abnormal" (1). Based on this, the anomaly monitoring model is trained. The training process of the discriminative model for the classification task is existing technology and is not specifically limited. For the samples, the operating features of the target generator under abnormal conditions and the operating features of the target generator under normal operating conditions can be collected in advance. As samples, the operating features of the target generator under normal operating conditions can be recorded as the control group, and the operating features of the target generator under abnormal operating conditions can be recorded as the sample group and the experimental group. Based on this, the discriminative model is trained so that the model can identify whether the target generator is abnormal based on the operating features of each dimension.

[0132] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for generating emergency maintenance plans for power generation terminals based on artificial intelligence, characterized in that, include: The load characteristics of the power grid to which the generator set is connected are obtained to identify the interference excitation period, and the load disturbance excitation value is determined based on the end value of the interference excitation period; Interference disturbance tests are conducted on generator sets, including limiting the operating parameters of individual generators one by one and obtaining the interference disturbance characteristics of the remaining generators within the limited time. Based on the load disturbance excitation value and interference disturbance characteristics, interference characterization parameters for the engine set are calculated to evaluate the interference state of the target generator affected by system interference. To address the interference state of the generator set, the generator set is monitored, including... The test operation characteristics of each generator in several dimensions are obtained during the interference disturbance test. Based on the test operation characteristics, a set of specific dimensions for each generator is constructed. Based on the set of specific dimensions, the operation characteristics of the generator corresponding to the low interference dimension during daily operation are extracted. Based on the operation characteristics, it is determined whether there is an abnormality in the generator set. Alternatively, extract all dimensions of the generator's operating characteristics to determine if the generator is malfunctioning; The step of constructing a set of specific dimensions for each generator based on test operation characteristics includes: filtering low-interference dimensions based on test operation characteristics, and constructing a set of specific dimensions based on low-interference dimensions. The process of acquiring the load characteristics of the power grid to which the generator set is connected and identifying the interference excitation period includes: Real-time monitoring of the dynamic sequence data of active power at the grid connection point, and calculation of the gradient of active power change; The ratio of the active power change gradient to the standard change gradient is determined as the change gradient factor. Real-time acquisition of power grid frequency offset is used to analyze the duration of frequency offset exceeding the offset threshold. If the gradient change factor is greater than a first predetermined threshold and the frequency duration is greater than a second predetermined threshold, then the time domain segment corresponding to the frequency duration is determined, and the time domain segment is identified as the interference excitation period. The process of determining the load disturbance excitation value based on the end-point value of the interference excitation period includes, The active power value at the end of the interference excitation period is used as the terminal value; The ratio of the terminal endpoint value to the standard terminal power value is used as the base value; The load disturbance coefficient is determined based on the changing gradient factor and the duration of the interference excitation period; The product of the base value and the load disturbance coefficient is determined as the load disturbance excitation value; By individually restricting the operating parameters of each generator, the interference disturbance characteristics of the remaining generators within the restricted time are obtained, including... Limit the operating parameters of each generator in the generator set, including frequency and voltage; The instantaneous frequency deviation peak value of each remaining generator in the generator set and the steady-state recovery time of each generator voltage to the allowable tolerance are collected synchronously. The maximum instantaneous frequency deviation peak value and the maximum steady-state recovery time value are defined as the interference disturbance characteristics; The process of calculating the interference characterization parameters for the engine set based on the load disturbance excitation value and interference disturbance characteristics includes, The ratio of the peak value of the instantaneous frequency deviation to the standard value of the instantaneous frequency deviation is determined as the instantaneous frequency parameter. The ratio of the maximum steady-state recovery time to the standard steady-state recovery time is determined as the recovery time parameter; The weighted summation of the load disturbance excitation value, instantaneous frequency parameter, and recovery time parameter is determined as the interferometric characterization parameter; The test operation characteristics of each generator in several dimensions are obtained during the interference disturbance test. Based on these test operation characteristics, a set of specific dimensions for each generator is constructed, including: The test operation characteristics of the target generator are collected in several dimensions, including stator winding temperature, bearing temperature, frequency, vibration, and active power. Dimensions with a sliding window standard deviation lower than a predetermined fluctuation threshold are selected as low-interference dimensions; Based on the selected low-interference dimensions, a set of specific dimensions for the target generator is constructed. Low-interference dimensions with a sliding window standard deviation below a predetermined fluctuation threshold were selected, including: Obtain the sliding window standard deviation of the corresponding values ​​of stator winding temperature, bearing temperature, frequency, vibration, and active power respectively; Each sliding window standard deviation is compared with its corresponding preset fluctuation threshold. If the sliding window standard deviation is lower than the fluctuation threshold, the dimension corresponding to the sliding window standard deviation is marked as a low interference dimension.

2. The method for generating emergency maintenance plans for power generation terminals based on artificial intelligence according to claim 1, characterized in that, Assess the interference state of the target generator under the influence of system interference, including: If the interference characterization parameter is greater than the interference characterization threshold, a strong interference state label is set for the target generator; If the interference characterization parameter is less than or equal to the interference characterization threshold, a weak interference state label is set for the target generator.

3. The method for generating emergency maintenance plans for power generation terminals based on artificial intelligence according to claim 1, characterized in that, Based on a specific set of dimensions, operational features corresponding to the generator's daily operation are extracted. Based on these operational features, it is determined whether the generator set exhibits any abnormalities, including... Based on the specific dimensional set of the target generator, the corresponding operational characteristics of the dimensions are collected; If a parameter in a specific dimension exceeds the preset anomaly threshold for that dimension, the target generator is determined to be abnormal.

4. The method for generating emergency maintenance plans for power generation terminals based on artificial intelligence according to claim 1, characterized in that, Extract all dimensions of the generator's operational characteristics to determine if the generator is experiencing any anomalies, including: Collect the operating characteristics of the target generator, including stator winding temperature, bearing temperature, frequency, vibration, and active power. The aforementioned operational features are input into a pre-trained anomaly detection model to determine if the target generator is abnormal.

Images