Dynamic evaluation method and system for real-time adjustment capability of power generation equipment of new energy power station

Abstract

The invention relates to the technical field of power system dispatching, in particular to a dynamic evaluation method and system for the real-time adjusting capacity of power generation equipment of a new energy power station, and the method comprises the steps: collecting multi-dimensional data, and obtaining a standardized data set through lightweight preprocessing; in combination with the rated parameters and the real-time operation state of the equipment, calculating core adjustment indexes of the equipment, generating differential labels of the equipment, and screening out invalid data to generate adjustment capability evaluation parameters of the equipment; inputting a layered dynamic evaluation model to obtain a risk prediction result; performing abnormity judgment by adopting a dynamic threshold algorithm, and generating equipment label adjustment suggestions and system optimization data according to a judgment result; and inputting the equipment regulation capability evaluation parameters, the system optimization data and the scene type label into an evaluation result application model, and calculating the equipment scheduling priority in combination with a power grid scheduling demand and a power station operation and maintenance target to generate a differentiated scheduling strategy. According to the scheme, the stability of the power system is effectively improved through double-domain coupling evaluation and a differentiation strategy.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and in particular to a method and system for dynamic evaluation of the real-time regulation capability of power generation equipment in new energy power plants. Background Technology

[0002] As the global energy structure transitions towards cleaner and lower-carbon energy, the installed capacity and penetration rate of new energy power generation systems such as photovoltaics and energy storage continue to increase in the power system, gradually becoming a core component of power supply. However, the intermittent and fluctuating characteristics of new energy power generation, as well as the diverse types of equipment and complex operating conditions of new energy power plants, pose significant challenges to existing technologies in assessing the regulation capabilities of new energy power plant equipment, grid coordinated dispatch, and risk management, as detailed below: During the operation of new energy power plants, the operational requirements of core components such as photovoltaic inverters, energy storage devices, power grids, and the environment exhibit significant time-domain and inter-domain differences. Photovoltaic inverters are prone to power fluctuations under sudden increases or decreases in irradiance, requiring real-time adjustment of MPPT tracking accuracy to maintain stable output (typically requiring ≥98%). Energy storage devices need to rapidly release power to fill the gap during sudden grid load shortages, while under normal operating conditions, they need to maintain a stable SOC within the optimal range of 20%-80% in the medium term to prevent battery aging. The power grid needs to dynamically allocate equipment regulation capabilities based on real-time voltage and frequency deviations to avoid parameter overruns (typically requiring voltage deviation ≤2% of rated voltage and frequency deviation ≤0.2Hz). However, traditional new energy power plant management solutions often employ static parameter evaluation and fixed threshold control, relying on data feedback from single devices. This makes it difficult to simultaneously address the dual needs of "short-term emergency regulation" and "medium-term stable operation," and fails to fully consider dynamic factors such as environmental fluctuations, equipment aging, and grid topology changes. This results in underutilized equipment regulation potential, low thermal resource allocation efficiency, significant energy waste, and even problems such as grid parameter overruns or equipment overload damage.

[0003] Meanwhile, multi-source data from new energy power plants (such as 100ms-level active / reactive power of inverters, 50ms-level charging and discharging power of energy storage, 500ms-level voltage and frequency data of the power grid, and 1-minute-level irradiance and temperature data of the environment) suffer from problems such as large differences in acquisition frequency, inconsistent transmission protocols (such as Modbus-TCP, IEC61850, MQTT), and instantaneous data interference. Traditional data processing methods can only filter outliers and cannot effectively integrate the spatiotemporal dimensions of the "equipment regulation domain" and the "power grid demand domain," making it difficult to support accurate regulation capability assessment. In addition, existing scheduling and operation and maintenance strategies lack differentiated adaptation capabilities and cannot adjust control logic according to equipment characteristics (such as high-response and large-capacity types) and core power grid needs (such as frequency regulation and voltage regulation), resulting in insufficient adaptability of the system under different operating scenarios and limited new energy consumption rate. Summary of the Invention

[0004] This invention uses multi-source data processing, dual-domain coupled evaluation, and differentiated strategies to accurately quantify the regulation capabilities of new energy power plant equipment, thereby achieving efficient grid-coordinated dispatching and risk prevention and control, and improving the new energy absorption rate and power system stability.

[0005] The technical solution proposed in this invention is: a method for dynamic evaluation of the real-time regulation capability of power generation equipment in a new energy power plant, the method comprising: Collect multi-dimensional data and obtain a standardized dataset through lightweight preprocessing; Based on a standardized dataset, combined with the equipment's rated parameters and real-time operating status, the core adjustment indicators of the equipment are calculated, and the equipment's differentiated labels are generated. Using a dual-domain coupled weighting algorithm and cluster aggregation logic, the equipment's adjustment contribution and cluster aggregation capability are determined. After filtering out invalid data, the equipment adjustment capability evaluation parameters are generated. The equipment regulation capacity assessment parameters are input into the hierarchical dynamic assessment model, and combined with real-time power grid demand and environmental prediction data, risk prediction results are obtained. Based on the risk prediction results, a dynamic threshold algorithm is used to determine anomalies, and equipment tag adjustment suggestions and system optimization data are generated based on the determination results. The evaluation results are input into the model by combining equipment regulation capability assessment parameters, system optimization data, and scenario type labels. The model is then used to calculate equipment scheduling priorities and generate differentiated scheduling strategies by combining power grid dispatching requirements with power plant operation and maintenance objectives.

[0006] Preferably, the specific process for obtaining the standardized dataset is as follows: Acquire photovoltaic inverter operation data, energy storage device data, grid operating condition data, environmental data, and equipment fault data; divide parameters into equipment regulation domain and grid demand domain according to a preset rule base, assign corresponding labels to parameters in both domains, bind the labels via SQL, and store them in a time-series database; use 3D... The criteria include constructing a sliding window to filter outliers, replacing out-of-range data with linear interpolation, unifying the units and timestamps of all parameters, and aligning data from different acquisition frequencies through linear interpolation.

[0007] Preferably, the specific process for obtaining the device differentiation label is as follows: By combining standardized datasets, equipment rated parameters, and real-time operating status, the system calculates the equipment's real-time active power regulation range, active power regulation rate, reactive power regulation capacity, voltage regulation accuracy, command response delay, and regulation duration. In extreme scenarios, a scenario correction coefficient and grid disturbance power are introduced to correct the regulation duration. Based on preset index thresholds, high-response equipment must simultaneously meet the active power regulation rate and command response delay requirements; large-capacity equipment must meet the requirement of real-time active power regulation range covering the rated power ratio or reactive power regulation capacity covering the rated apparent power ratio; and long-duration equipment must meet the requirement that the regulation duration is not less than a specified multiple of the grid demand regulation duration. Based on the judgment results, differentiated labels are generated for the equipment, and the equipment code and label type are clearly marked in the labels.

[0008] Preferably, the specific process for obtaining the equipment adjustment capability evaluation parameters is as follows: A dual-domain coupled weighted algorithm is adopted to generate a sensitivity matrix by combining real-time power grid topology data and operating parameters, and to calculate the adjustment contribution of a single device to power grid demand. Based on the core needs of the power grid and the differentiated labels of the devices, cluster aggregation is carried out to calculate the overall adjustment rate of the cluster and the total reactive power adjustment capacity, and invalid device data with excessive fault ratios are filtered out. The adjustment contribution of a single device, the cluster aggregation capability, the boundary prediction value of extreme scenarios, the comprehensive evaluation score of the device and the cluster, and the demand matching degree are integrated to form a device adjustment capability evaluation parameter covering the individual adjustment potential of the device, the cluster coordination capability, and the adaptability to power grid demand.

[0009] Preferably, the specific process for obtaining the risk prediction result is as follows: Using equipment regulation capacity assessment parameters as the core input, combined with real-time grid demand data and 1-hour environmental forecast data, a hierarchical dynamic assessment model is imported. Through dual-domain coupled assessment, the regulation contribution of a single device to grid demand is clarified. Based on the differentiated labels of the equipment, cluster aggregation assessment is carried out to determine the comprehensive regulation capacity of different types of equipment clusters. Boundary prediction assessment is used to correct the indicator boundaries under extreme scenarios. The results of each stage of dual-domain coupled assessment, cluster aggregation assessment and boundary prediction assessment are integrated to output risk prediction results including extreme scenario boundary prediction values, early warning signals, comprehensive assessment scores of equipment and clusters, and the matching degree between equipment and grid demand.

[0010] Preferably, the specific process for obtaining the device label adjustment suggestion and the system optimization data is as follows: Based on the strategy execution results and early warning verification data, the strategy execution effect verification is initiated. The effectiveness of the power grid dispatch strategy is determined by calculating the equipment command response deviation rate and the proportion of compliant equipment. The effectiveness of the operation and maintenance measures is verified by comparing the improvement rate of core indicators before and after operation and maintenance. The rationality of the early warning rules is verified by statistically analyzing the actual failure rate after the early warning. Based on the verification results, the dual-domain coupling weight is fine-tuned using the small-batch gradient descent method with the goal of minimizing the response deviation rate. The extreme scenario correction coefficient and the influence factors in the boundary prediction formula are optimized by combining the failure rate to generate updated model parameters. The aging coefficient is calculated by combining the battery cycle count, charge and discharge depth, ambient temperature and rated cycle life in the equipment's full life cycle data. The regional aggregation weight is adjusted according to the load ratio and line impedance after the power grid topology change, and the equipment label judgment threshold is dynamically adjusted. The above data are integrated to form system optimization data and equipment label adjustment suggestions respectively.

[0011] Preferably, the specific process for obtaining the differentiated scheduling strategy is as follows: Using equipment regulation capacity assessment parameters, system optimization data, and scenario type labels as inputs, and combining real-time grid demand with power plant operation and maintenance objectives, a grid dispatch strategy is initiated. A dispatch priority calculation model is constructed based on equipment regulation contribution and comprehensive evaluation score. The equipment regulation contribution is taken from the dual-domain coupled evaluation results, and the comprehensive evaluation score is obtained by weighted summation. The matching degree between a single device and grid demand is split from the cluster matching degree according to the device's rated power ratio, and the dispatch priority of each device is calculated. According to the priority ranking results, differentiated instructions are issued to different types of devices in combination with the cluster aggregation scheme. Short-term high-frequency regulation instructions are assigned to high-priority, high-response devices, and continuous voltage regulation instructions are assigned to large-capacity devices. The instructions are synchronized to the device controller through the corresponding communication protocol to form a differentiated dispatch strategy adapted to grid demand and device characteristics.

[0012] The present invention also provides a dynamic evaluation system for the real-time regulation capability of power generation equipment in a new energy power plant, the system being used to execute the aforementioned dynamic evaluation method for the real-time regulation capability of power generation equipment in a new energy power plant.

[0013] The present invention also provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the aforementioned method for dynamic evaluation of the real-time regulation capability of power generation equipment in a new energy power plant.

[0014] The beneficial effects of this invention are: 1. By dividing multi-source data into "equipment regulation domain" and "grid demand domain" and combining dynamic sensitivity matrix to calculate dual-domain coupling weights, on the one hand, it can accurately quantify the regulation contribution of a single device to grid demand. Compared with traditional static threshold assessment, it can reduce the misjudgment rate of equipment regulation capacity by more than 30%, effectively avoiding overload damage caused by overestimation of equipment capacity or grid regulation command failure caused by underestimation, and ensuring that grid voltage and frequency are stable within the rated range. On the other hand, it can carry out cluster aggregation based on the differentiated characteristics of equipment to form functional frequency and voltage regulation clusters, which can improve the overall regulation rate of the cluster by 20%-40% and increase the total reactive power regulation capacity utilization rate to more than 90%, effectively filling the grid load gap and enhancing the coordinated stability capability of new energy power plants and the grid.

[0015] 2. Through the improved version "3 The lightweight preprocessing process, including outlier filtering, multi-frequency data alignment, and unit standardization, can improve the integrity of multi-source data (100ms-level power, 500ms-level voltage, etc.) to over 99.5%, eliminate coupling barriers caused by differences in data format and frequency, and reduce the processing time of dual-domain data by 50%, providing efficient support for subsequent evaluation. At the same time, by correcting the boundary of extreme scenario indicators based on environmental prediction data and equipment status, it can trigger risk warnings such as sudden changes in irradiance and equipment cluster failures 5-10 minutes in advance, reducing the probability of grid parameters exceeding limits in extreme scenarios by more than 60%, avoiding operational accidents caused by power fluctuations or cluster regulation failures, and significantly improving the reliability of power plant operation.

[0016] 3. By allocating differentiated instructions to equipment with different characteristics through a scheduling priority algorithm, the frequency regulation instruction response rate of high-response equipment can be increased to over 90%, and the voltage regulation task completion efficiency of large-capacity equipment can be improved by 35%. At the same time, the ineffective energy consumption of equipment is reduced, and the renewable energy consumption rate is increased by 5%-10%. By formulating classified operation and maintenance strategies according to equipment tags, the fault repair efficiency of high-response equipment can be improved by 40%, and the SOC decay rate of long-duration energy storage can be controlled below 0.1% / day. By combining execution feedback and incremental learning to fine-tune model parameters, the evaluation accuracy can be gradually optimized over the operating time, adapting to dynamic scenarios such as equipment aging and grid topology changes, and extending the applicability of the solution to the entire life cycle of the power plant. Attached Figure Description

[0017] Figure 1 A flowchart of a method for dynamic evaluation of the real-time regulation capability of power generation equipment in a new energy power plant; Figure 2 This is a flowchart illustrating the evaluation process of a dynamic evaluation method for the real-time regulation capability of power generation equipment in a new energy power plant. Detailed Implementation

[0018] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0019] It is understood that the term "a" should be understood as "at least one" or "one or more," that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0020] like Figure 1 and Figure 2 As shown, the photovoltaic inverter's operating data is collected in real time by the inverter's built-in monitoring module. Parameters such as active / reactive power, output voltage, and current at the 100ms level are transmitted to the edge gateway via the Modbus-TCP protocol. The MPPT tracking accuracy feedback signal (MPPT tracking accuracy threshold set to...) is also included. The output is directly from the inverter controller, while the cooling system's operating status is determined by a combination of temperature sensors and fan operation signals. In the energy storage device's operating data, the 50ms-level charge / discharge power curve is uploaded in real-time by the energy storage converter's PCS, with the PCS response delay upper limit set to [value missing]. The SOC value of the battery pack is calculated by collecting the voltage and current of each individual cell through the battery management system (BMS), and then using a second-order RC equivalent circuit model combined with the Kalman filter algorithm. ,in Remaining battery power This is the rated capacity, taking into account the temperature compensation coefficient. (Low temperature environment) , normal temperature ,high temperature ) and aging coefficient Dynamic correction , The actual cycle count of the battery is determined by fitting data from the full life cycle test of 100 groups of the same model of battery. The PCS command response delay is obtained by recording the time difference between command issuance and power change. The battery cell voltage balance status and charge / discharge cutoff voltage threshold are retrieved from the BMS parameter configuration. In the grid operating condition data, 500ms-level three-phase voltage / frequency waveform data at the grid connection point are collected by a high-precision power quality monitoring device. The real-time load gap of the regional power grid is issued by the SCADA system of the power grid dispatch center through a dedicated data link of the IEC61850-8-1MMS protocol. Line topology switching signals (such as switch on / off status) are obtained from the distribution network automation system. The voltage deviation value is based on the Newton-Raphson method power flow calculation model (convergence accuracy). (Iteration count ≤ 10 times), combined with line impedance (Taken from power distribution network GIS system, accuracy) The data is obtained from the calculation of the node power distribution correction. In the environmental data, the real-time irradiance at the 1-minute level is collected by the polycrystalline silicon sensor deployed in the power station (range 0-2000W / ㎡). The ambient temperature and wind speed are obtained by the temperature and humidity sensor and anemometer of the meteorological station. The short-term forecast data at 1 hour (including the irradiance fluctuation range and extreme weather warning) is synchronized with the API interface of the third-party meteorological platform. At the same time, satellite remote sensing meteorological data is introduced to improve the timeliness of the forecast. The equipment fault alarm signal is generated by the fault diagnosis module of the inverter, energy storage BMS, sensor and other equipment, which includes the fault occurrence time, standardized fault code and fault level (such as warning, serious), and is pushed to the data acquisition terminal in real time through the MQTT protocol.

[0021] After these multi-source data are aggregated to the edge computing node through their respective acquisition links, the dual-domain data classification and tagging engine is first activated. Based on the preset "parameter-domain association rule base" combined with real-time grid demand and equipment status, the data is dynamically classified: parameters that directly reflect the equipment's regulation potential (such as the inverter's current active / reactive power, energy storage SOC, PCS response delay, and MPPT tracking accuracy) are classified into the "equipment regulation domain" and assigned a structured label "REG-parameter type-equipment ID" (such as "REG-P-Inv001" representing the active power of inverter 001). The D code corresponds one-to-one with the equipment code in the power plant asset ledger to ensure traceability. At the same time, parameters directly related to the power grid operation requirements (such as grid connection point voltage deviation, regional load gap, topology switching signal, and frequency deviation) are classified into the "Power Grid Demand Domain" and assigned the label "GRD-Parameter Type-Region ID" (such as "GRD-ΔU-Area05" representing the voltage deviation of region A05). The region ID is determined by the power grid dispatching zoning plan. After classification, the label is bound to the original data through SQL statements and stored in the specified partition of the time series database, laying the foundation for data classification for subsequent targeted processing.

[0022] Following the completion of dual-domain data classification and labeling, a multi-dimensional data preprocessing workflow is then executed to ensure data quality: first, outlier filtering is performed using an improved version of the "3σ criterion," and a 10-minute sliding window (1-minute window step) is constructed for each parameter to calculate the mean of the data within the window. (formula , The amount of data within the window is determined by the sampling frequency, such as 100ms-level power data within 10 minutes. ) and standard deviation (formula , (For a single data item within the window), if the data satisfy Then it is judged as an outlier (e.g., irradiance 2500). Outliers such as a sudden voltage spike from 10kV to 12kV are replaced using linear interpolation (the average of the three valid data points before and after the outlier is used as the replacement value) to avoid data loss affecting subsequent calculations. After outlier handling, unit standardization is performed by unifying parameter units using a preset conversion function (e.g., converting energy storage capacity from kWh to kWh). Photovoltaic power is converted from MW to kW and voltage from V to kV. Timestamps are synchronized to the UTC+8 time zone using Python's datetime library and are accurate to the millisecond level, eliminating interference caused by differences in units and time bases. After unifying units and timestamps, data alignment is finally completed. For data with different acquisition frequencies (such as 100ms-level power and 500ms-level voltage), linear interpolation is used to upscale low-frequency data to the 100ms level, ensuring that the dual-domain data are completely matched in the time dimension, providing consistent data support for subsequent "device-grid" dual-domain coupling calculations.

[0023] After all data preprocessing is completed and the data quality meets the standards, the scene dynamic triggering module is activated. Scene judgment logic is built based on multi-dimensional thresholds to adapt to different operating conditions: For typical scenarios, three core conditions must be met simultaneously—1. The maximum fluctuation value of irradiance within 10 minutes. ( , , (These are the maximum and minimum irradiance values ​​within the window, respectively); 2. Equipment failure rate. ( The number of faulty devices currently triggering alarms is obtained by deduplicating device IDs from fault signals in real-time statistics. (The total number of this type of equipment in the power plant, retrieved from the asset ledger); 3. Key power grid parameters are within the rated range (voltage deviation). Frequency deviation , (This refers to the grid's rated voltage). If any condition is not met, an extreme scenario is immediately triggered and the scenario type is marked: If... Then it is marked as "Extreme Scene - Sudden Irradiation Change", if Then it is marked as "Extreme Scenario - Device Cluster Failure". or The scenario is then marked as “Extreme Scenario - Power Grid Parameter Exceeds Limits”. At the same time, the module automatically records key data snapshots at the moment the scenario is triggered (such as the irradiance curve, list of faulty equipment, and power grid voltage waveform at the time of triggering) and stores them in the scenario event library. This provides scenario samples for subsequent model iterations and facilitates fault tracing.

[0024] After the entire process of data collection, classification, preprocessing, and scene determination, four core deliverables are ultimately formed: First, a standardized dataset with dual-domain structured labels. Each data entry includes parameter values, dual-domain labels, device / region ID, and collection timestamp, achieving a data integrity rate of over 99.5%, directly supporting subsequent evaluation calculations. Second, scene type labeling results. For regular scenes, the output is "SCENE-N-timestamp," while for extreme scenes, the output is "SCENE-E-trigger type-timestamp" (e.g., "SCENE-E-irradiation abrupt change-20240520143000"), providing a basis for scene adaptation in subsequent dynamic evaluations. Third, an abnormal data processing report, detailing the original value of abnormal data, the time of occurrence, and the determination criteria (specific details omitted). , The system includes four main functions: (1) the initial value, (2) the replacement value and (3) the processing method, and (4) the scene trigger snapshot, which includes multi-source data screenshots at the trigger time and a judgment threshold comparison table. This can intuitively present the cause of the scene trigger and provide a reference for fault analysis and operation and maintenance optimization.

[0025] After completing the standardized data output with dual-domain labels, scenario type marking (regular scenarios are marked "SCENE-N-timestamp", extreme scenarios are marked "SCENE-E-trigger type-timestamp"), and abnormal data removal records, the output standardized dual-domain data (including parameters such as power, SOC, and response status in the equipment regulation domain, and voltage deviation and frequency deviation in the grid demand domain), scenario type marking results, and equipment rated parameters (inverter rated power) are used as the basis for the output. Rated energy storage capacity Inverter rated apparent power (Etc., retrieve the matching equipment code from the power plant asset ledger) as input, and first calculate the six core regulation indicators.

[0026] When calculating the real-time active power regulation range, the photovoltaic inverter needs to incorporate real-time irradiance data from the environmental data. With ambient temperature The MPPT tracking efficiency under the current operating conditions is obtained by interpolating the irradiance-temperature-power characteristic curves provided by the inverter manufacturer. Then substitute into the formula Determine the maximum active power output and the minimum active power output. According to industry standards, 5% of the rated power is taken, that is... The energy storage end relies on the current SOC value in the device's regulation domain. Calculate the maximum active power output during charging. Take the rated charging power of the energy storage And corrected linearly according to SOC (when SOC≤20%) 20%-80% (It linearly decays to 0 when it reaches 80%-100%). Similarly, the rated discharge power is taken during discharge. Correction, minimum active power output This is the negative value of the rated charging power during reverse charging, i.e. .

[0027] Active power regulation rate By extracting two active power samples within one second of each other from the standardized data. , (timestamp interval) Substitute into the formula Calculations ensure the system can reflect the equipment's maximum power change within one second. Reactive power regulation capacity. Based on the current active power output in the device regulation domain With the inverter's rated apparent power Calculated based on apparent power constraints, the formula is as follows: The negative sign represents the absorption of reactive power, and the positive sign represents the generation of reactive power.

[0028] Voltage regulation accuracy Take the actual voltage at the grid connection point in the power grid demand domain. With the rated voltage of the power grid The absolute value of the difference (obtained from the power grid parameter configuration), the formula is: Command response delay By comparing the timestamps of the adjustment command issuance Timestamp of when device power began to change (Extracted from equipment action record data), substituted into the formula calculate.

[0029] Adjust duration In terms of photovoltaics, the effective irradiance duration predicted in the short-term 1-hour environmental data is directly adopted. The energy storage side combines the current SOC value. Energy storage minimum protection SOC ( (obtained from energy storage BMS parameters) Rated capacity and current adjustment power Substitute into the formula calculate( For the energy storage converter conversion efficiency, an industry-standard value of 0.92-0.95 is used. If the scenario type is marked as an extreme scenario, an additional scenario correction factor needs to be introduced. (Scene of sudden change in irradiation) Equipment failure scenarios (Determined based on training with historical extreme cases) and power grid disturbance. (Taken from load gap data in the power grid demand domain), the formula is modified to: This allows for the correction of indicators under extreme scenarios.

[0030] After calculating the six core indicators, differentiated labels for the equipment are generated based on the indicator thresholds: high-response equipment must simultaneously meet the active power regulation rate. And command response delay Large-capacity equipment must meet the requirement that the real-time active power regulation range covers more than 80% of the rated power. Or, the reactive power regulation capacity covers more than 80% of the rated apparent power. Long-duration equipment must meet the requirement that the regulation duration is not less than twice the regulation duration required by the power grid. , (Refrigeration duration demand parameters taken from the power grid demand domain).

[0031] The final output includes extreme scenario correction values ​​(such as...) , The results of the six core regulation indicators, as well as the differentiated labels that clearly mark the equipment codes and label types (high response / large capacity / long duration), provide accurate quantitative data and characteristic identification of equipment regulation capabilities for carrying out dual-domain coupled computing and cluster aggregation modeling in hierarchical dynamic evaluation.

[0032] In the tiered dynamic evaluation phase, the results of the six core regulation indicators (including extreme scenario correction values, such as the real-time active power regulation range) are first used as the basis for evaluation. Active power regulation rate Reactive power regulation capacity Voltage regulation accuracy Command response delay Adjust duration ) and equipment differentiation labels (high response / high capacity / long duration), combined with real-time grid demand (load gap) Voltage regulation requirements Adjusting duration requirements ) and 1-hour environmental prediction data (such as predicted irradiance) Predicted temperature Initiate dual-domain coupling evaluation: First, acquire real-time power grid topology data (such as line impedance). Node voltage ) and current operating parameters (such as grid frequency) Total load power Based on this, a sensitivity matrix is ​​dynamically generated. Taking a voltage regulation scenario as an example, the formula is used to generate the sensitivity matrix. Calculate the first The coupling weight of reactive power regulation of the equipment to voltage deviation, among which For the first Coupling weights of the devices Investment in the equipment (Taken from reactive power regulation capacity index) Voltage correction amount generated after reactive power (obtained through power grid flow calculation, i.e., based on line impedance) (Calculate node voltage change with node power change). The initial voltage deviation of the power grid without any equipment regulation (based on real-time collected node voltages) With rated voltage Difference acquisition, i.e. ), The total number of devices participating in voltage regulation; in frequency regulation scenarios, the coupling weight formula is adjusted to... ,in For the first The coupling weight of the active power regulation rate of the equipment to the frequency deviation. For this device Frequency correction amount generated during adjustment (based on grid inertia constant) (Calculation of power change) The initial frequency deviation without equipment adjustment ( , (The frequency is the rated frequency of the power grid). The coupling weight is updated every 500ms according to the power grid operating conditions. Then, the adjustment contribution of a single device is calculated in combination with the device adjustment index, such as the voltage regulation contribution. ( The total reactive power regulation capacity of all equipment, i.e. ), frequency modulation contribution ( The total active power regulation rate for all devices, i.e. This allows for the quantification of the value of a single device in alleviating grid demand.

[0033] After completing the dual-domain coupling assessment, the cluster aggregation assessment phase begins: first, the aggregation priority is determined based on the current core needs of the power grid; if the core need is frequency regulation (i.e., , If the frequency deviation threshold is set to 0.1Hz, then the marked "high-response" devices will be selected first. and , (Rated power of the equipment), through the formula Calculate the overall active power regulation rate of the cluster, where For the integrated frequency modulation rate of the trunking, For the first Rated power of high-response equipment (taken from equipment parameter library). Total rated power for high-response equipment ( If the core requirement is voltage regulation (i.e.) If the voltage deviation threshold is set to 0.05kV, then "large capacity" equipment will be prioritized for screening. , (Rated reactive power capacity of the equipment), using the formula Calculate the total reactive power regulation capacity of the cluster, where The total reactive power capacity for voltage regulation in the cluster. For the first The failure coefficient of large-capacity equipment is set at 0 (0 for no faults, 0.1 for single module failures, and 1 for overall failures, based on equipment fault signals); simultaneously, abnormal equipment (equipment with a failure rate exceeding 5%) is automatically eliminated. , (Total number of devices of this type) to avoid lowering the cluster's regulation capability, and then formulate differentiated control schemes based on the aggregation results, such as assigning short-term high-frequency regulation commands to high-response devices and assigning continuous voltage regulation commands to large-capacity devices.

[0034] Subsequently, boundary prediction and assessment are conducted. First, the current scene type is determined (based on scene labeling). If it is an extreme scene (such as an irradiance increase of ≥500 ppm within 10 minutes), then... If SOC ≤ 20% or equipment failure rate ≥ 3%, then the adjustment index boundary is adjusted based on 1-hour environmental prediction data and real-time equipment status: taking the duration of energy storage adjustment as an example, the formula is used to adjust the boundary. Calculate the adjustment duration boundary in extreme scenarios, where To adjust the duration of cluster operation in extreme scenarios, This refers to the duration of adjustment in a typical scenario (taken from the adjustment duration metric). The scene correction coefficients are set to 1.2 for strong radiation scenes and 0.8 for equipment failure scenes, and are generated based on a historical extreme scene case library. Power grid disturbance ( , (This refers to the current total power generation from new energy sources); the boundary of the active power regulation range is determined by the formula... Correction, among which This refers to the range of active power adjustment for the cluster under extreme scenarios. This represents the active power regulation range under normal scenarios (taken from the real-time active power regulation range index). This represents the current irradiance (data collected in real time). The predicted irradiance is calculated for one hour; after correction, the boundary prediction value is compared with the grid demand. ( (for grid regulation time requirements) or If the deviation exceeds 20%, an early warning signal will be triggered. The warning level is divided according to the degree of deviation (a level 1 warning is given if the deviation exceeds 20%, and a level 2 warning is given if the deviation is between 10% and 20%).

[0035] Finally, a comprehensive evaluation score and demand matching degree are calculated: the comprehensive score is calculated using a weighted summation formula. ,in For the first The overall evaluation score of the equipment (out of 1). For the first Voltage regulation accuracy of the equipment (taken from voltage regulation accuracy index). The maximum command response latency across all devices. Maximum adjustment duration among all devices; overall cluster score , The total power of all participating equipment is determined; the demand matching degree is expressed by the formula. Calculation, where The degree of matching between the cluster and the power grid demand (value 0-1). Minimum active power regulation rate required for grid frequency regulation (based on) (and power grid inertia calculation). Minimum reactive power regulation capacity required for grid voltage regulation (based on) (Based on line impedance calculation), the final output is the adjustment contribution of each device ( , Cluster aggregation adjustment capability () , ), extreme scenario boundary prediction values ​​( , ) and early warning signals, equipment / cluster comprehensive evaluation score ( , ) and demand matching degree ( ).

[0036] In the evaluation result application and strategy output phase, the first step is to obtain the overall score (equipment overall score). Cluster Comprehensive Score ), Demand matching degree Cluster aggregation solutions (such as high-response equipment cluster frequency regulation solutions and large-capacity equipment cluster voltage regulation solutions), and boundary early warning signals (including early warning levels and deviation data, such as the corresponding level one early warning). Using this as input, the power grid dispatch strategy is initiated: first, based on the equipment's regulation contribution (voltage regulation contribution)... FM contribution (Taken from the dual-domain coupled evaluation results) and the comprehensive score to construct a scheduling priority calculation model, the formula is as follows: ,in For the first The scheduling priority of each device (values ​​range from 0 to 1, with higher values ​​indicating higher priority). The contribution of the equipment to the core needs of the power grid (if the current core need is frequency regulation, then take...). For voltage regulation, take The core demand is determined through real-time demand parameters of the power grid, such as... (Then it is determined to be a frequency modulation requirement). Matching degree between individual devices and grid demand (through cluster matching degree) Segmented according to the proportion of rated power of the equipment, i.e. , Rated power of the equipment (Set the total power for the cluster); based on the priority ranking results and the cluster aggregation scheme, issue differentiated instructions, such as issuing "100ms-level active power adjustment instructions, with adjustment range" to the top 30% of high-response devices. ( (For active power regulation rate of equipment), issue "reactive power regulation command" to large-capacity equipment, with target reactive power output. ( (To adjust the reactive power capacity of the equipment), and at the same time synchronize the instructions to the equipment controller via the Modbus-TCP protocol to ensure that the instructions are executed in real time.

[0037] After the power grid dispatch strategy is output, the power plant operation and maintenance strategy is formulated: first, based on the generated equipment differentiation tags and comprehensive scores, the key points of operation and maintenance are extracted by category—for high-response equipment ( and ), with a focus on monitoring command response delays Contribution to FM ,when 10% or more up from the historical average When the power consumption drops by 5%, a "Response Performance Optimization Recommendation" is generated, which includes recommendations such as "Check the PCS command receiving module" and "Calibrate the power change timestamp acquisition accuracy"; for long-duration energy storage devices ( The focus is on analyzing the duration of the adjustment. The battery SOC degradation rate (calculated using 30 consecutive days of SOC change data collected by the BMS, using the formula is...) ),like shorten by 15% or The system sends out a "Battery Health Status Detection Work Order," which specifies the detection items as "Individual Cell Voltage Balance" and "PCS Conversion Efficiency." "(Taken from the indicator calculation parameters); if the output boundary warning signal (such as the first-level warning corresponding to)" If a snapshot (taken from the scenario determination results) is automatically triggered based on the scenario, a pre-maintenance work order will be generated. For example, when an "Extreme Scenario - Sudden Irradiance Change" warning is issued, the work order will include "Check the MPPT tracking efficiency of the photovoltaic inverter". Targeted operations such as "cleaning the surface of the irradiance sensor" are performed, and work orders are pushed to the maintenance terminal via the MQTT protocol and simultaneously recorded in the equipment maintenance log.

[0038] Finally, an absorption optimization strategy was formulated: first, based on the boundary prediction values ​​of extreme scenarios (such as... , ) and grid demand ( , The formula for adjusting the regional renewable energy absorption capacity is as follows: ,in This refers to the adjusted absorption capacity under extreme scenarios. This represents the basic absorption capacity under normal scenarios (calculated based on the average absorption data of normal scenarios over the past 30 days). The range of active power adjustment of the cluster under extreme scenarios (taken from boundary prediction). The grid load gap is represented by data from the grid demand domain; based on the corrected absorption capacity, an adjustment suggestion for the absorption rate is output, such as when... At that time, it was recommended to "reduce the regional photovoltaic absorption rate from 95% to 90%, prioritizing energy storage absorption"; simultaneously, based on the distribution of equipment tags, resource allocation recommendations were provided, for example, if the proportion of high-response energy storage equipment is less than 20%, it was recommended to "add 2 units with a rated power of 500kW". Regarding energy storage devices, if the proportion of large-capacity inverters is less than 30%, it is recommended to "reduce the rated apparent power of the existing 10 inverters". The upgrade from 1000kVA to 1250kVA was completed. The final output of the power grid optimization plan was in the form of a visual report, which included a comparison curve of power grid capacity, a resource allocation list and expected benefits (such as a 2% reduction in curtailment rate after the power grid is adjusted). The report was uploaded to the power grid dispatch center and the power plant management platform to support subsequent resource allocation decisions.

[0039] During the feedback and model iteration phase, the strategy execution results (including equipment command response data and operation and maintenance effect data) and early warning verification data are used as inputs to initiate the strategy execution effect verification: for power grid dispatch command responses, the actual response delay of the equipment is extracted. (Obtained from the time difference between command reception and power change recorded by the device controller, taken from command response data) and the response delay threshold required by the issued command. (For example, if the command threshold for a high-response device is set to 50ms), the formula can be used to... Calculate the response deviation rate, if Devices are judged as "responsive and up to standard" if they fail to meet the standard, and otherwise marked as "responsive and up to standard"; the percentage of compliant devices is also tallied. ( In response to the number of compliant equipment, (Total number of devices receiving instructions), if The scheduling strategy is therefore effective overall. Regarding the effectiveness of power plant operation and maintenance, compare the changes in key indicators before and after maintenance, such as the active power regulation rate of high-response equipment after maintenance. (Taken from operation and maintenance effect data) Compared with before operation and maintenance (Taken from historical indicators), through formula Calculate the improvement rate, if The maintenance measures are effective; based on the early warning verification data, the actual failure rate after the early warning signal is triggered is statistically analyzed. ( This represents the actual number of devices that malfunctioned after the warning was issued. (Total number of devices that triggered the warning), if If the warning rules are reasonable, then the warning thresholds need to be optimized.

[0040] After verifying the execution effect, fine-tuning of model parameters is initiated based on the verification results: for dual-domain coupled weights (such as voltage regulation scenarios) The method employs mini-batch gradient descent in incremental learning, with the "response bias rate" as the key factor. Using "minimize" as the objective function, 500 device data points marked as "abnormal response" in the validation were selected as incremental samples, and the learning rate was set to... Through formula Update coupling weights ( For the updated weights, (This refers to the original weights). For example, if an inverter has a high response deviation rate, its corresponding reactive power regulation coupling weight is reduced, lowering its subsequent scheduling priority. For extreme scenario correction rules, if the early warning verification is in progress... Then adjust the scene correction coefficient. For example, the original coefficient in the case of equipment failure. According to the formula Revised to This improves the conservatism of indicator calculations in extreme scenarios; simultaneously, if the actual adjustment duration deviates from the predicted value by more than 20% in scenarios of sudden irradiation changes, the irradiation impact factor in the boundary prediction formula is optimized, such as by... In the formula With irradiation fluctuation value (Taken from environmental data) Hooked, corrected to This enhances the accuracy of predictions.

[0041] Finally, combining equipment lifecycle data with power grid topology change records, the equipment tagging criteria were adjusted: for equipment aging, the battery cycle count collected by the BMS was used. (Taken from the power plant asset ledger) and rated cycle life (Equipment factory parameters), calculate aging coefficient ,like (If the equipment enters the late aging stage), the response latency threshold for high-response equipment will be relaxed from 50ms to 60ms, using the formula... Dynamic adjustments are implemented; for changes in the power grid topology (such as a signal indicating the addition of two distribution lines during a line topology switch), the regional weight during cluster aggregation is adjusted, and the aggregation weight of the original region ID "Area05" is changed. Based on the load ratio after the addition of new lines (The ratio of newly added line load to total regional load is taken from data from the power grid dispatch center), expressed by the formula. Increase the aggregation priority of devices in this area.

[0042] The final output consists of three types of results: first, a strategy execution performance report, including the response achievement rate. Operation and maintenance indicator improvement rate Early warning accuracy The system includes: 1) core data, a list of abnormal equipment and improvement suggestions; 2) updated model parameters, including the adjusted dual-domain coupling weight matrix, extreme scenario correction coefficient table and boundary prediction formula, which can be directly imported into the evaluation model; and 3) equipment label adjustment suggestions, clarifying the label threshold correction scheme for equipment with different aging levels and the regional aggregation weight adjustment rules after changes in power grid topology, providing an updated basis for label generation, and storing all iteration results in the model version library for easy traceability and secondary iteration.

[0043] For example, taking the operating conditions of a 100MW photovoltaic power station + 20MWh energy storage power station from 10:00 to 11:00 on June 15, 2024 as an example, data is obtained through the data acquisition link: Active power is collected in 100ms increments via an inverter monitoring module (Modbus-TCP protocol). (Rated power) reactive power MPPT tracking accuracy (Meets the ≥98% threshold set in the document), cooling fan PWM duty cycle is 60% (determined to be normal operation); Data collected in 50ms increments via PCS+BMS (IEC61850 protocol) indicates charge / discharge power. (Negative sign indicates discharge), current SOC value PCS response delay (Below the document's limit of ≤100ms), individual battery cell voltage difference (Meets the equilibrium judgment criteria); A high-precision power quality monitoring device (0.2-level accuracy) collects data at the 500ms level, including the actual voltage at the grid connection point. (Rated voltage) ), power grid frequency Regional load gap Line impedance (Taken from the power distribution network GIS system, accuracy ±5%) Polycrystalline silicon sensors collect real-time irradiance data in 1-minute increments. Ambient temperature Connect to a third-party meteorological platform API to obtain 1-hour forecast irradiance. Simultaneously, satellite remote sensing meteorological data (resolution 1km×1km) is introduced. The device diagnostic module pushes alerts via the MQTT protocol; no fault alarms are detected, and the fault ratio is reported. ( This represents the number of faulty devices. (Total number of devices of this type).

[0044] Based on the classification of "parameter-domain association rule base": Equipment regulation domain: Assign structured labels, such as “REG-P-Inv001” (Inverter 001 active power 85kW), “REG-SOC-ESS01” (Energy storage 01 current SOC 75%), “REG-Tresp-ESS01” (Energy storage 01 response delay 35ms). The equipment ID in the label corresponds one-to-one with the power plant asset ledger code. Grid demand domain: Assign the label “GRD-ΔU-Area01” (voltage deviation in area A01) The regions are categorized as follows: “GRD-ΔP-Area01” (A01 area load deficit 8MW) and “GRD-f-Area01” (A01 area frequency deviation 0.05Hz). The region ID is determined according to the power grid dispatching zoning plan. After classification, tags are bound using SQL statements and stored in a time-series database (partitioned into three levels: document “time + region ID + equipment type”, with time partitioning at 1-hour granularity).

[0045] Outlier filtering: Adopts the improved document "3" The "Guideline" constructs a 10-minute sliding window for irradiation data. ), calculate the mean Standard deviation No data satisfies No replacement required; Unit and time synchronization: The power unit is kW and the voltage is kV. The timestamp is synchronized to the UTC+8 time zone (accurate to milliseconds). Data alignment: 500ms-level voltage data is upscaled to 100ms-level through linear interpolation to match power data; Scene determination: Irradiation fluctuation over 10 minutes Failure rate Voltage deviation Frequency deviation This meets the typical document scenario requirements and is marked "SCENE-N-202406151000".

[0046] Photovoltaic inverter: According to the formula, combined with , Insertion is worthwhile Maximum active power output Minimum active power output ; Energy storage devices: (20%-80% range), rated discharge power Maximum active power output (Discharge), minimum active power output (Charging, the negative sign indicates reverse power).

[0047] Extracting power values ​​at 1-second intervals from standardized data (formula) ): Inverter: , , ; Energy storage: , , .

[0048] According to the apparent power constraint formula Inverter rated apparent power Currently meritorious , (A negative sign indicates the absorption of reactive power, and a positive sign indicates the generation of reactive power.)

[0049] According to the formula : .

[0050] According to the formula ( The time the instruction was issued. (Power change time): Energy storage command issuance =10:00:00.000, power change =10:00:00.035, .

[0051] Photovoltaic side: The duration of irradiation is directly predicted using a 1-hour forecast. ; Energy storage: According to the formula for conventional scenarios , Rated capacity , (Industry standard value), current regulating power , .

[0052] Determined based on indicator thresholds: High-response type: Energy storage (After conversion) and Marked as "High Response Type - ESS01"; High-capacity type: Inverter Marked as "Large Capacity Type - Inv001"; Long-term sustainable energy storage (Power grid demand regulation duration) ), marked "long-lasting type - ESS01".

[0053] Power flow calculations using the Newton-Raphson method (convergence accuracy) Sensitivity matrix generation: A total of 500 voltage-regulating inverters are involved, with each unit put into operation... Reactive power generates voltage correction. Initial voltage deviation According to the coupling weight formula , .

[0054] Total reactive power regulation capacity Inverter voltage regulation contribution ; Energy storage frequency regulation contribution (Similarly, the calculation is based on the frequency correction and the total adjustment rate).

[0055] Based on the aggregation strategy, 10 high-response energy storage units (total rated power) were selected. According to the formula single unit , , .

[0056] Select 300 high-capacity inverters (without faults) According to the formula , .

[0057] Boundary prediction: No correction is needed in normal scenarios. No warning; Overall equipment score: according to the formula Inverter , , , , ; Cluster requirement matching degree: according to the formula Minimum frequency regulation rate of power grid Minimum voltage regulation capacity , (Values ​​range from 0 to 1, with 1.16 indicating a perfect match with redundancy).

[0058] According to the scheduling priority formula : Energy storage , (Core requirements for FM) , (Higher priority than inverter); Instructions issued: Three high-response energy storage units execute "100ms-level frequency regulation commands." "50 high-capacity inverters execute reactive power regulation commands," The command is synchronized to the device controller via the Modbus-TCP protocol.

[0059] High-response energy storage: (Historical average 32ms, up 9.4%) 10%), no optimization required; Long-duration energy storage: SOC decay rate The system pushes out a "monthly battery balance test work order," specifying the test results for individual cell voltage balance and PCS conversion efficiency. No warning signal: No need to generate pre-maintenance work orders.

[0060] According to the formula for absorption capacity : Typical scenarios Currently, the output of new energy sources is 85 + 12 = 97. It is recommended to "maintain a photovoltaic absorption rate of 95% and an energy storage absorption rate of 100%, with no need for new equipment for the time being," and output a comparison curve of absorption capacity and a resource allocation list.

[0061] According to the verification formula: Response compliance rate: Actual response delay of 3 energy storage units threshold Response deviation rate compliance rate ; Operation and maintenance results: After inverter operation and maintenance Improvement rate ; Early warning accuracy: No early warning, accuracy is 100%.

[0062] Dual-domain coupled weights: No outliers, maintain ; Equipment tag threshold: number of energy storage cycles Next, rated life Second, aging coefficient The label threshold remains unchanged.

[0063] Strategy Execution Performance Report: Response Compliance Rate 100%, Operation and Maintenance Indicator Improvement Rate 5%, Early Warning Accuracy Rate 100%, Attached is a list of abnormal devices (no abnormalities); Updated model parameters: dual-domain coupling weight matrix, extreme scenario correction coefficients (Sudden change in irradiation); Recommendation for adjusting equipment labels: The current judgment criteria are applicable and no adjustment is required.

[0064] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0065] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0066] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A method for dynamically evaluating real-time adjustment capability of a power generation device of a new energy power plant, characterized in that, The method comprises: Collect multi-dimensional data, obtain standardized data set through light pretreatment; Based on the standardized data set, combined with the rated parameters and real-time operating state of the equipment, the core regulation index of the equipment is calculated, the equipment differentiation label is generated, the double-domain coupling weight algorithm and the cluster aggregation logic are used to determine the equipment regulation contribution and the cluster aggregation ability, and after excluding invalid data, the equipment regulation ability evaluation parameter is generated; The equipment regulation ability evaluation parameter is input into the hierarchical dynamic evaluation model, combined with the real-time demand of the power grid and the environmental prediction data, the risk prediction result is obtained; According to the risk prediction result, the dynamic threshold algorithm is used for abnormal judgment, and the equipment label adjustment suggestion and system optimization data are generated according to the judgment result; The equipment regulation ability evaluation parameter, system optimization data and scene type label are input into the evaluation result application model, combined with the power grid dispatching demand and power station operation and maintenance target, the equipment dispatching priority is calculated, and the differentiated dispatching strategy is generated.

2. The method according to claim 1, wherein, The specific obtaining process of the standardized data set is as follows: Obtain photovoltaic inverter operation data, energy storage device data, power grid operating condition data, environmental data and equipment failure data; divide the parameters into equipment adjustment domain and power grid demand domain according to the preset rule library, respectively give corresponding labels to the parameters of the two domains, bind the labels through SQL and store them in the time sequence database; adopt 3 The criterion is used to filter abnormal values in a sliding window, and out-of-range data is replaced by linear interpolation. The units of all parameters and timestamps are unified, and the data with different collection frequencies are aligned through linear interpolation.

3. The method of claim 2, wherein the method further comprises: The specific obtaining process of the equipment differentiation label is as follows: Combined with the standardized data set, the rated parameters and the real-time operating state of the equipment, the real-time active regulation range, the active regulation rate, the reactive regulation capacity, the voltage regulation accuracy, the instruction response delay and the regulation duration of the equipment are calculated; In extreme scenarios, introduce scene correction coefficient and grid disturbance power to correct the regulation duration; According to the preset index threshold, high response type equipment needs to meet the requirements of active regulation rate and instruction response delay, large capacity type equipment needs to meet the requirements of real-time active regulation range covering rated power ratio or reactive regulation capacity covering rated apparent power ratio, and long duration type equipment needs to meet the requirements of regulation duration not less than specified multiple of grid demand regulation duration; According to the judgment result, the equipment generates differentiated label, and the equipment code and label type are marked in the label.

4. The method of claim 3, wherein the method further comprises: The specific obtaining process of the equipment regulation ability evaluation parameter is as follows: Using double-domain coupling weight algorithm, combined with the real-time topology data and working condition parameters of the power grid, the sensitivity matrix is generated, the regulation contribution of single equipment to the power grid demand is calculated; According to the core demand of the power grid, the cluster aggregation is carried out based on the equipment differentiation label, the cluster comprehensive regulation rate and the total reactive regulation capacity are calculated, and the invalid equipment data with fault proportion exceeding the standard is excluded; The single equipment regulation contribution, cluster aggregation ability, extreme scene boundary prediction value, equipment and cluster comprehensive evaluation score and demand matching degree are integrated to form the equipment regulation ability evaluation parameter covering the individual regulation potential of the equipment, the cluster cooperation ability and the adaptability to the power grid demand.

5. The method of claim 4, wherein the method further comprises: The specific obtaining process of the risk prediction result is as follows: The device adjustment capability evaluation parameter is taken as the core input, combined with the real-time demand data of the power grid and the 1-hour environment prediction data, the hierarchical dynamic evaluation model is introduced, and the adjustment contribution of a single device to the power grid demand is determined through double-domain coupling evaluation; based on the differentiated labels of the devices, the cluster aggregation evaluation is carried out to determine the comprehensive adjustment capability of different types of device clusters, and the index boundary in the extreme scenario is corrected through boundary prediction evaluation; the results of double-domain coupling evaluation, cluster aggregation evaluation and boundary prediction evaluation are integrated to output the risk prediction results including the boundary prediction value in the extreme scenario, the early warning signal, the comprehensive evaluation score of the device and the cluster, and the matching degree of the device and the power grid demand.

6. The method of claim 5, wherein the method further comprises: The specific obtaining process of the device label adjustment suggestion and the system optimization data is as follows: Based on the policy execution result and the early warning verification data, the policy execution effect verification is started, the effectiveness of the power grid dispatching strategy is determined by calculating the device instruction response deviation rate and the proportion of devices meeting the standard, the effect of the operation and maintenance measures is verified by comparing the core index improvement rate before and after the operation and maintenance, and the rationality of the early warning rule is verified by counting the actual fault occurrence rate after the early warning; Based on the verification result, the small batch gradient descent method is used to fine-tune the double-domain coupling weight with the minimum response deviation rate as the target, the influence factor in the boundary prediction formula is optimized by combining the extreme scenario correction coefficient and the fault occurrence rate, and the updated model parameters are generated; the aging coefficient is calculated by combining the battery cycle number, the charging and discharging depth, the environmental temperature and the rated cycle life in the device full life cycle data, the regional aggregation weight is adjusted according to the load proportion and the line impedance after the change of the power grid topology, and the device label determination threshold is dynamically adjusted; the above data is integrated to form the system optimization data and the device label adjustment suggestion respectively.

7. The method of claim 5, wherein the method further comprises: The specific obtaining process of the differentiated dispatching strategy is as follows: Taking the device adjustment capability evaluation parameter, the system optimization data and the scene type label as the input, combining the real-time demand of the power grid and the operation and maintenance target of the power station, the power grid dispatching strategy is formulated; a dispatching priority calculation model is constructed based on the device adjustment contribution and the comprehensive evaluation score, the device adjustment contribution is obtained from the double-domain coupling evaluation result, the comprehensive evaluation score is obtained by weighted summation, the matching degree of a single device and the power grid demand is split from the cluster matching degree according to the proportion of the device rated power, and the dispatching priority of each device is calculated; according to the priority sorting result, the differentiated instructions are issued to different types of devices according to the cluster aggregation scheme, the short-time high-frequency adjustment instruction is allocated to the high-priority high-response device, and the continuous voltage regulation instruction is allocated to the large-capacity device, and the instruction is synchronized to the device controller through the corresponding communication protocol to form the differentiated dispatching strategy adapting to the power grid demand and the device characteristics.

8. A new energy power station power generation equipment real-time regulation capability dynamic evaluation system, characterized in that, The system is used to execute the real-time adjustment capability dynamic evaluation method of the new energy power station power generation device.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to realize the real-time adjustment capability dynamic evaluation method of the new energy power station power generation device in any one of claims 1-7.

Images