New energy station operation simulation assessment method and system based on comprehensive early warning

Abstract

The invention provides a new energy station operation simulation assessment method and system based on comprehensive early warning, and relates to the field of new energy power generation, and the method comprises the steps: collecting operation and fault data, constructing a parameter propagation network to recognize fluctuation abnormity, and generating a risk propagation path; a turning point is determined according to the fluctuation change rate, and adjustment and disturbance time periods are divided; constructing a staged evolution scene, and injecting a control instruction and an interference signal; and evaluating the adjustment and anti-interference capability based on the response and the deviation between the instability trajectory and the safety boundary, and generating an assessment result. According to the method, early warning and targeted evaluation of the operation risk of the new energy station are realized.

Description

Technical Field

[0001] This invention relates to new energy power generation technology, and in particular to a method and system for simulating and assessing the operation of new energy power plants based on comprehensive early warning. Background Technology

[0002] As the proportion of renewable energy power generation continues to increase, the safe and stable operation of renewable energy power plants is becoming increasingly important to the reliability of the power system. Currently, renewable energy power plants face challenges such as variable weather conditions, frequent equipment failures, and complex control systems, requiring effective early warning mechanisms and assessment methods to ensure their stable operation. Traditional assessments of renewable energy power plant operation are mainly based on static indicators, which are difficult to effectively simulate the dynamic response characteristics under complex operating environments and cannot fully reflect the plant's regulation and disturbance resistance capabilities under abnormal operating conditions.

[0003] Existing assessment methods for new energy power plants mostly employ single-indicator evaluation, often focusing only on static deviations in parameters such as output power and voltage. They lack analysis of the dynamic fluctuation characteristics and propagation patterns of operating parameters, failing to anticipate potential fault risks and evolution paths, leading to untimely warnings and inaccurate measures. Traditional assessment methods fail to differentiate the challenges faced by new energy power plants at different operational stages, particularly distinguishing between periods requiring system self-regulation and periods subject to external disturbances. This results in assessments that do not accurately reflect the plant's ability to cope with different fault types and development stages, lacking specificity. The existing assessment system lacks a comprehensive evaluation mechanism for the dynamic response capabilities of new energy power plants, failing to simultaneously consider the plant's regulation and disturbance rejection capabilities. It also cannot verify the plant's comprehensive response capabilities through simulation of real-world fault scenarios, resulting in significant discrepancies between assessment results and actual operating conditions, and failing to provide effective guidance for plant optimization. Summary of the Invention

[0004] This invention provides a method and system for simulating and assessing the operation of new energy power stations based on comprehensive early warning, which can solve the problems in the prior art.

[0005] A first aspect of this invention provides a method for simulating and evaluating the operation of new energy power stations based on comprehensive early warning, comprising: Collect real-time operation data and historical fault data of new energy power stations; Extract the fluctuation characteristics of each operating parameter before the failure from historical failure data, calculate the correlation coefficient between each fluctuation characteristic, and construct a parameter propagation network based on the correlation coefficient; Real-time operational data is input into the parameter propagation network to identify initial fluctuation anomalies, and propagation sequences related to the initial fluctuation anomalies are extracted from the parameter propagation network to generate risk propagation paths. Extract the rate of change of fluctuation of the parameters running on the risk propagation path from historical failure data, determine the fluctuation inflection point, and divide the adjustment period and disturbance period based on the fluctuation inflection point; Based on the risk propagation path, a phased evolution scenario is constructed. Corrective control commands are injected during the adjustment period, and quantitative interference signals are injected during the disturbance period. The response trajectory of the corrective control commands and the instability trajectory of the quantitative interference signals are recorded. The dynamic deviation between the response trajectory and the preset safety boundary is calculated to obtain the regulation capability score, and the dynamic deviation between the instability trajectory and the preset safety boundary is calculated to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result.

[0006] Extracting the fluctuation characteristics of various operating parameters before the fault occurred from historical fault data, calculating the correlation coefficients between the fluctuation characteristics, and constructing a parameter propagation network based on the correlation coefficients includes: Locate the fault trigger time from historical fault data, extract a historical data segment of a preset duration from the fault trigger time as the end time, extract the fluctuation amplitude and fluctuation period of each operating parameter within the historical data segment, and combine them to form the fluctuation characteristics of each operating parameter. Calculate the time delay correlation coefficient between the fluctuation characteristics of any two operating parameters, and screen the operating parameter pairs whose time delay correlation coefficient reaches a preset strength; Determine which operating parameter fluctuates first and which fluctuates later in the operating parameter pair. Set the operating parameter that fluctuates first as the propagation source node and the operating parameter that fluctuates later as the propagation target node. Use the time delay correlation coefficient between the propagation source node and the propagation target node as the inter-node correlation coefficient. A parameter propagation network is constructed based on the source node, the target node, and the correlation coefficient between the nodes. The nodes of the parameter propagation network are the operating parameters, the edges are the connections from the source node to the target node, and the weight of the edge is the correlation coefficient between the nodes.

[0007] Real-time operational data is input into the parameter propagation network to identify initial fluctuation anomalies. Propagation sequences related to the initial fluctuation anomalies are extracted from the parameter propagation network to generate risk propagation paths, including: The real-time running data is input into the parameter propagation network, the real-time fluctuation characteristics of each running parameter in the real-time running data are extracted, the deviation between the real-time fluctuation characteristics and the fluctuation characteristics of the corresponding nodes in the parameter propagation network is calculated, the nodes whose deviation exceeds the preset deviation threshold are marked as the initial fluctuation anomalies, and the fluctuation direction of the initial fluctuation anomalies is recorded. Based on the node position of the initial fluctuation anomaly in the parameter propagation network, search for downstream nodes along the directed edges, extract the real-time fluctuation characteristics of the corresponding operating parameters of the downstream nodes, calculate the fluctuation direction corresponding to the real-time fluctuation characteristics of the downstream nodes, and select downstream nodes whose fluctuation direction is consistent with the fluctuation direction of the initial fluctuation anomaly to form a candidate propagation node set. Extract the correlation coefficients between nodes in the parameter propagation network for each node in the candidate propagation node set, calculate the path correlation strength from the initial fluctuation anomaly to each node in the candidate propagation node set, and filter the nodes and corresponding propagation paths with path correlation strength greater than a preset strength threshold to form a propagation sequence related to the initial fluctuation anomaly. Risk propagation paths are generated based on the propagation hierarchy and path relevance of nodes in the propagation sequence.

[0008] Extract the rate of change of fluctuations in parameters along the risk propagation path from historical failure data, determine the inflection points of fluctuations, and divide the adjustment period and disturbance period based on the inflection points: Construct sliding observation windows of different durations, extract fluctuation data of operating parameters on the risk propagation path, calculate the mean and variance of the fluctuation data, select the optimal observation window according to the minimum variance criterion, and use the optimal observation window to resample the operating parameters on the risk propagation path to obtain the fluctuation sampling sequence. The fluctuation sampling sequence is decomposed into multiple fluctuation components, the energy distribution value of the multiple fluctuation components is calculated, the target fluctuation component is selected according to the energy distribution value, and the target fluctuation component is recombined to generate a fluctuation change sequence. Calculate the change between adjacent sampling points in the fluctuation change sequence to obtain the fluctuation change rate of the running parameters on the risk propagation path. Segment the fluctuation change rate, identify the change direction of the fluctuation change rate, and mark the sampling points where the change direction reverses as fluctuation inflection candidate points. The amplitude ratio and duration of fluctuation inflection candidate points are calculated to form fluctuation morphology characteristics, and fluctuation inflection points are obtained by screening based on the fluctuation morphology characteristics. The fluctuation sequence is segmented based on the turning point of the fluctuation. The time period when the fluctuation rate changes from negative to positive is divided into the adjustment period, and the time period when the fluctuation rate changes from positive to negative is divided into the disturbance period.

[0009] A phased evolution scenario is constructed based on the risk propagation path. Corrective control commands are injected during the adjustment period, and quantized disturbance signals are injected during the disturbance period. The response trajectories of the corrective control commands and the instability trajectories of the quantized disturbance signals are recorded, including: Collect state variables of adjacent control nodes along the risk propagation path, calculate the mean and fluctuation amplitude of the state variables, determine the evolution boundary, and divide the evolution boundary into regulation evolution scenario and disturbance evolution scenario based on the fluctuation inflection point; In the context of the regulation and evolution scenario, a correction control command is generated, the fluctuation frequency and fluctuation amplitude of the correction control command are calculated, and the correction control command with the minimum fluctuation frequency and the fluctuation amplitude that meets the preset fluctuation range is injected into the adjacent control node at time intervals to obtain the correction input signal of the control node. Collect the state response data corresponding to the correction input signal of the control node, calculate the rate of change and acceleration of the state response data, construct the response trajectory, and calculate the duration of the correction control command based on the response trajectory; Within the disturbance evolution scenario, the fluctuation period is divided based on the duration of action to generate a quantized interference signal. The amplitude distribution of the quantized interference signal is calculated, and the quantized interference signal whose amplitude distribution meets the preset distribution range is injected into the adjacent control node to obtain the interference input of the control node. Based on the disturbance input of the control node, the state fluctuation data is collected, the amplitude change rate and frequency offset of the state fluctuation data are calculated, the instability trajectory is constructed, and the response trajectory and instability trajectory are output.

[0010] In a disturbance evolution scenario, generating quantized interference signals by dividing the fluctuation period based on the duration of action includes: The duration of action is divided into multiple fluctuation periods according to a preset ratio coefficient, and the average period of each fluctuation period is calculated. A cosine waveform reference signal is generated within the mean period, and the peak and trough positions of the reference signal are used as feature points to construct a time series of feature points. The time series of feature points are subjected to nonlinear transformation to generate multiple sets of distorted feature point sequences. The reference signal is reconstructed in segments based on the distorted feature point sequences to obtain multiple sets of distorted interference signals. Calculate the spectral distribution and energy distribution of the distorted interference signal, and construct the wave characteristic vector of the distorted interference signal; The matching degree between the fluctuation feature vector and the preset disturbance template is calculated, and the distortion interference signal with the highest matching degree is selected as the quantized interference signal.

[0011] The dynamic deviation between the response trajectory and the preset safety boundary is calculated to obtain the regulation capability score, and the dynamic deviation between the instability trajectory and the preset safety boundary is calculated to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result, which includes: Extract the upper and lower threshold values ​​and standard frequency values ​​of the preset safety boundary to generate safety boundary parameters; The response trajectory is segmented according to the sampling time sequence, and the fluctuation deviation between each segment of the response trajectory and the safety boundary parameter is calculated to construct the fluctuation amplitude deviation sequence and the fluctuation frequency deviation sequence. The cumulative duration of the amplitude exceeding the safety boundary parameter in the fluctuation amplitude deviation sequence is calculated to obtain the amplitude exceeding the limit duration, and the cumulative magnitude of the deviation from the standard frequency value in the fluctuation frequency deviation sequence is calculated to obtain the frequency offset. The duration of amplitude exceeding the limit and the frequency offset are converted according to a preset mapping coefficient to generate a regulation capability score; The instability trajectory is segmented according to the fluctuation period, and the fluctuation change rate of each segment of the instability trajectory relative to the safety boundary parameter is calculated to construct the fluctuation rate sequence and the fluctuation acceleration sequence. Calculate the number of abrupt changes in the fluctuation rate sequence and the transition amplitude in the fluctuation acceleration sequence, and convert the number of abrupt changes and the transition amplitude according to a preset mapping coefficient to generate an anti-disturbance capability score; The assessment results are generated by combining the adjustment capability score and the disturbance resistance score according to a preset weighting.

[0012] A second aspect of this invention provides a new energy power station operation simulation and assessment system based on comprehensive early warning, comprising: The first unit is used to collect real-time operation data and historical fault data of new energy power stations; The second unit is used to extract the fluctuation characteristics of each operating parameter before the occurrence of the fault from historical fault data, calculate the correlation coefficient between each fluctuation characteristic, and construct the parameter propagation network based on the correlation coefficient. The third unit is used to input real-time operating data into the parameter propagation network, identify initial fluctuation anomalies, extract propagation sequences related to initial fluctuation anomalies from the parameter propagation network, and generate risk propagation paths. The fourth unit is used to extract the rate of change of fluctuations of the parameters running on the risk propagation path from historical fault data, determine the inflection point of fluctuations, and divide the adjustment period and the disturbance period based on the inflection point of fluctuations. The fifth unit is used to construct a phased evolution scenario based on the risk propagation path, inject corrective control commands during the adjustment period, inject quantitative interference signals during the disturbance period, and record the response trajectory of the corrective control commands and the instability trajectory of the quantitative interference signals. The sixth unit is used to calculate the dynamic deviation between the response trajectory and the preset safety boundary to obtain the regulation capability score, and to calculate the dynamic deviation between the instability trajectory and the preset safety boundary to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result.

[0013] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] In this embodiment, by extracting parameter fluctuation characteristics from historical fault data and constructing a parameter propagation network, early identification of abnormal states and prediction of risk propagation paths are achieved, improving the accuracy and foresight of fault warnings. The method of determining fluctuation inflection points based on the fluctuation change rate and dividing the adjustment period and disturbance period makes the assessment process more closely resemble actual operating scenarios, improving the realism and relevance of the simulation. By constructing a phased evolution scenario and injecting corrective control commands and quantized interference signals at different time periods, a comprehensive assessment of the regulation and disturbance rejection capabilities of new energy power plants is achieved. The dynamic deviation calculation method is used to evaluate the performance of the response trajectory and instability trajectory relative to the safety boundary. Combining the regulation capability score and the disturbance rejection capability score to generate the assessment result helps guide new energy power plants to improve operational stability and reliability. Replacing actual fault testing with simulated assessment reduces safety risks and provides technical support for the safe operation and performance optimization of new energy power plants. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the simulation assessment method for the operation of new energy power stations based on comprehensive early warning, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the risk propagation path control logic of an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0018] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0019] Figure 1 This is a flowchart illustrating the simulation and assessment method for the operation of new energy power stations based on comprehensive early warning, as described in an embodiment of the present invention. Figure 1 As shown, the method includes: Collect real-time operation data and historical fault data of new energy power stations; Extract the fluctuation characteristics of each operating parameter before the failure from historical failure data, calculate the correlation coefficient between each fluctuation characteristic, and construct a parameter propagation network based on the correlation coefficient; Real-time operational data is input into the parameter propagation network to identify initial fluctuation anomalies, and propagation sequences related to the initial fluctuation anomalies are extracted from the parameter propagation network to generate risk propagation paths. Extract the rate of change of fluctuation of the parameters running on the risk propagation path from historical failure data, determine the fluctuation inflection point, and divide the adjustment period and disturbance period based on the fluctuation inflection point; Based on the risk propagation path, a phased evolution scenario is constructed. Corrective control commands are injected during the adjustment period, and quantitative interference signals are injected during the disturbance period. The response trajectory of the corrective control commands and the instability trajectory of the quantitative interference signals are recorded. The dynamic deviation between the response trajectory and the preset safety boundary is calculated to obtain the regulation capability score, and the dynamic deviation between the instability trajectory and the preset safety boundary is calculated to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result.

[0020] In one optional implementation, the fluctuation characteristics of each operating parameter before the fault occurred are extracted from historical fault data, the correlation coefficient between each fluctuation characteristic is calculated, and a parameter propagation network is constructed based on the correlation coefficient, including: Locate the fault trigger time from historical fault data, extract a historical data segment of a preset duration from the fault trigger time as the end time, extract the fluctuation amplitude and fluctuation period of each operating parameter within the historical data segment, and combine them to form the fluctuation characteristics of each operating parameter. Calculate the time delay correlation coefficient between the fluctuation characteristics of any two operating parameters, and screen the operating parameter pairs whose time delay correlation coefficient reaches a preset strength; Determine which operating parameter fluctuates first and which fluctuates later in the operating parameter pair. Set the operating parameter that fluctuates first as the propagation source node and the operating parameter that fluctuates later as the propagation target node. Use the time delay correlation coefficient between the propagation source node and the propagation target node as the inter-node correlation coefficient. A parameter propagation network is constructed based on the source node, the target node, and the correlation coefficient between the nodes. The nodes of the parameter propagation network are the operating parameters, the edges are the connections from the source node to the target node, and the weight of the edge is the correlation coefficient between the nodes.

[0021] In the process of locating the fault trigger point, the precise trigger point is determined by analyzing the abnormal indicator change patterns in historical fault data. Specifically, the numerical change trends of various operating parameters are monitored, and when a parameter value exceeds the normal operating range or undergoes a drastic change, that moment is recorded as a potential fault trigger point. Through cross-validation of multiple parameters, the precise moment of fault triggering is finally determined. Taking a wind power station as an example, when the vibration amplitude of the wind turbine suddenly increases from the normal 2 mm to 8 mm, and the bearing temperature rises from 65 degrees Celsius to 85 degrees Celsius, and this change is completed within 30 seconds, the moment when the vibration amplitude begins to change abnormally is defined as the fault trigger point.

[0022] After determining the fault trigger time, historical data segments of a preset duration are extracted and analyzed, using this time as the endpoint. The preset duration is determined based on the equipment type and fault characteristics; for wind power equipment, a data segment typically selected is 24 to 72 hours prior to the fault. Within this time window, the fluctuation amplitude and fluctuation period characteristics of each operating parameter are extracted. The fluctuation amplitude is quantified by calculating the standard deviation or range of the parameter values, and the fluctuation period is identified by using a Fast Fourier Transform to determine the main frequency components of parameter changes. For the wind turbine gearbox oil temperature parameter, in the 48 hours prior to the fault, its fluctuation amplitude might be 5.2 degrees Celsius, with the main fluctuation periods being 6 hours and 12 hours. These values ​​combined constitute the fluctuation characteristic vector of this parameter.

[0023] The cross-correlation function method is used to calculate the correlation coefficient between the fluctuation characteristics of different operating parameters and the time delay relationship. For any two fluctuation characteristic sequences of operating parameters, the correlation coefficient under different time delays is calculated through a sliding time window, and the maximum value of the correlation coefficient and its corresponding time delay are identified. The mathematical expression of the cross-correlation function is the normalized covariance of two signals under different time delays, and its value ranges from -1 to 1. When the maximum time delay correlation coefficient between the wind turbine speed fluctuation characteristics and the generator temperature fluctuation characteristics is 0.82, corresponding to a time delay of 15 minutes, it indicates that the speed fluctuation has a strong predictive effect on the generator temperature change.

[0024] When screening pairs of operating parameters, a preset threshold for the correlation coefficient of time delay is set, typically above 0.6. Only when the maximum time delay correlation coefficient of two operating parameters exceeds this threshold are they considered meaningful parameter pairs for further analysis. This screening mechanism ensures that the propagation network contains only parameters with significant correlations, avoiding the influence of weak correlations or noise interference.

[0025] The direction of propagation in a pair of operating parameters is determined by analyzing the temporal sequence of fluctuations. For parameter pairs with established correlation, the starting times of their fluctuation characteristic changes are compared. The parameter that fluctuates significantly first is identified as the propagation source, and the parameter that fluctuates later is identified as the propagation target. The determination method is based on time series analysis of the parameter's rate of change; the moment when the rate of change of the parameter first exceeds a set threshold is the start of the fluctuation. In the relationship between gearbox vibration and oil temperature, if the vibration amplitude begins to increase 2 hours before the fault, while the oil temperature begins to rise 1.5 hours before the fault, then the vibration parameter is the propagation source node, and the oil temperature parameter is the propagation target node.

[0026] The propagation network is constructed by treating operating parameters as network nodes and the propagation relationships between parameters as directed edges. Each node represents a specific operating parameter, and its attributes include parameter type, normal value range, and fluctuation characteristics. Directed edges connect the source node and the target node, with the edge direction indicating the direction of propagation influence. The edge weight is set as the time delay correlation coefficient between the corresponding parameter pairs. The propagation network constructed in this way can clearly demonstrate the path and intensity of fault propagation, providing important information for fault early warning and diagnosis.

[0027] After the network is constructed, critical propagation paths and core nodes can be identified using network topology analysis. Nodes with high degree centrality are typically critical parameters, and their abnormal fluctuations have a significant impact on the overall operational status. Nodes with high betweenness centrality act as bridges during fault propagation and are key monitoring targets. By analyzing the structural characteristics of the propagation network, monitoring strategies can be optimized, focusing on parameter changes along critical propagation paths.

[0028] In this embodiment, a propagation relationship network among parameters is established, enabling the visualization and quantitative analysis of the fault propagation mechanism. The propagation network not only reveals the causal relationships and influence strength among various operating parameters but also provides a scientific basis for early fault warning. By monitoring the state changes of the propagation source nodes, abnormal trends in downstream parameters can be predicted in advance, thereby achieving more accurate and timely fault warnings. The propagation network provides clear path guidance for fault tracing analysis, helping to quickly locate the root cause of the fault and improve fault handling efficiency.

[0029] In one optional implementation, real-time operational data is input into a parameter propagation network to identify initial fluctuation anomalies. Propagation sequences related to these initial fluctuation anomalies are extracted from the parameter propagation network to generate risk propagation paths, including: The real-time running data is input into the parameter propagation network, the real-time fluctuation characteristics of each running parameter in the real-time running data are extracted, the deviation between the real-time fluctuation characteristics and the fluctuation characteristics of the corresponding nodes in the parameter propagation network is calculated, the nodes whose deviation exceeds the preset deviation threshold are marked as the initial fluctuation anomalies, and the fluctuation direction of the initial fluctuation anomalies is recorded. Based on the node position of the initial fluctuation anomaly in the parameter propagation network, search for downstream nodes along the directed edges, extract the real-time fluctuation characteristics of the corresponding operating parameters of the downstream nodes, calculate the fluctuation direction corresponding to the real-time fluctuation characteristics of the downstream nodes, and select downstream nodes whose fluctuation direction is consistent with the fluctuation direction of the initial fluctuation anomaly to form a candidate propagation node set. Extract the correlation coefficients between nodes in the parameter propagation network for each node in the candidate propagation node set, calculate the path correlation strength from the initial fluctuation anomaly to each node in the candidate propagation node set, and filter the nodes and corresponding propagation paths with path correlation strength greater than a preset strength threshold to form a propagation sequence related to the initial fluctuation anomaly. Risk propagation paths are generated based on the propagation hierarchy and path relevance of nodes in the propagation sequence.

[0030] In the processing of real-time operational data input parameter propagation networks, it is necessary to collect and extract features from various sensor data of new energy power plants in real time. Sensor data includes key operational parameters such as wind turbine speed, blade angle, gearbox temperature, generator current, and transformer load. For each operational parameter, its real-time fluctuation characteristics are extracted using a sliding time window method. The length of the time window is determined based on the parameter's response characteristics, typically ranging from 15 minutes to 2 hours. Real-time fluctuation characteristics include two dimensions: fluctuation amplitude and fluctuation frequency. The fluctuation amplitude is obtained by calculating the standard deviation of the data within the time window, and the fluctuation frequency is extracted using a short-time Fourier transform to obtain the main frequency components. When the standard deviation of the wind turbine speed within a 30-minute time window is 12 revolutions per minute, and the main fluctuation frequency is 0.05 Hz, these values ​​constitute the real-time fluctuation feature vector of the speed parameter.

[0031] The deviation calculation process compares and analyzes the real-time fluctuation characteristics with the historical fluctuation characteristics of the corresponding nodes in the parameter propagation network. The historical fluctuation characteristics are derived from the training data used to construct the parameter propagation network, representing the fluctuation patterns of each operating parameter under normal conditions. The deviation calculation employs Euclidean distance or cosine similarity methods to quantify the degree of difference between real-time and historical characteristics. The Euclidean distance formula is the square root of the sum of the squares of the differences between the components of two feature vectors, while cosine similarity reflects the degree of similarity through the cosine of the angle between the two vectors. For the gearbox oil temperature parameter, when its real-time fluctuation feature vector is [3.2, 0.08] and the historical normal fluctuation feature vector is [1.8, 0.03], the calculated Euclidean distance is 1.46, exceeding the preset deviation threshold of 1.2. Therefore, this node is marked as having an initial fluctuation anomaly.

[0032] The direction of fluctuation is recorded by analyzing the trend of parameter value changes. Fluctuation direction is categorized into three types: upward, downward, and oscillating. This is determined by calculating the first derivative of the parameter value within a time window or the slope of a linear regression. An upward trend is defined as a positive slope exceeding a set threshold, a downward trend as a negative slope exceeding the threshold in absolute value, and an oscillating state as a slope in absolute value less than the threshold. For generator temperature parameters, upon detecting an initial abnormal fluctuation, if the temperature change rate within 30 minutes increases by 0.15 degrees Celsius per minute, the fluctuation direction is recorded as upward.

[0033] Downstream node search is performed based on the directed edge structure of the parameter propagation network. Starting from the initial fluctuation anomaly node, all directly connected downstream nodes are traversed along the direction of the directed edges. A breadth-first search algorithm is used during traversal to ensure that downstream nodes are visited in an orderly manner according to the propagation hierarchy. For each downstream node, the real-time fluctuation characteristics of its corresponding operating parameters are extracted, and its current fluctuation direction is calculated. In photovoltaic power plants, when the inverter DC voltage exhibits an initial fluctuation anomaly and the fluctuation direction is decreasing, the downstream nodes searched include parameter nodes such as inverter AC output power, grid-side voltage, and power factor.

[0034] The selection of candidate propagation nodes is based on the principle of fluctuation direction consistency. Only when the real-time fluctuation direction of a downstream node is consistent with the initial fluctuation anomaly direction is it included in the candidate propagation node set. A certain time delay is allowed in the fluctuation direction consistency determination because signal propagation takes time. The delay time is determined based on the physical correlation between parameters; the delay between electrical parameters is typically on the order of seconds, while the delay between mechanical parameters may be on the order of minutes. In energy storage battery management, when the voltage of a single battery cell drops abnormally, if the voltage of its downstream battery module and the total voltage of the battery pack also show a downward trend within 5 minutes, these nodes are added to the candidate propagation node set.

[0035] The calculation of path correlation strength comprehensively considers the weights of each edge along the propagation path and the path length. For the path from the initial fluctuation anomaly node to the candidate propagation node, the overall propagation strength is calculated by multiplying the inter-node correlation coefficients of each edge along the path. To avoid excessive attenuation of the correlation strength for long paths, a path length correction factor is introduced, which is a negative exponential function of the path length. The formula for calculating the corrected path correlation strength is the original propagation strength multiplied by the path length correction factor. When the path from the wind turbine blade vibration node to the gearbox wear node contains two edges with edge weights of 0.75 and 0.68 respectively, the original propagation strength is 0.51. After considering a correction factor of 0.9 with a path length of 2, the final path correlation strength is 0.459.

[0036] The propagation sequence is constructed by selecting nodes and their corresponding propagation paths whose path correlation strength is greater than a preset threshold. The preset threshold is set according to the type of site and the required early warning accuracy, and generally ranges from 0.3 to 0.6. Nodes that meet the criteria are sorted according to the propagation level to form an ordered propagation sequence. The propagation level is determined based on the shortest path length of the node from the initial anomaly node in the propagation network, and nodes within the same level are arranged in descending order of path correlation strength.

[0037] The risk propagation path generation process integrates node information and path correlation strength data from the propagation sequence. The propagation path is represented as a directed graph, where nodes represent operational parameters, edges represent propagation relationships, and the thickness of the edges reflects the magnitude of the propagation strength. The path generation algorithm prioritizes propagation paths with high correlation strength while maintaining path continuity and integrity. For cases with multiple parallel propagation paths, a weighted merging based on path correlation strength is used to form a comprehensive risk assessment result.

[0038] In this embodiment, dynamic monitoring and visualization of the propagation path of operational risks at new energy power plants are achieved. By identifying initial fluctuation anomalies and tracking the risk propagation process in real time, early warnings can be provided before a full-blown fault occurs, effectively reducing the risk of equipment damage. Propagation path analysis reveals the intrinsic mechanism of fault spread, providing a basis for formulating precise response strategies. Quantitative assessment of path correlation strength makes risk level classification more objective and accurate, improving the reliability and practicality of early warnings.

[0039] In one optional implementation, the rate of change of fluctuations in parameters along the risk propagation path is extracted from historical fault data to determine the inflection point of fluctuations. Based on the inflection point, the adjustment period and the disturbance period are divided, including: Construct sliding observation windows of different durations, extract fluctuation data of operating parameters on the risk propagation path, calculate the mean and variance of the fluctuation data, select the optimal observation window according to the minimum variance criterion, and use the optimal observation window to resample the operating parameters on the risk propagation path to obtain the fluctuation sampling sequence. The fluctuation sampling sequence is decomposed into multiple fluctuation components, the energy distribution value of the multiple fluctuation components is calculated, the target fluctuation component is selected according to the energy distribution value, and the target fluctuation component is recombined to generate a fluctuation change sequence. Calculate the change between adjacent sampling points in the fluctuation change sequence to obtain the fluctuation change rate of the running parameters on the risk propagation path. Segment the fluctuation change rate, identify the change direction of the fluctuation change rate, and mark the sampling points where the change direction reverses as fluctuation inflection candidate points. The amplitude ratio and duration of the fluctuation inflection candidate points are calculated to form the fluctuation pattern characteristics. Fluctuation inflection points are then selected based on the fluctuation pattern characteristics. The fluctuation sequence is segmented based on the turning point of the fluctuation. The time period when the fluctuation rate changes from negative to positive is divided into the adjustment period, and the time period when the fluctuation rate changes from positive to negative is divided into the disturbance period.

[0040] The duration of the sliding observation window can be set to different time scales such as 5 minutes, 10 minutes, 15 minutes, and 30 minutes. For the risk propagation path data within each sliding observation window, the mean and variance of the fluctuation data are calculated. The mean μ is calculated as the arithmetic mean of all data points within the window, and the variance σ... 2 The variance is calculated as the sum of the squares of the differences between each data point and the mean, divided by the number of data points. The optimal observation window is selected based on the minimum variance criterion; that is, the sliding observation window with the smallest variance is chosen as the optimal observation window. For photovoltaic power generation scenarios, when the incident irradiance parameter fluctuates significantly over a certain period, the variance of a 15-minute window might be 0.023, while the variance of a 10-minute window is 0.018, and the variance of a 30-minute window is 0.035. In this case, a 10-minute window is selected as the optimal observation window. The operating parameters along the risk propagation path are resampled using the selected optimal observation window to obtain a fluctuating sampling sequence. An equal-interval sampling method is used during the resampling process to ensure that the sampling points are evenly distributed along the time axis.

[0041] The fluctuating sample sequence is decomposed into multiple fluctuating components through hierarchical decomposition. Hierarchical decomposition employs wavelet transform to break down the fluctuating sample sequence into fluctuating components at different frequency levels. Specifically, the Daubechies wavelet function is used for a 5-level decomposition, yielding 5 detail components D1 to D5 and 1 approximate component A5. The energy distribution value of each fluctuating component is calculated by dividing the sum of squares of the component coefficients by the total energy. For wind farm wind speed parameters, the energy percentages might be: D1 8%, D2 15%, D3 42%, D4 22%, D5 9%, and A5 4%. Target fluctuating components are selected based on their energy distribution values, choosing those with an energy percentage exceeding a preset threshold. If the threshold is set to 20%, then D3 and D4 are the target fluctuating components. The selected target fluctuating components are then recombined using an inverse transform to generate a fluctuating change sequence.

[0042] The change in the fluctuation sequence between adjacent sampling points is calculated, i.e., the rate of change of the operating parameters along the risk propagation path is obtained by calculating the difference in parameter values ​​at adjacent time points. For inverter output voltage parameters, if the interval between two adjacent sampling points is 1 minute, the voltage is 220.5V at time t and 221.2V at time t+1, then the rate of change at that time point is 0.7V / min. The rate of change is segmented, and a piecewise linear fitting method is used to divide the rate of change sequence into multiple linear segments. The direction of change of the rate of change is identified by calculating the derivative or first-order difference of the rate of change to determine its trend. Sampling points where the direction of change reverses are marked as candidate points of fluctuation inflection, i.e., points where the rate of change changes from rising to falling or from falling to rising. For the charging and discharging power parameters of the energy storage system, the rate of change may be +2.5kW / min at time t1, +0.8kW / min at time t2, and -1.2kW / min at time t3, then time t3 is a candidate point of fluctuation inflection.

[0043] The fluctuation amplitude ratio and duration are calculated for candidate fluctuation inflection points to constitute fluctuation morphology characteristics. The fluctuation amplitude ratio is calculated as the ratio of the peak values ​​before and after the candidate point, and the duration is calculated as the time it takes for the fluctuation to move from one peak to the next. For photovoltaic module temperature parameters, if a candidate point has a peak value of 45.2℃ before and a peak value of 42.8℃ after, with a duration of 25 minutes, then the fluctuation amplitude ratio is 1.056 and the duration is 25 minutes. Fluctuation inflection points are selected based on fluctuation morphology characteristics. The selection criteria are that candidate points with a fluctuation amplitude ratio exceeding 1.05 and a duration exceeding 20 minutes are identified as fluctuation inflection points.

[0044] The fluctuation sequence is segmented based on the identified inflection points, with each inflection point serving as a segment boundary. The time period when the fluctuation rate changes from negative to positive is designated as the adjustment period, and the time period when the fluctuation rate changes from positive to negative is designated as the disturbance period. For the output power parameters of a grid-connected photovoltaic inverter, if the fluctuation rate changes from -1.8 kW / min to +2.2 kW / min at 9:15 and from +1.5 kW / min to -2.5 kW / min at 10:40, then 9:15 to 10:40 is the adjustment period, and the period after 10:40 is the disturbance period.

[0045] In this embodiment, by constructing an optimal observation window and multi-layer wavelet decomposition, key fluctuation features of operating parameters along the risk propagation path can be effectively extracted, fluctuation inflection points can be accurately identified, and accurate division of adjustment and disturbance periods can be achieved. This data-driven fluctuation feature extraction and segmentation method can significantly improve the accuracy of identifying the operating status of new energy power plants, providing a reliable basis for subsequent fault early warning and operational assessment. It can adapt to the volatility and intermittency of new energy power generation, reduce false alarm and false negative rates, improve the generalization ability and robustness of the early warning model, and provide technical support for the safe and efficient operation of new energy power plants.

[0046] like Figure 2 The diagram shown illustrates the risk propagation path control logic flowchart of this embodiment.

[0047] In one optional implementation, a phased evolution scenario is constructed based on the risk propagation path. Corrective control commands are injected during the adjustment period, and quantized interference signals are injected during the disturbance period. The response trajectory of the corrective control commands and the instability trajectory of the quantized interference signals are recorded, including: Collect state variables of adjacent control nodes along the risk propagation path, calculate the mean and fluctuation amplitude of the state variables, determine the evolution boundary, and divide the evolution boundary into regulation evolution scenario and disturbance evolution scenario based on the fluctuation inflection point; In the context of the regulation and evolution scenario, a correction control command is generated, the fluctuation frequency and fluctuation amplitude of the correction control command are calculated, and the correction control command with the minimum fluctuation frequency and the fluctuation amplitude that meets the preset fluctuation range is injected into the adjacent control node at time intervals to obtain the correction input signal of the control node. Collect the state response data corresponding to the correction input signal of the control node, calculate the rate of change and acceleration of the state response data, construct the response trajectory, and calculate the duration of the correction control command based on the response trajectory; Within the disturbance evolution scenario, the fluctuation period is divided based on the duration of action to generate a quantized interference signal. The amplitude distribution of the quantized interference signal is calculated, and the quantized interference signal whose amplitude distribution meets the preset distribution range is injected into the adjacent control node to obtain the interference input of the control node. Based on the disturbance input of the control node, the state fluctuation data is collected, the amplitude change rate and frequency offset of the state fluctuation data are calculated, the instability trajectory is constructed, and the response trajectory and instability trajectory are output.

[0048] The collected state variables include operating parameters such as voltage, current, power, and frequency. The mean and fluctuation amplitude of the state variables are calculated. The mean is calculated as the arithmetic mean of the collected data within the time window, and the fluctuation amplitude is calculated as the difference between the maximum and minimum values. Evolutionary boundaries are determined, i.e., the upper and lower limits of normal operation for each state variable. For example, the normal operating range of AC voltage for a photovoltaic power station is 198V to 235V. Based on fluctuation inflection points, the evolutionary boundaries are divided into regulation evolution scenarios and disturbance evolution scenarios. The regulation evolution scenario corresponds to the time period when the fluctuation rate changes from negative to positive, and the disturbance evolution scenario corresponds to the time period when the fluctuation rate changes from positive to negative. For the DC bus voltage parameters of the photovoltaic power station, when the rate of change of fluctuation changes from -3.5V / min to +2.8V / min from 9:30 to 11:45 AM, and from 11:45 to 2:20 PM from +2.8V / min to -4.2V / min, the period from 9:30 to 11:45 AM is classified as the regulation evolution scenario, and the period from 11:45 to 2:20 PM is classified as the disturbance evolution scenario.

[0049] Corrective control commands are generated within the regulation and evolution scenario. These commands are generated using a Fourier transform method, specifically by performing a Fourier transform on historical data of state variables within the scenario, extracting the main frequency components, and reconstructing the control signal based on these components. The fluctuation frequency and amplitude of the corrective control commands are calculated. The fluctuation frequency is calculated as the number of fluctuation cycles per unit time, and the fluctuation amplitude is calculated as half the difference between the peak and trough. Corrective control commands with the minimum fluctuation frequency and amplitude meeting the preset fluctuation range are injected into adjacent control nodes at time intervals. For wind farm pitch angle control, corrective control commands with fluctuation frequencies of 0.05Hz, 0.1Hz, and 0.15Hz are generated, corresponding to fluctuation amplitudes of 1.2 degrees, 0.8 degrees, and 0.6 degrees, respectively. The preset fluctuation range is 0.5 degrees to 1.5 degrees. Therefore, a corrective control command with a fluctuation frequency of 0.05Hz and a fluctuation amplitude of 1.2 degrees is selected and injected into the wind turbine control system at 5-second time intervals to obtain the corrective input signal for the control node.

[0050] The system collects state response data corresponding to the corrected input signal from the control node. This data includes changes in state variables after the control node receives the corrected control command. The rate of change and acceleration of the state response data are calculated. The rate of change is the amount of change of the state variable per unit time, and the acceleration is the rate of change of the rate of change. For active power control of a photovoltaic inverter, after the power adjustment correction command is injected, the inverter output power is collected to increase from 80kW to 85kW within 10 seconds at a rate of 0.5kW / s; in the following 10 seconds, the power increases from 85kW to 87kW at a rate of 0.2kW / s, with an acceleration of -0.03kW / s². A response trajectory is constructed, represented as a curve of the state variable changing over time. Discrete state response data points are connected using spline interpolation to form a continuous trajectory. The duration of the corrected control command is calculated based on the response trajectory. The duration is the time required for the state variable to change from its initial value to a stable value. For the charging power control of the energy storage system, after the correction command is injected, it takes 35 seconds for the charging power to be adjusted from 50kW to 75kW to reach stability, so the duration of action is 35 seconds.

[0051] Within the disturbance evolution scenario, quantized interference signals are generated by dividing the fluctuation period based on the duration of the control command. The fluctuation period is set to 1.5 to 2.5 times the duration of the control command. The quantized interference signals are generated using a stochastic process model, including random sequences such as Gaussian white noise and autoregressive moving average. The amplitude distribution of the quantized interference signals is calculated, represented by the probability density function of each amplitude value. Quantized interference signals with amplitude distributions within a preset range are injected into adjacent control nodes to obtain the interference input for the control nodes. For photovoltaic power plant irradiance intensity interference, the generated average value is 800 W / m². 2 The standard deviation is 150 W / m 2 The Gaussian-distributed interference signal has an amplitude mainly distributed around 500 W / m. 2 Up to 1100W / m 2 Between these values, the distribution falls within the preset range of 450W / m. 2 Up to 1150W / m 2 It is injected into the site environmental monitoring system every 60 seconds.

[0052] State fluctuation data is collected based on the interference input of the control node. This data records the changes in the control node's state variables under the influence of the interference signal. The amplitude change rate and frequency offset of the state fluctuation data are calculated. The amplitude change rate represents the speed of change of the fluctuation amplitude, and the frequency offset represents the deviation of the fluctuation frequency from the normal value. For wind farm grid-connected frequency control, after the injection of a frequency interference signal, the amplitude of the station's output frequency fluctuation increases from 0.1Hz to 0.25Hz within 30 seconds, with an amplitude change rate of 0.005Hz / s; the dominant frequency shifts from the normal 50Hz to 50.18Hz, with a frequency offset of 0.18Hz. An instability trajectory is constructed, representing the process of the state variables deviating from the stable operating point under the influence of the interference. Dynamic characteristics are extracted from the state fluctuation data using a phase space reconstruction method. For wind farm active power control, the instability trajectory can be represented as a closed curve on the power-frequency plane; a larger curve area indicates lower system stability. The output response trajectory and instability trajectory form a complete description of the system's dynamic behavior, providing data support for subsequent stability assessment.

[0053] In this embodiment, by constructing a phased evolution scenario and generating and injecting modified control commands and quantified interference signals, a comprehensive capture and accurate simulation of the dynamic response characteristics of new energy power plants under different operating conditions is achieved. The time-segmentation based on fluctuation inflection points accurately identifies key periods of system regulation and disturbance resistance capabilities, improving the accuracy and efficiency of simulation assessments. The construction of response trajectories and instability trajectories provides a direct reflection of the system's steady-state and dynamic characteristics, offering reliable data support for a comprehensive early warning mechanism. This solves the problem that traditional assessment methods struggle to address the random fluctuation characteristics of new energy power plants, improves the accuracy and timeliness of early warnings, enhances the system's safe operation guarantee capabilities, and plays a crucial technical supporting role in promoting high-proportion new energy access and safe and stable operation.

[0054] In one optional implementation, generating a quantized interference signal by dividing the fluctuation period based on the duration of action within the perturbation evolution scenario includes: The duration of action is divided into multiple fluctuation periods according to a preset ratio coefficient, and the average period of each fluctuation period is calculated. A cosine waveform reference signal is generated within the mean period, and the peak and trough positions of the reference signal are used as feature points to construct a time series of feature points. The time series of feature points are subjected to nonlinear transformation to generate multiple sets of distorted feature point sequences. The reference signal is reconstructed in segments based on the distorted feature point sequences to obtain multiple sets of distorted interference signals. Calculate the spectral distribution and energy distribution of the distorted interference signal, and construct the wave characteristic vector of the distorted interference signal; The matching degree between the fluctuation feature vector and the preset disturbance template is calculated, and the distortion interference signal with the highest matching degree is selected as the quantized interference signal.

[0055] The duration segmentation process divides the overall duration of the disturbance into equal segments based on the temporal characteristics of the disturbance evolution scenario, using a preset scaling factor. This preset scaling factor is typically set between 0.15 and 0.25 to ensure that the segmented fluctuation periods reflect both the continuity of the disturbance and facilitate subsequent signal processing. For a wind speed disturbance scenario in a wind farm, assuming an overall duration of 120 minutes, segmentation using a scaling factor of 0.2 yields six fluctuation periods, each lasting 20 minutes. The mean period of each fluctuation period is obtained by calculating the arithmetic mean of the durations of all periods, serving as the time reference for subsequent baseline signal generation. When the durations of the six fluctuation periods are 18 minutes, 21 minutes, 19 minutes, 22 minutes, 20 minutes, and 20 minutes, the mean period is 20 minutes.

[0056] The cosine waveform reference signal is generated within a defined mean period, using a standard cosine function as the basic signal form. The mathematical expression for the cosine waveform is the amplitude multiplied by the cosine function, where the angular frequency is calculated based on the mean period. The angular frequency equals 2π divided by the mean period, ensuring the reference signal completes a full waveform cycle within one mean period. For a 20-minute mean period, the angular frequency is π / 600 radians per second, and the amplitude of the reference signal is set to 1 unit. Peaks correspond to moments when the cosine function value is 1, and troughs correspond to moments when the cosine function value is -1. Within the 20-minute period, peaks occur at the 5th and 15th minutes, and troughs occur at the 10th and 20th minutes; these positions constitute a time series of characteristic points.

[0057] Nonlinear transformations apply nonlinear mapping functions to the time series of feature points, generating distorted feature point sequences with varying degrees of deformation. These nonlinear mapping functions include power function transformations, exponential function transformations, and trigonometric function transformations. Power function transformations use powers between 1.5 and 2.5 to transform the original time coordinates; exponential function transformations use natural exponential functions to nonlinearly stretch the time coordinates; and trigonometric function transformations use sine or cosine functions to periodically modulate the time coordinates. When the original feature point time series is transformed with a power of 2, the peak position at the 5th minute is transformed to the 2.5th minute, and the trough position at the 10th minute is transformed to the 5th minute, generating a distorted feature point sequence with time compression characteristics.

[0058] The segmented reconstruction process locally modifies the original reference signal based on the sequence of distorted feature points, generating a new signal waveform through interpolation and fitting methods. The reconstruction process divides the mean period into multiple time segments, each defined by adjacent distorted feature points. Within each time segment, cubic spline interpolation connects adjacent distorted feature points, ensuring the reconstructed signal has continuous first and second derivatives. By applying different nonlinear transformation functions, multiple sets of interference signals with different distortion characteristics can be generated. When modeling photovoltaic power generation disturbances, the distorted interference signal generated using exponential transformation exhibits time compression during the power rise phase and time extension during the power fall phase.

[0059] The spectral distribution calculation employs Fast Fourier Transform (FFT) to perform frequency domain analysis on the distorted interference signal, obtaining the amplitude distribution of the signal at different frequency components. FFT converts the time-domain signal into a frequency-domain representation, revealing the signal's frequency components and phase characteristics. The spectral distribution calculates the power spectral density by squared the amplitude of each frequency component, reflecting the distribution of signal energy along the frequency axis. Energy distribution calculation is based on Passevar's theorem, equating the total energy of the time-domain signal to the sum of the energies of each frequency component in the frequency domain. Spectral analysis of the energy storage battery charging and discharging disturbance signal shows that the main energy is concentrated in the low-frequency band from 0.1 Hz to 0.5 Hz, corresponding to the time constant characteristics of the battery response.

[0060] The construction of the fluctuation eigenvector combines key parameters of spectral and energy distribution to form a multidimensional vector describing signal characteristics. The eigenvector includes statistical parameters such as dominant frequency, spectral width, energy concentration, skewness, and kurtosis. The dominant frequency is the frequency component with the largest amplitude in the spectrum; the spectral width reflects the dispersion of signal frequency components; the energy concentration represents the proportion of the dominant frequency component to the total energy; and skewness and kurtosis describe the asymmetry and sharpness of the spectral distribution. In the eigenvector of transformer load fluctuation, the dominant frequency is 0.3 Hz, the spectral width is 0.8 Hz, and the energy concentration is 0.75, constituting a 5-dimensional eigenvector.

[0061] The matching degree calculation employs similarity metrics such as Euclidean distance or cosine similarity to evaluate the degree of matching between the fluctuation feature vector of the distorted interference signal and the preset disturbance template. The preset disturbance template is pre-established based on the typical disturbance characteristics of new energy equipment, encompassing standard disturbance patterns under various operating conditions. Euclidean distance measures similarity by calculating the square root of the sum of the squares of the differences between the components of the feature vector; a smaller distance indicates a higher matching degree. Cosine similarity assesses directional similarity by calculating the cosine of the angle between two vectors; a value closer to 1 indicates a higher matching degree. For wind turbine blade angle disturbance simulation, when the Euclidean distance between the feature vector of a certain set of distorted interference signals and the preset template is 0.15, that signal is selected as the optimal quantized interference signal.

[0062] In this embodiment, nonlinear transformation and spectrum matching techniques are used to generate quantized interference signals that closely resemble real disturbance characteristics, effectively simulating equipment response behavior under various complex operating conditions. The construction of fluctuation feature vectors and template matching mechanisms ensure that the generated interference signals have good representativeness and applicability, improving the realism and effectiveness of operational simulation assessments. This interference signal generation method based on signal processing and pattern recognition has strong versatility and scalability, and can adapt to the disturbance characteristic modeling needs of different types of new energy equipment, providing a reliable technical foundation for site operation status assessment and fault prediction.

[0063] In one optional implementation, the dynamic deviation between the response trajectory and the preset safety boundary is calculated to obtain the regulation capability score, and the dynamic deviation between the instability trajectory and the preset safety boundary is calculated to obtain the disturbance rejection capability score. The combination of the regulation capability score and the disturbance rejection capability score to generate the assessment result includes: Extract the upper and lower threshold values ​​and standard frequency values ​​of the preset safety boundary to generate safety boundary parameters; The response trajectory is segmented according to the sampling time sequence, and the fluctuation deviation between each segment of the response trajectory and the safety boundary parameter is calculated to construct the fluctuation amplitude deviation sequence and the fluctuation frequency deviation sequence. The cumulative duration of the amplitude exceeding the safety boundary parameter in the fluctuation amplitude deviation sequence is calculated to obtain the amplitude exceeding the limit duration, and the cumulative magnitude of the deviation from the standard frequency value in the fluctuation frequency deviation sequence is calculated to obtain the frequency offset. The duration of amplitude exceeding the limit and the frequency offset are converted according to a preset mapping coefficient to generate a regulation capability score; The instability trajectory is segmented according to the fluctuation period, and the fluctuation change rate of each segment of the instability trajectory relative to the safety boundary parameter is calculated to construct the fluctuation rate sequence and the fluctuation acceleration sequence. Calculate the number of abrupt changes in the fluctuation rate sequence and the transition amplitude in the fluctuation acceleration sequence, and convert the number of abrupt changes and the transition amplitude according to a preset mapping coefficient to generate an anti-disturbance capability score; The assessment results are generated by combining the adjustment capability score and the disturbance resistance score according to a preset weighting.

[0064] In one implementation method, preset safety boundary parameters are first obtained, including an upper threshold, a lower threshold, and a standard frequency value. These parameters can be preset according to system stability requirements. For example, in power system frequency regulation, the upper threshold can be set to 50.2Hz, the lower threshold to 49.8Hz, and the standard frequency value to 50Hz.

[0065] After acquiring the response trajectory data, it is segmented according to the sampling time sequence. The 1000 consecutive sampling points are divided into 10 equal-length time periods, with each sampling point lasting 10 milliseconds. For each response trajectory segment, its fluctuation deviation from the safety boundary parameters is calculated. The fluctuation deviation includes both the fluctuation amplitude deviation and the fluctuation frequency deviation.

[0066] The fluctuation amplitude deviation is calculated as the difference between the current response value and the upper or lower limit of the safety boundary. For example, when the response value is 50.3Hz and the upper limit threshold is 50.2Hz, the fluctuation amplitude deviation is 0.1Hz. If the response value is between the upper and lower limits, the fluctuation amplitude deviation is zero. By calculating for all segments, a complete fluctuation amplitude deviation sequence is constructed.

[0067] The fluctuation frequency deviation is calculated as the difference between the fluctuation period in the response trajectory and the corresponding period of the standard frequency value. In practice, the spectral characteristics of the response trajectory can be extracted using a Fast Fourier Transform (FFT) to calculate the difference between the dominant frequency and the standard frequency. For example, if the dominant frequency of the response trajectory is 0.5 Hz and the standard frequency is 0.3 Hz, then the fluctuation frequency deviation is 0.2 Hz. This calculation is repeated for all segments to form a fluctuation frequency deviation sequence.

[0068] Based on the constructed deviation sequence, the duration of amplitude exceeding the limit is calculated. Specifically, this is done by summing the times corresponding to the elements greater than zero in the fluctuation amplitude deviation sequence. For example, in a 10-second response trajectory, if the amplitude exceeds the safety boundary for 2 seconds, then the duration of amplitude exceeding the limit is 2 seconds.

[0069] Simultaneously, the frequency offset is calculated by summing the absolute values ​​of all elements in the fluctuation frequency deviation sequence. If the frequency deviations of a certain response trajectory in 10 segments are [0.1, 0.2, 0.1, 0.3, 0.2, 0.1, 0.2, 0.3, 0.1, 0.2] Hz, then the frequency offset is 1.8 Hz.

[0070] The calculated amplitude over-limit duration and frequency offset are converted into a regulation capability score using a preset mapping coefficient. The specific mapping relationship can be expressed as: Regulation capability score = 100 - (amplitude over-limit duration × mapping coefficient 1 + frequency offset × mapping coefficient 2).

[0071] When the amplitude exceeds the limit for 2 seconds, the frequency offset is 1.8Hz, the mapping coefficient 1 is 10, and the mapping coefficient 2 is 15, the adjustment ability score is 100-(2×10+1.8×15)=100-(20+27)=53 points.

[0072] For handling unstable trajectories, the trajectory is first segmented according to the fluctuation period. The period boundaries are determined by detecting peak points in the trajectory. For example, if five peaks are detected consecutively, the trajectory can be divided into four complete period segments.

[0073] For each unstable trajectory segment, calculate the rate of change of fluctuation, including fluctuation velocity and fluctuation acceleration. The fluctuation velocity is the difference between adjacent sampling points divided by the sampling time interval, and the fluctuation acceleration is the difference between adjacent velocities divided by the sampling time interval. Process all segmented data to construct complete fluctuation velocity and fluctuation acceleration sequences.

[0074] Identify abrupt change points in the fluctuation rate sequence, i.e., points where the rate value changes beyond a preset threshold. For example, if the preset threshold is 0.5 Hz / s, a point where an adjacent rate change exceeds this value is marked as an abrupt change point. Count the number of all abrupt change points.

[0075] Simultaneously, the transition amplitudes in the wave acceleration sequence are calculated, which are the differences between the maximum and minimum values ​​in the acceleration sequence. For example, the maximum value in the acceleration sequence is 2 Hz / s. 2 The minimum value is -1.5Hz / s 2 The transition amplitude is then 3.5 Hz / s. 2 .

[0076] The number of mutation points and the transition amplitude are converted into an immunity score according to a preset mapping coefficient: Immunity score = 100 - (Number of mutation points × Mapping coefficient 3 + Transition amplitude × Mapping coefficient 4). When the number of mutation points is 8, the transition amplitude is 3.5 Hz / s. 2 When the mapping coefficient 3 is 2 and the mapping coefficient 4 is 5, the anti-interference ability score is 100-(8×2+3.5×5)=100-(16+17.5)=66.5 points.

[0077] The final assessment result is obtained by combining the adjustment capability score and the disturbance rejection capability score: Assessment result = Adjustment capability score × Combined weight 1 + Disturbance rejection capability score × Combined weight 2. When the adjustment capability score is 53 points, the disturbance rejection capability score is 66.5 points, the combined weight 1 is 0.6, and the combined weight 2 is 0.4, the assessment result = 53 × 0.6 + 66.5 × 0.4 = 31.8 + 26.6 = 58.4 points.

[0078] This scoring mechanism can comprehensively evaluate the system's responsiveness and stability under different operating conditions, providing a quantitative basis for system optimization. By adjusting various mapping coefficients and combined weights, the focus of the evaluation system can be flexibly configured according to actual application needs, improving the practicality of the evaluation results.

[0079] A second aspect of this invention provides a simulation and assessment system for the operation of new energy power stations based on comprehensive early warning, the system comprising: The first unit is used to collect real-time operation data and historical fault data of new energy power stations; The second unit is used to extract the fluctuation characteristics of each operating parameter before the occurrence of the fault from historical fault data, calculate the correlation coefficient between each fluctuation characteristic, and construct the parameter propagation network based on the correlation coefficient. The third unit is used to input real-time operating data into the parameter propagation network, identify initial fluctuation anomalies, extract propagation sequences related to initial fluctuation anomalies from the parameter propagation network, and generate risk propagation paths. The fourth unit is used to extract the rate of change of fluctuations of the parameters running on the risk propagation path from historical fault data, determine the inflection point of fluctuations, and divide the adjustment period and the disturbance period based on the inflection point of fluctuations. The fifth unit is used to construct a phased evolution scenario based on the risk propagation path, inject corrective control commands during the adjustment period, inject quantitative interference signals during the disturbance period, and record the response trajectory of the corrective control commands and the instability trajectory of the quantitative interference signals. The sixth unit is used to calculate the dynamic deviation between the response trajectory and the preset safety boundary to obtain the regulation capability score, and to calculate the dynamic deviation between the instability trajectory and the preset safety boundary to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result.

[0080] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0081] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0082] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0083] 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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A simulation and assessment method for the operation of new energy power stations based on comprehensive early warning, characterized in that, include: Collect real-time operation data and historical fault data of new energy power stations; Extract the fluctuation characteristics of each operating parameter before the failure from historical failure data, calculate the correlation coefficient between each fluctuation characteristic, and construct a parameter propagation network based on the correlation coefficient; Real-time operational data is input into the parameter propagation network to identify initial fluctuation anomalies, and propagation sequences related to the initial fluctuation anomalies are extracted from the parameter propagation network to generate risk propagation paths. Extract the rate of change of fluctuation of the parameters running on the risk propagation path from historical failure data, determine the fluctuation inflection point, and divide the adjustment period and disturbance period based on the fluctuation inflection point; Based on the risk propagation path, a phased evolution scenario is constructed. Corrective control commands are injected during the adjustment period, and quantitative interference signals are injected during the disturbance period. The response trajectory of the corrective control commands and the instability trajectory of the quantitative interference signals are recorded. The dynamic deviation between the response trajectory and the preset safety boundary is calculated to obtain the regulation capability score, and the dynamic deviation between the instability trajectory and the preset safety boundary is calculated to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result.

2. The method according to claim 1, characterized in that, Extracting the fluctuation characteristics of various operating parameters before the fault occurred from historical fault data, calculating the correlation coefficients between the fluctuation characteristics, and constructing a parameter propagation network based on the correlation coefficients includes: Locate the fault trigger time from historical fault data, extract a historical data segment of a preset duration from the fault trigger time as the end time, extract the fluctuation amplitude and fluctuation period of each operating parameter within the historical data segment, and combine them to form the fluctuation characteristics of each operating parameter. Calculate the time delay correlation coefficient between the fluctuation characteristics of any two operating parameters, and screen the operating parameter pairs whose time delay correlation coefficient reaches a preset strength; Determine which operating parameter fluctuates first and which fluctuates later in the operating parameter pair. Set the operating parameter that fluctuates first as the propagation source node and the operating parameter that fluctuates later as the propagation target node. Use the time delay correlation coefficient between the propagation source node and the propagation target node as the inter-node correlation coefficient. A parameter propagation network is constructed based on the source node, the target node, and the correlation coefficient between the nodes. The nodes of the parameter propagation network are the operating parameters, the edges are the connections from the source node to the target node, and the weight of the edge is the correlation coefficient between the nodes.

3. The method according to claim 1, characterized in that, Real-time operational data is input into the parameter propagation network to identify initial fluctuation anomalies. Propagation sequences related to the initial fluctuation anomalies are extracted from the parameter propagation network to generate risk propagation paths, including: The real-time running data is input into the parameter propagation network, the real-time fluctuation characteristics of each running parameter in the real-time running data are extracted, the deviation between the real-time fluctuation characteristics and the fluctuation characteristics of the corresponding nodes in the parameter propagation network is calculated, the nodes whose deviation exceeds the preset deviation threshold are marked as the initial fluctuation anomalies, and the fluctuation direction of the initial fluctuation anomalies is recorded. Based on the node position of the initial fluctuation anomaly in the parameter propagation network, search for downstream nodes along the directed edges, extract the real-time fluctuation characteristics of the corresponding operating parameters of the downstream nodes, calculate the fluctuation direction corresponding to the real-time fluctuation characteristics of the downstream nodes, and select downstream nodes whose fluctuation direction is consistent with the fluctuation direction of the initial fluctuation anomaly to form a candidate propagation node set. Extract the correlation coefficients between nodes in the parameter propagation network for each node in the candidate propagation node set, calculate the path correlation strength from the initial fluctuation anomaly to each node in the candidate propagation node set, and filter the nodes and corresponding propagation paths with path correlation strength greater than a preset strength threshold to form a propagation sequence related to the initial fluctuation anomaly. Risk propagation paths are generated based on the propagation hierarchy and path relevance of nodes in the propagation sequence.

4. The method according to claim 1, characterized in that, Extract the rate of change of fluctuations in parameters along the risk propagation path from historical failure data, determine the inflection points of fluctuations, and divide the adjustment period and disturbance period based on the inflection points: Construct sliding observation windows of different durations, extract fluctuation data of operating parameters on the risk propagation path, calculate the mean and variance of the fluctuation data, select the optimal observation window according to the minimum variance criterion, and use the optimal observation window to resample the operating parameters on the risk propagation path to obtain the fluctuation sampling sequence. The fluctuation sampling sequence is decomposed into multiple fluctuation components, the energy distribution value of the multiple fluctuation components is calculated, the target fluctuation component is selected according to the energy distribution value, and the target fluctuation component is recombined to generate a fluctuation change sequence. Calculate the change between adjacent sampling points in the fluctuation change sequence to obtain the fluctuation change rate of the running parameters on the risk propagation path. Segment the fluctuation change rate, identify the change direction of the fluctuation change rate, and mark the sampling points where the change direction reverses as fluctuation inflection candidate points. The amplitude ratio and duration of fluctuation inflection candidate points are calculated to form fluctuation morphology characteristics, and fluctuation inflection points are obtained by screening based on the fluctuation morphology characteristics. The fluctuation sequence is segmented based on the turning point of the fluctuation. The time period when the fluctuation rate changes from negative to positive is divided into the adjustment period, and the time period when the fluctuation rate changes from positive to negative is divided into the disturbance period.

5. The method according to claim 1, characterized in that, A phased evolution scenario is constructed based on the risk propagation path. Corrective control commands are injected during the adjustment period, and quantized disturbance signals are injected during the disturbance period. The response trajectories of the corrective control commands and the instability trajectories of the quantized disturbance signals are recorded, including: Collect state variables of adjacent control nodes along the risk propagation path, calculate the mean and fluctuation amplitude of the state variables, determine the evolution boundary, and divide the evolution boundary into regulation evolution scenario and disturbance evolution scenario based on the fluctuation inflection point; In the context of the regulation and evolution scenario, a correction control command is generated, the fluctuation frequency and fluctuation amplitude of the correction control command are calculated, and the correction control command with the minimum fluctuation frequency and the fluctuation amplitude that meets the preset fluctuation range is injected into the adjacent control node at time intervals to obtain the correction input signal of the control node. Collect the state response data corresponding to the correction input signal of the control node, calculate the rate of change and acceleration of the state response data, construct the response trajectory, and calculate the duration of the correction control command based on the response trajectory; Within the disturbance evolution scenario, the fluctuation period is divided based on the duration of action to generate a quantized interference signal. The amplitude distribution of the quantized interference signal is calculated, and the quantized interference signal whose amplitude distribution meets the preset distribution range is injected into the adjacent control node to obtain the interference input of the control node. Based on the disturbance input of the control node, the state fluctuation data is collected, the amplitude change rate and frequency offset of the state fluctuation data are calculated, the instability trajectory is constructed, and the response trajectory and instability trajectory are output.

6. The method according to claim 5, characterized in that, In a disturbance evolution scenario, generating quantized interference signals by dividing the fluctuation period based on the duration of action includes: The duration of action is divided into multiple fluctuation periods according to a preset ratio coefficient, and the average period of each fluctuation period is calculated. A cosine waveform reference signal is generated within the mean period, and the peak and trough positions of the reference signal are used as feature points to construct a time series of feature points. The time series of feature points are subjected to nonlinear transformation to generate multiple sets of distorted feature point sequences. The reference signal is reconstructed in segments based on the distorted feature point sequences to obtain multiple sets of distorted interference signals. Calculate the spectral distribution and energy distribution of the distorted interference signal, and construct the wave characteristic vector of the distorted interference signal; The matching degree between the fluctuation feature vector and the preset disturbance template is calculated, and the distortion interference signal with the highest matching degree is selected as the quantized interference signal.

7. The method according to claim 1, characterized in that, The dynamic deviation between the response trajectory and the preset safety boundary is calculated to obtain the regulation capability score, and the dynamic deviation between the instability trajectory and the preset safety boundary is calculated to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result, which includes: Extract the upper and lower threshold values ​​and standard frequency values ​​of the preset safety boundary to generate safety boundary parameters; The response trajectory is segmented according to the sampling time sequence, and the fluctuation deviation between each segment of the response trajectory and the safety boundary parameter is calculated to construct the fluctuation amplitude deviation sequence and the fluctuation frequency deviation sequence. The cumulative duration of the amplitude exceeding the safety boundary parameter in the fluctuation amplitude deviation sequence is calculated to obtain the amplitude exceeding the limit duration, and the cumulative magnitude of the deviation from the standard frequency value in the fluctuation frequency deviation sequence is calculated to obtain the frequency offset. The duration of amplitude exceeding the limit and the frequency offset are converted according to a preset mapping coefficient to generate a regulation capability score; The instability trajectory is segmented according to the fluctuation period, and the fluctuation change rate of each segment of the instability trajectory relative to the safety boundary parameter is calculated to construct the fluctuation rate sequence and the fluctuation acceleration sequence. Calculate the number of abrupt changes in the fluctuation rate sequence and the transition amplitude in the fluctuation acceleration sequence, and convert the number of abrupt changes and the transition amplitude according to a preset mapping coefficient to generate an anti-disturbance capability score; The assessment results are generated by combining the adjustment capability score and the disturbance resistance score according to a preset weighting.

8. A new energy power station operation simulation and assessment system based on comprehensive early warning, used to implement the method of any one of claims 1-7, characterized in that, include: The first unit is used to collect real-time operation data and historical fault data of new energy power stations; The second unit is used to extract the fluctuation characteristics of each operating parameter before the occurrence of the fault from historical fault data, calculate the correlation coefficient between each fluctuation characteristic, and construct the parameter propagation network based on the correlation coefficient. The third unit is used to input real-time operating data into the parameter propagation network, identify initial fluctuation anomalies, extract propagation sequences related to initial fluctuation anomalies from the parameter propagation network, and generate risk propagation paths. The fourth unit is used to extract the rate of change of fluctuations of the parameters running on the risk propagation path from historical fault data, determine the inflection point of fluctuations, and divide the adjustment period and the disturbance period based on the inflection point of fluctuations. The fifth unit is used to construct a phased evolution scenario based on the risk propagation path, inject corrective control commands during the adjustment period, inject quantitative interference signals during the disturbance period, and record the response trajectory of the corrective control commands and the instability trajectory of the quantitative interference signals. The sixth unit is used to calculate the dynamic deviation between the response trajectory and the preset safety boundary to obtain the regulation capability score, and to calculate the dynamic deviation between the instability trajectory and the preset safety boundary to obtain the disturbance rejection capability score. The regulation capability score and the disturbance rejection capability score are combined to generate the assessment result.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images