A multi-level management method for real-time data of new energy station power generation equipment

Abstract

The application discloses a kind of new energy field station power generation equipment real-time data multilevel management method, it is related to the data multilevel management technical field of photovoltaic power generation field station, this method is by obtaining field station electronic distribution map and divides sub-region, and collects voltage, current, temperature and power and so on multidimensional data at equipment and sub-region level, after abnormal elimination and normalization processing, constructs AI dynamic multilevel management model, exports management state vector and response evaluation result.Through calculating fluctuation abnormal coefficient, sub-region management response consistency coefficient and management synergy efficiency coefficient, respectively corresponding component, sub-region and cross-zone level operation state determination, and with the comparison analysis of pre-set three kinds of threshold, gradually trigger management strategy.The application can realize photovoltaic field station equipment level, sub-region level, cross-zone level and field station level real-time collaborative management and dynamic optimization, improve system overall operation stability and strategy response efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data multi-level management of photovoltaic power generation stations, in particular to a multi-level management method for real-time data of new energy station power generation equipment. BACKGROUND

[0002] With the rapid development of new energy technology, new energy power generation stations such as photovoltaic and wind power are being constructed on a large scale and gradually put into operation across the country. These stations generally have the characteristics of wide geographical distribution, diverse equipment types, and complex operating states. In particular, in photovoltaic power generation systems, voltage, current, and temperature data at the component level, power output and collaborative control parameters at the sub-region level, and comprehensive performance indicators and strategy responses at the station level constitute a typical multi-level operating data system.

[0003] However, the current energy efficiency management of new energy stations is still relatively extensive, relying on traditional methods such as fixed thresholds, empirical rules, and regular maintenance, which cannot effectively address the real-world problems of diverse data sources, frequent state changes, and complex level coordination. In particular, during photovoltaic power generation, the collected operating data often has problems such as noise interference, parameter drift, and time asynchronization due to factors such as natural environmental fluctuations, equipment aging, and electrical interference, directly affecting the accuracy of state judgment and strategy execution.

[0004] In addition, the existing energy consumption management system lacks understanding of the dynamic coupling relationship between different data levels, and often cannot accurately distinguish between component-level abnormalities, sub-region fluctuations, or station-level strategy imbalances when executing management strategies, leading to management response deviations and even system coordination failures. A more prominent problem is that the existing system lacks unified fusion and intelligent analysis capabilities for multi-source heterogeneous data, making it impossible to achieve component-level, sub-region-level, cross-region-level, and station-level coordinated management. The data interface, processing algorithm, and response strategy between different levels are fragmented, increasing the complexity of system operation and maintenance and limiting the improvement of intelligent and refined management of new energy stations. SUMMARY

[0005] To address the deficiencies of the prior art, the present application provides a multi-level management method for real-time data of new energy station power generation equipment to solve the problems mentioned in the background art.

[0006] To achieve the above purpose, the present application is implemented by the following technical solution: a multi-level management method for real-time data of new energy station power generation equipment, comprising the following steps:

[0007] Step 1: Obtain the electronic distribution map of the photovoltaic station and divide it into sub-regions; collect voltage, current, and temperature at the component level; and collect output power and collaborative behavior data at the sub-region level.

[0008] Step two, collect data voltage, current and temperature, and perform abnormal rejection, normalization and sliding window smoothing processing;

[0009] Step three, construct a multi-scale deep learning model, fuse device to station multi-level data and governance history information, train to generate AI dynamic multi-level governance model with governance perception, structure modeling and strategy adaptation capability, output governance state vector, feature map and response evaluation results, for assisting intelligent strategy generation and dynamic adaptation;

[0010] Step four, through real-time monitoring of the component operating state of the sub-area of the photovoltaic station, combining the data mean and fluctuation level of the historical operation period of the sub-area, calculating the fluctuation abnormality coefficient BDYC, and comparing and analyzing with the first threshold Q1, to judge whether the component fluctuation in the sub-area is normal, if not, give strategy;

[0011] Step five, by monitoring the response difference of multiple sub-areas under unified governance strategy, calculating the sub-area governance response consistency coefficient XYYZ, and comparing and analyzing with the second threshold Q2, to judge whether the governance response consistency of the sub-area is qualified, if not, give strategy;

[0012] Step six, through real-time monitoring of the response consistency, strategy coupling degree and overall complementary effect among multiple sub-areas after governance, calculating the governance coordination efficiency coefficient ZLXN, and comparing and analyzing with the third threshold Q3, to judge whether the cross-regional governance coordination effect is qualified, if not, give strategy.

[0013] Preferably, step one includes:

[0014] S11, pre-acquire the electronic distribution map of the photovoltaic power station, divide the photovoltaic power station into several sub-areas, and mark on the electronic distribution map;

[0015] S12, real-time monitoring of the equipment of the sub-area, through deploying edge computing perception device and temperature sensor in photovoltaic module, collecting original running data at module level, including voltage value, current value and temperature value;

[0016] S13, by deploying intelligent data gateway and multi-sub-area data synchronization interface, collecting original running data at sub-area level, including output power and sub-area collaborative interaction original behavior data.

[0017] Preferably, step two includes:

[0018] S21, by adopting IQR outlier detection and Min-Max normalization technology, the original sensing data of voltage, current and temperature are subjected to outlier elimination and normalization processing; by introducing a sliding window filtering algorithm, the voltage, current and temperature data are subjected to time series smoothing processing, short-term fluctuation interference is suppressed, and trend characteristics are retained.

[0019] Preferably, step three comprises:

[0020] S31, by constructing a deep learning model with multi-scale perception ability, the model is based on device layer-sub-region layer-cross-region layer-station layer, four-layer management structure, by fusing multi-dimensional running state data, historical management strategy feedback and response results after management, a neural network model with management perception, structure dependent modeling and strategy adaptation evaluation ability is established, through multi-dimensional perception of multi-device component running data, sub-region management behavior, regional collaborative strategy and global station management strategy, a training data set with fused structure time series characteristics is constructed; a multi-task label set is constructed based on management history data; and joint training is carried out through a unified objective function, the comprehensive modeling ability of the model for different hierarchical management characteristics is improved, and finally the trained model is used as an AI dynamic multi-level management model, which outputs photovoltaic power station management state vector, structure dependent feature map and multi-dimensional response evaluation matrix, and outputs the auxiliary generation and dynamic adaptation of intelligent management strategy.

[0021] Preferably, step four comprises:

[0022] S41, by real-time monitoring the component running state of the sub-region of the photovoltaic station, combining the data mean and fluctuation level of the historical running period of the sub-region, and after non-dimensional processing, the fluctuation abnormality coefficient BDYC is calculated and obtained, and the formula is as follows:

[0023]

[0024] In the formula, n represents the number of components participating in monitoring in the sub-region, represents the current period voltage value of the i-th component, represents the historical mean value of the component voltage, represents the historical standard deviation value of the voltage, represents the current period current value of the i-th component, represents the historical mean value of the component current, represents the historical standard deviation value of the current, represents the difference between the temperature of the i-th component and the temperature of the adjacent component, represents the reference temperature value of the component environment in the period, represents an extremely small constant to prevent the denominator from being zero, and w1, w2 and w3 represent weight coefficients.

[0025] Preferably, step four further comprises:

[0026] S42, by presetting the first threshold Q1 in advance, and comparing and analyzing the fluctuation abnormal coefficient BDYC with the first threshold Q1, a first evaluation result is obtained, including:

[0027] When the fluctuation abnormal coefficient BDYC < the first threshold Q1, it indicates that the component fluctuation in the sub-region is normal, and there is no deviation from the historical stable state, and continuous monitoring is performed.

[0028] When the fluctuation abnormal coefficient BDYC ≥ the first threshold Q1, it indicates that the component fluctuation in the sub-region is abnormal, and there is a deviation from the historical stable state, a first warning instruction is triggered, and a first strategy is generated: marking the abnormal occurrence time, component number and physical coordinates; comparing with the historical database, identifying repeated abnormal points, and performing thermal control and voltage adjustment governance measures on the repeated abnormal points; starting the photovoltaic field area governance response consistency analysis mechanism.

[0029] Preferably, step five includes:

[0030] S51, when the first warning instruction is received, the photovoltaic field area governance response consistency analysis mechanism is started, the response difference of multiple sub-regions under the unified governance strategy is monitored, and the fluctuation abnormal coefficient BDYC is combined for dimensionless processing, and a sub-region governance response consistency coefficient XYYZ is calculated and obtained, and the formula is as follows:

[0031]

[0032] In the formula, m represents the number of sub-regions participating in governance, represents the average output power of the jth sub-region after governance, represents the reference average output power before governance, represents a minimum constant to prevent the denominator from being zero, represents the fluctuation abnormal coefficient standard deviation of all devices in the sub-region, and represents a weight coefficient.

[0033] Preferably, step five further includes:

[0034] S52, by presetting the second threshold Q2 in advance, and comparing and analyzing the sub-region governance response consistency coefficient XYYZ with the second threshold Q2, a second evaluation result is obtained, including:

[0035] When the sub-region governance response consistency coefficient XYYZ ≤ the second threshold Q2, it indicates that the governance response consistency of the sub-region is qualified, the strategy remains unchanged, and continuous monitoring is performed.

[0036] When the sub-region governance response consistency coefficient XYYZ is greater than the second threshold Q2, it indicates that the governance response consistency of the sub-region is unqualified, a second early warning instruction is triggered, and a second strategy is generated: starting the governance sub-region redistribution and strategy redeployment mechanism.

[0037] Preferably, step six includes:

[0038] S61, when receiving the second early warning instruction, starting the governance sub-region redistribution and strategy redeployment mechanism, and monitoring the response consistency, strategy coupling degree, and overall complementary effect among the multiple sub-regions in real time after governance. After dimensionless processing, the governance synergy efficiency coefficient ZLXN is calculated and obtained, and the formula is as follows:

[0039]

[0040] In the formula, represents the standard deviation of the sub-region governance response consistency coefficient after governance, represents the mean value of the sub-region governance response consistency coefficient, represents the number of sub-region combinations participating in the synergy calculation, represents the governance post-strategy response offset between the hth pair of sub-regions, represents the synergy reference value of the hth pair of sub-regions before governance, represents a small constant to prevent the denominator from being zero, and represents a weight adjustment factor.

[0041] Preferably, step six further includes:

[0042] S62, by presetting a third threshold Q3 in advance, and comparing and analyzing the governance synergy efficiency coefficient ZLXN with the third threshold Q3, a third evaluation result is obtained, including:

[0043] When the governance synergy efficiency coefficient ZLXN is less than or equal to the third threshold Q3, it indicates that the cross-region governance synergy effect is qualified, no adjustment is made, and continuous monitoring is performed;

[0044] When the governance synergy efficiency coefficient ZLXN is greater than the third threshold Q3, it indicates that the cross-region governance synergy effect is unqualified, a third early warning instruction is triggered, and a third strategy is generated: starting the photovoltaic power station level global governance optimization process to unify the multi-level strategy synergy;

[0045] S63, based on the response characteristics of the device component level fluctuation abnormality coefficient BDYC, the sub-region level sub-region governance response consistency coefficient XYYZ, and the cross-region level governance synergy efficiency coefficient ZLXN, the device-sub-region-cross-region-station four-level control link is reconstructed, and the data synergy hub is deployed;

[0046] S64, based on the evolution track of the hth pair of sub-region interaction data and the governance synergistic effect coefficient ZLXN, construct The dynamic correction model adopts the minimum mean square error (MSE) criterion to regress and approximate the residual fitting error of the offset. The sensitivity threshold of the constraint control system is used to adjust the output power, voltage, current and temperature control strategy, and the EMS and PCS are linked to implement dynamic control fine tuning on a 10-second time scale.

[0047] The application provides a multi-level management method for real-time data of new energy station power generation equipment.

[0048] (1) The multi-level management method for real-time data of new energy station power generation equipment, by collecting voltage, current, temperature, power and other multi-dimensional real-time data in the component layer and sub-region layer, and combining AI model to structure the multi-level data and output dynamic management strategy, can effectively identify abnormal fluctuations and operation deviations, improve the timeliness and accuracy of fault discovery, and ensure stable operation of the photovoltaic station system.

[0049] (2) The multi-level management method for real-time data of new energy station power generation equipment, which adopts an AI multi-scale deep learning model to fuse historical management information and current equipment state, realizes intelligent matching between strategy and equipment, and judges the adaptation effect of strategy in different sub-regions through management response consistency coefficient, can implement strategy re-deployment for unqualified areas, and effectively avoid waste of management resources and uneven distribution.

[0050] (3) The multi-level management method for real-time data of new energy station power generation equipment, by calculating the management synergistic effect coefficient in real time, quantitatively evaluating the cross-region response consistency, strategy coupling and complementarity, if the synergistic effect is unqualified, the area redistribution and synergistic strategy re-deployment mechanism is automatically triggered, the synergistic control and intelligent linkage ability of the whole station level is improved.

[0051] (4) The multi-level management method for real-time data of new energy station power generation equipment, by introducing a closed-loop mechanism of continuous abnormality identification, early warning trigger, strategy generation, response evaluation and re-optimization, judging the management effect through three types of threshold, and dynamically updating the strategy logic combined with the AI model, realizing the adaptive evolution of the management strategy, and enhancing the intelligent management ability of the photovoltaic station. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 The application provides a multi-level management method for real-time data of new energy station power generation equipment. DETAILED DESCRIPTION

[0053] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0054] Embodiment 1

[0055] Please refer to Figure 1 The present application provides a multi-level management method for real-time data of a new energy station power generation device, comprising the following steps:

[0056] Step one, obtaining a photovoltaic station electronic distribution map and dividing sub-regions; collecting voltage, current and temperature at the component layer; collecting output power and cooperative behavior data at the sub-region layer;

[0057] Step two, performing abnormal elimination, normalization and sliding window smoothing processing on the collected data voltage, current and temperature;

[0058] Step three, constructing a multi-scale deep learning model, fusing device to station multi-level data and management history information, training to generate an AI dynamic multi-level management model with management perception, structure modeling and strategy adaptation capability, outputting management state vector, feature map and response evaluation result, for assisting intelligent strategy generation and dynamic adaptation;

[0059] Step four, through real-time monitoring of the component running state of the sub-regions of the photovoltaic station, combining the data mean and fluctuation level of the historical running period of the sub-regions, calculating and obtaining the fluctuation abnormality coefficient BDYC, and comparing and analyzing with the first threshold Q1, judging whether the component fluctuation in the sub-region is normal, if not, giving a strategy;

[0060] Step five, through monitoring the response difference of multiple sub-regions under a unified management strategy, calculating and obtaining the sub-region management response consistency coefficient XYYZ, and comparing and analyzing with the second threshold Q2, judging whether the management response consistency of the sub-region is qualified, if not, giving a strategy;

[0061] Step six, through real-time monitoring of the response consistency, strategy synergistic coupling degree and overall complementary effect among multiple sub-regions after management, calculating and obtaining the management synergistic efficiency coefficient ZLXN, and comparing and analyzing with the third threshold Q3, judging whether the cross-region management synergistic effect is qualified, if not, giving a strategy.

[0062] In this embodiment, first, an electronic distribution map of the photovoltaic station to be treated is obtained, and the photovoltaic station is divided into several sub-regions according to factors such as terrain, component layout and historical maintenance strategy. At the component level, the voltage, current and temperature data of each photovoltaic component are obtained by integrating the acquisition device on each photovoltaic component. At the sub-region level, the output power data and the cooperative behavior data related to the treatment strategy of each sub-region are collected by using the edge device. The collected voltage, current and temperature data are subjected to outlier rejection, normalization processing and sliding window smoothing processing to remove spikes, unify dimensions and reduce noise interference, thereby providing stable input for subsequent modeling. A deep learning treatment model based on multi-scale input features is constructed. The model integrates multi-level structural data from the component level, the sub-region level to the station level, and combines historical treatment records, response effects, fault labels and other information for supervised learning training. Finally, a dynamic AI treatment model with treatment perception ability, structure modeling ability and strategy adaptation ability is formed. The model output results include but are not limited to: a vector representing the current treatment state, a response evaluation result of the treatment strategy execution, and the output result can be used to assist the dynamic generation and matching adaptation of intelligent strategies. Based on the real-time running data of the component layer, combined with the running mean and standard deviation of each sub-region in the historical running period, the fluctuation abnormality coefficient BDYC is constructed. The calculation method considers the deviation degree between the current real-time fluctuation amplitude and the historical stable features. The coefficient BDYC is compared with the set first threshold Q1. If BDYC>Q1, it is determined that there is abnormal fluctuation in the components in the sub-region, and the system triggers the strategy regulation suggestion or the response treatment mechanism. Under the condition of unified treatment strategy implementation, the response indicators of each sub-region are statistically analyzed, and the sub-region treatment response consistency coefficient XYYZ is calculated and obtained. The coefficient reflects the consistency degree of the response effect of different sub-regions under the same treatment strategy. XYYZ is compared with the set second threshold Q2. If XYYZ<Q2, it means that there is a significant difference in the response of the treatment strategy in different regions, and the strategy adjustment suggestion is triggered to improve the consistency of the treatment. After the treatment of multiple sub-regions is completed, the treatment coordination efficiency coefficient ZLXN is constructed by real-time monitoring the coupling degree of the strategy response, the matching degree of the response synergy curve and the overall output complementarity and other parameters, which is used to measure the regional coordination level of the entire photovoltaic station after treatment. ZLXN is compared with the set third threshold Q3. If ZLXN<Q3, it means that there is a lack of coordination efficiency in the treatment across regions, and the strategy structure or the coordination path needs to be optimized. Through regional division, multi-dimensional data acquisition, multi-scale AI modeling and three types of response coefficient judgment mechanisms, the embodiment effectively realizes the multi-region intelligent treatment, abnormal response identification and coordination efficiency optimization in the complex treatment scene of the photovoltaic station, and has good adaptability and promotion value.

[0063] Embodiment 2

[0064] This embodiment is an explanation in embodiment 1, please refer to Figure 1 Specifically, step one includes:

[0065] S11, the electronic distribution map of the photovoltaic power station is acquired in advance, and the photovoltaic power station is divided into several sub-regions, and marked on the electronic distribution map;

[0066] S12, real-time monitoring is performed on the equipment of the sub-region, edge computing sensing devices and temperature sensors are deployed on the photovoltaic module to collect original operation data at the module level, including voltage value, current value and temperature value;

[0067] S13, by deploying intelligent data gateway and multi-sub-region data synchronization interface, collecting original data at the sub-region level, including output power and sub-region cooperative interaction original behavior data.

[0068] In this embodiment, first, the structure presetting and multi-source data sensing operation of the photovoltaic power station are performed, specifically including: the electronic distribution map of the target photovoltaic power station is acquired in advance, and the distribution map can be generated by the GIS system or automatically generated by combining the aerial photography of the unmanned aerial vehicle with the construction drawing of the station. According to the component layout density, terrain distribution, power supply structure or historical maintenance unit and other factors, the entire photovoltaic station is divided into several sub-regions, and is marked in the electronic distribution map. Each sub-region is used as a relatively independent governance unit, which is used for subsequent differentiated analysis and cooperative governance modeling. For the photovoltaic modules in each sub-region, edge computing sensing devices with local computing and data collection capabilities are deployed on the modules, combined with integrated temperature sensors, to realize real-time sensing of the operation state of the modules. The collected original operation data; at the sub-region level, by deploying intelligent data gateway with protocol conversion and multi-channel communication capability, and combining multi-sub-region data synchronization interface, the original operation data of the sub-region is collected. Realize the multi-granularity, real-time, multi-dimensional operation data collection and structure modeling of the photovoltaic station from the module layer to the sub-region layer, and provide basic data support for subsequent governance modeling and evaluation.

[0069] Embodiment 3

[0070] This embodiment is an explanation in embodiment 2, please refer to Figure 1 Specifically, step two includes:

[0071] S21, by using IQR outlier detection and Min-Max normalization technology, the original sensing data of voltage, current and temperature are subjected to outlier elimination and normalization processing; by introducing a sliding window filtering algorithm, the voltage, current and temperature data are subjected to time series smoothing processing, short-term fluctuation interference is suppressed, and trend characteristics are retained.

[0072] In this embodiment, for the photovoltaic module level raw operation data collected in step one, the following data cleaning and feature retention processing steps are performed to improve the accuracy and stability of subsequent modeling analysis: using IQR outlier detection method, the voltage value, current value and temperature value collected by the module layer are identified and removed. After completing the abnormality removal, the Min-Max normalization algorithm is used to standardize the remaining normal data, which maps the data of different physical quantities to the interval [0, 1] to eliminate the influence of dimension difference on subsequent model training, further suppress the interference of short-term measurement error and environmental disturbance on the model, and introduce a sliding window filtering algorithm to smooth the time series data of voltage, current and temperature. Set the sliding window size to w, and each time take the current time as the center to extract the data of several time points before and after for weighted average or simple average processing; the smoothed data can effectively retain long-term trend characteristics while eliminating local noise fluctuations, improving the timeliness and robustness of the downstream governance model. The module layer data processed by this step has strong stability, consistency and normalization, which is suitable for feature input of multi-scale deep learning governance model, and helps to improve the model's perception ability of operation trend, mutation and abnormal response.

[0073] Embodiment 4

[0074] This embodiment is an explanation and description in embodiment 3, please refer to Figure 1 , specifically, step three includes:

[0075] S31, by constructing a deep learning model with multi-scale perception ability, the model is based on device layer-sub-region layer-cross-region layer-site layer, four-layer governance structure, by fusing multi-dimensional operation state data, historical governance strategy feedback and response results after governance, a neural network model with governance perception, structure dependent modeling and strategy adaptation evaluation ability is established, through multi-dimensional perception of multi-device component operation data, sub-region governance behavior, regional coordination strategy and global site governance strategy, a training data set with fused structure time series features is constructed; based on the governance history data, a multi-task label set is constructed; and through unified objective function, joint training is performed to improve the comprehensive modeling ability of the model to different level governance features, finally the trained model is used as AI dynamic multi-level governance model, which outputs photovoltaic power station governance state vector, structure dependent feature map and multi-dimensional response evaluation matrix, and outputs the auxiliary generation and dynamic adaptation of intelligent governance strategy.

[0076] In this embodiment, in order to realize the comprehensive perception of the management state of the photovoltaic power station from the component layer to the station layer, the structure relationship modeling and the strategy response evaluation, an AI dynamic multi-level management model with multi-scale perception ability is proposed. According to the management structure of the photovoltaic station, it is divided into four levels: device layer, sub-area layer, cross-area layer and global station layer. On the basis of the multi-level management structure, a deep learning model combining structure dependence and time sequence dynamics is constructed. The model has the ability of cross-layer modeling, structure perception and strategy adaptation evaluation, which significantly improves the understanding depth of the operation state of the photovoltaic station and the intelligent level of the management strategy.

[0077] Embodiment 5

[0078] This embodiment is an explanation and description in embodiment 4. Please refer to Figure 1 , specifically, step four includes:

[0079] S41, by monitoring the component operation state of the sub-area of the photovoltaic station in real time, combining the data mean and fluctuation level of the historical operation period of the sub-area, and after dimensionless processing, the fluctuation abnormal coefficient BDYC is calculated and obtained, the formula is as follows:

[0080]

[0081] In the formula, n represents the number of components participating in monitoring of the sub-area, represents the current period voltage value of the i-th component, represents the historical mean value of the component voltage, represents the standard deviation value of the voltage history, represents the current period current value of the i-th component, represents the historical mean value of the component current, represents the standard deviation value of the current history, represents the difference between the temperature of the i-th component and the adjacent component, represents the reference temperature value of the component environment in the period, represents an extremely small constant to prevent the denominator from being zero, w1, w2 and w3 represent weight coefficients, , and .

[0082] In this embodiment, in order to identify the fluctuation abnormal situation of the operation state of each sub-area in the photovoltaic station in real time, a fluctuation abnormal coefficient BDYC calculation method based on fusion analysis of historical statistical characteristics and current monitoring data is proposed. The coefficient takes voltage, current and temperature as the core indexes, compares the relative deviation degree of the current monitoring value and the historical mean value, and is composed of standard deviation normalization and temperature offset, which improves the rigor of the calculation.

[0083] Embodiment 6​​

[0084] This embodiment is an explanation and illustration in embodiment 5, please refer to Figure 1 , specifically, step four further comprises:

[0085] S42, by presetting the first threshold Q1 in advance, and comparing and analyzing the fluctuation abnormal coefficient BDYC with the first threshold Q1, the first evaluation result is obtained, including:

[0086] When the fluctuation abnormal coefficient BDYC is less than the first threshold Q1, it indicates that the component fluctuation in the sub-region is normal, there is no deviation from the historical stable state, and continuous monitoring is carried out;

[0087] When the fluctuation abnormal coefficient BDYC is greater than or equal to the first threshold Q1, it indicates that the component fluctuation in the sub-region is abnormal, there is deviation from the historical stable state, a first early warning instruction is triggered, and a first strategy is generated: marking the abnormal occurrence time, component number and physical coordinates; Compared with the historical database, the repeated abnormal points are identified, and the repeated abnormal points are subjected to thermal control and voltage adjustment governance measures; Start the photovoltaic field area governance response consistency analysis mechanism.

[0088] In this embodiment, for the sub-regional component operation state fluctuation in the photovoltaic field area, the system performs the steps of consistency judgment of management as follows: through the preset first threshold Q1, the calculated fluctuation abnormal coefficient BDYC in the sub-region is compared and analyzed to obtain a first evaluation result for judging whether there is an abnormal behavior deviating from the historical stable state in the region. If the fluctuation abnormal coefficient BDYC < Q1, it is considered that the component operation state fluctuation in the current sub-region belongs to the normal range, and there is no significant deviation from the historical stable characteristics, and the system maintains the existing monitoring mode without triggering the management action. If the fluctuation abnormal coefficient BDYC ≥ Q1, it is determined that there is an abnormal fluctuation characteristic in the sub-region, which deviates from the historical stable operation state, and the system immediately triggers a first early warning instruction and generates a first management strategy: automatically marking the time point of the abnormal occurrence, the number of the corresponding abnormal component, and the physical coordinates of the component in the photovoltaic field area; calling the historical operation database to compare and analyze the component abnormal information to identify whether it is a repetitive abnormal point: if the component has appeared similar fluctuation abnormality in the past period and the triggering frequency exceeds the set threshold such as 3 times, it is determined as a repetitive abnormal point; for such repetitive abnormal components, the system automatically performs thermal control management measures such as reducing the load, adjusting the power of the cooling fan, and combining the current voltage offset to perform corresponding voltage adjustment operation; starting the photovoltaic field area management response consistency analysis mechanism to evaluate the difference degree of the abnormal response strategy and the previous management behavior, and ensuring the consistency and cooperation of the response direction, parameter adjustment range and strategy execution time in the multi-component and multi-region collaborative management. Through the above steps, the system realizes intelligent judgment and hierarchical response of the component fluctuation state in the sub-region, and provides reliable data support and strategy basis for the effectiveness and consistency of subsequent management measures, as shown in Table 1:

[0089] Table 1 Step four calculation example table

[0090]

[0091] Embodiment 7

[0092] This embodiment is an explanation and description in embodiment 6, please refer to Figure 1 , specifically, step five includes:

[0093] S51, when receiving the first early warning instruction, starting the photovoltaic field area management response consistency analysis mechanism to monitor the response difference of multiple sub-regions under the unified management strategy, combining the fluctuation abnormal coefficient BDYC, and calculating the sub-regional management response consistency coefficient XYYZ after non-dimensional treatment, the formula is as follows:

[0094]

[0095] In the formula, m represents the number of sub-regions participating in management, represents the average output power of the jth sub-region after governance, represents the reference average output power before governance, represents a minimum constant to prevent the denominator from being zero, represents the standard deviation of the fluctuation anomaly coefficient of all devices in the sub-region, and represents the weight coefficient, , , and .

[0096] In this embodiment, when the system receives the first early warning instruction, it automatically enters the photovoltaic field area governance response consistency analysis stage, which is used to evaluate the response consistency degree of different sub-regions under the unified governance strategy, so as to guide the subsequent strategy fine-tuning and local differentiated response.

[0097] Embodiment 8

[0098] This embodiment is an explanation and description in embodiment 7. Please refer to Figure 1 Specifically, step five further comprises:

[0099] S52, by presetting a second threshold Q2 in advance, and comparing and analyzing the sub-region governance response consistency coefficient XYYZ with the second threshold Q2, a second evaluation result is obtained, which comprises:

[0100] When the sub-region governance response consistency coefficient XYYZ is less than or equal to the second threshold Q2, it indicates that the governance response consistency of the sub-region is qualified, the strategy remains unchanged, and continuous monitoring is carried out.

[0101] When the sub-region governance response consistency coefficient XYYZ is greater than the second threshold Q2, it indicates that the governance response consistency of the sub-region is unqualified, a second early warning instruction is triggered, and a second strategy is generated: starting the governance sub-region redistribution and strategy redeployment mechanism.

[0102] In this embodiment, the governance response consistency coefficient XYYZ is further evaluated and analyzed. By comparing with the preset second threshold Q2, the system can accurately determine whether the current sub-regional governance response reaches the expected consistency level. When XYYZ≤Q2, it indicates that the responses of different sub-regions to the unified strategy are small, the governance effect is relatively coordinated, the system keeps the existing strategy unchanged, and only enters the continuous monitoring stage; when XYYZ>Q2, it indicates that the responses of some sub-regions to the governance strategy deviate greatly, and there is a problem of insufficient execution coordination. The system triggers the second warning instruction and enters the strategy optimization process, automatically executes the "governance sub-region redistribution and strategy redeployment mechanism", to realize the fine and adaptive strategy adjustment process of partition governance. For example, in the numerical simulation of the foregoing embodiment, if the second threshold Q2=40 is set, since the actual calculation result XYYZ≈42.66, the system judges that the governance response consistency is unqualified, and then triggers the second warning and executes the subsequent optimization operation, such as adjusting the individualized voltage control strategy for sub-region 3, as shown in Table 2:

[0103] Table 2 Step five calculation example table

[0104]

[0105] Embodiment 9

[0106] This embodiment is an explanation and description in embodiment 8, please refer to Figure 1 , specifically, step six includes:

[0107] S61, when receiving the second warning instruction, start the governance sub-region redistribution and strategy redeployment mechanism, and monitor the response consistency, strategy coupling degree and overall complementary effect among the multiple sub-regions after governance in real time. After dimensionless processing, the governance coordination efficiency coefficient ZLXN is calculated and obtained, and the formula is as follows:

[0108]

[0109] In the formula, represents the standard deviation of the sub-regional governance response consistency coefficient after governance, represents the mean value of the sub-regional governance response consistency coefficient, represents the number of sub-region combinations participating in the coordination calculation, represents the governance response offset between the hth pair of sub-regions, represents the coordination reference value of the hth pair of sub-regions before governance, represents a small constant to prevent the denominator from being zero, and represent weight adjustment factors, , , and .

[0110] In this embodiment, by constructing the governance synergy coefficient ZLXN, the response consistency and synergy adaptation degree of multiple sub-regions in the photovoltaic field area after strategy redeployment are quantitatively evaluated. By introducing the ratio of the standard deviation of the response consistency coefficient of the sub-regions after governance to the mean value, the synergy deviation behavior caused by the difference in equipment type, local meteorological disturbance or mismatched control strategy can be effectively identified. Further combining the response deviation amount between sub-regions after governance with the synergy benchmark value before governance, the adaptation efficiency of the governance strategy after reconstruction can be accurately measured.

[0111] Embodiment 10

[0112] This embodiment is an explanation and description in embodiment 9. Please refer to Figure 1 , specifically, step six further comprises:

[0113] S62, by presetting a third threshold value Q3 in advance, and comparing and analyzing the governance synergy coefficient ZLXN with the third threshold value Q3, a third evaluation result is obtained, which includes:

[0114] When the governance synergy coefficient ZLXN is less than or equal to the third threshold value Q3, it indicates that the cross-regional governance synergy effect is qualified, and no adjustment is made, and continuous monitoring is carried out;

[0115] When the governance synergy coefficient ZLXN is greater than the third threshold value Q3, it indicates that the cross-regional governance synergy effect is unqualified, a third early warning instruction is triggered, and a third strategy is generated: starting the photovoltaic power station level global governance optimization process, so that the multi-level strategy is coordinated and unified;

[0116] S63, based on the response characteristics of the fluctuation abnormal coefficient BDYC of the equipment component level, the sub-regional governance response consistency coefficient XYYZ of the sub-regional level and the governance synergy coefficient ZLXN of the cross-regional level, the equipment-sub-regional-cross-regional-station four-level control link is reconstructed, and the data synergy hub is deployed;

[0117] S64, based on the evolution track of the hth pair of sub-regional interaction data and the governance synergy coefficient ZLXN, a dynamic correction model is constructed, and the least mean square error MSE criterion is used to regress and approximate the fitting residual of the deviation; is used to constrain the sensitivity threshold of the control system when adjusting the output power, voltage, current and temperature control strategy, and through the linkage of EMS and PCS, dynamic control fine tuning is implemented on the time scale of 10 seconds.

[0118] In this embodiment, the third threshold Q3 is introduced to judge the governance synergy efficiency coefficient ZLXN, the application can realize the accurate identification of the cross-regional governance effect of the photovoltaic station, and then trigger the global governance optimization process of the station level. When ZLXN exceeds the set threshold, the system automatically enters the unified stage of multi-level strategy, which not only guarantees the consistency of the synergy between sub-regions, but also avoids the problem of overall efficiency decline caused by regional isolated optimization. On this basis, the application further integrates the characteristics of the three types of indexes of the device level BDYC, the sub-region level XYYZ and the cross-region level ZLXN, and reconstructs the four-level control link of device-sub-region-cross-region-station to form a unified response architecture from local fluctuations to global scheduling. With the aid of the dynamic correction model constructed by the hth pair of sub-region interaction data and the evolution track of ZLXN, the system regresses and fits the strategy deviation trend based on the minimum mean square error (MSE) criterion, and forms a closed-loop error feedback mechanism. In actual deployment, by linking with the energy management system (EMS) and the power conversion system (PCS), this method can accurately fine-tune the output power, voltage, current and key temperature control indicators within a time scale of 10 seconds. For example, in a certain high-altitude grid-connected photovoltaic station, affected by sudden cloud cover, the temperature control strategy of some areas deviates, the system identifies the increase of ZLXN and triggers the optimization process, and completes the fine-tuning of the multi-region control strategy within 10 seconds, significantly improving the cooperative stability of the station under disturbance conditions, as shown in Table 3:

[0119] Table 3 Step six calculation example table

[0120]

[0121] The size of the threshold is set for easy comparison. The size of the threshold depends on the amount of sample data and the base number set by the person skilled in the art for each group of sample data; as long as it does not affect the proportional relationship of the parameters and the quantized values.

[0122] The above formulas are obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value. The coefficients in the formula are set by the person skilled in the art according to the actual situation. The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can make equivalent substitutions or changes within the technical scope disclosed by the present application according to the technical solution and inventive concept of the present application, which should be covered within the protection scope of the present application.

Claims

1. A multi-level management method for real-time data of new energy station power generation equipment, characterized in that, Comprise the following steps: Step one, obtain the electronic distribution map of photovoltaic station and divide the sub-region; Collect voltage, current, temperature at the component layer; Collect output power and cooperative behavior data at the sub-region layer; Step two, the collected data voltage, current and temperature are excluded, normalized and smoothed by sliding window; Step three, construct a multi-scale deep learning model, fuse device to station multi-level data and governance history information, train to generate AI dynamic multi-level governance model with governance perception, structure modeling and strategy adaptation ability, output governance state vector, feature spectrum and response evaluation result, for auxiliary intelligent strategy generation and dynamic adaptation; Step four, through real-time monitoring of the component running state of the sub-region of the photovoltaic station, combining the data mean and fluctuation level of the historical operation period of the sub-region, the fluctuation abnormality coefficient BDYC is calculated and compared with the first threshold Q1 to judge whether the component fluctuation in the sub-region is normal, if not, give strategy; Step five, through monitoring the response difference of multiple sub-regions under unified governance strategy, the sub-region governance response consistency coefficient XYYZ is calculated and compared with the second threshold Q2 to judge whether the governance response consistency of the sub-region is qualified, if not, give strategy; Step six, through real-time monitoring of the response consistency, strategy coupling degree and overall complementary effect among multiple sub-regions after governance, the governance synergy efficiency coefficient ZLXN is calculated and compared with the third threshold Q3 to judge whether the cross-region governance synergy effect is qualified, if not, give strategy; Step four includes: S41, through real-time monitoring of the component running state of the sub-region of the photovoltaic station, combining the data mean and fluctuation level of the historical operation period of the sub-region, after dimensionless processing, the fluctuation abnormality coefficient BDYC is calculated, the formula is as follows: In the formula, n represents the number of components participating in monitoring of the sub-region, represents the current cycle voltage value of the i-th component, represents the historical average value of the component voltage, represents the historical standard deviation value of the voltage, represents the current cycle current value of the i-th component, represents the historical average value of the component current, represents the historical standard deviation value of the current, represents the difference between the temperature of the i-th component and the adjacent component, represents the environmental reference temperature value of the component within the cycle, represents a minimum value constant for preventing the denominator from being zero, w1, w2, and w3 represent weight coefficients; Step five includes: S51, when receiving the first early warning instruction, start the photovoltaic field area governance response consistency analysis mechanism, monitor the response difference of multiple sub-regions under unified governance strategy, combine the fluctuation abnormality coefficient BDYC, after dimensionless processing, calculate the sub-region governance response consistency coefficient XYYZ, the formula is as follows: In the formula, m represents the number of sub-regions participating in management, represents the average output power of the jth sub-region after management, represents the reference average output power before management, represents a minimum constant to prevent the denominator from being zero, represents the standard deviation of the fluctuation anomaly coefficient of all devices in the sub-region, and represents a weight coefficient; Step six includes: S61, when receiving the second early warning instruction, start the governance sub-region redistribution and strategy redeployment mechanism, real-time monitor the response consistency, strategy coupling degree and overall complementary effect among multiple sub-regions after governance, after dimensionless processing, calculate the governance synergy efficiency coefficient ZLXN, the formula is as follows: In the formula, represents the standard deviation of the consistency coefficient of the sub-regional governance response after governance, represents the mean of the consistency coefficient of the sub-regional governance response after governance, represents the number of sub-regional combinations participating in collaborative calculation, represents the governance strategy response deviation between the hth pair of sub-regions after governance, represents the collaborative benchmark value of the hth pair of sub-regions before governance, represents a minimum constant to prevent the denominator from being zero, and represents a weight adjustment factor.

2. The multi-level governance method of real-time data of a new energy station power generation equipment according to claim 1, characterized in that, Step one includes: S11, obtain the electronic distribution map of photovoltaic power station in advance, divide the photovoltaic power station into several sub-regions, and mark on the electronic distribution map; S12, real-time monitor the equipment of the sub-region, collect the original running data of the component layer through deploying edge computing perception device and temperature sensor on photovoltaic component, including voltage value, current value and temperature value; S13, by deploying intelligent data gateway and multi-sub-area data synchronization interface, collecting sub-area level operation raw data, including output power and sub-area cooperative interaction raw behavior data.

3. The multi-level governance method of real-time data of a new energy station power generation equipment according to claim 2, characterized in that, Step two includes: S21, by adopting IQR outlier detection and Min-Max normalization technology, the raw sensor data of voltage, current and temperature are subjected to outlier elimination and normalization processing; by introducing a sliding window filtering algorithm, the voltage, current and temperature data are subjected to time series smoothing processing, short-term fluctuation interference is suppressed, and trend characteristics are retained.

4. The multi-level governance method of real-time data of a new energy station power generation equipment according to claim 3, characterized in that, Step three includes: S31, by constructing a deep learning model with multi-scale perception ability, the model is based on device layer-sub-area layer-cross area layer-station layer, four-layer governance structure, by fusing multi-dimensional operation state data, historical governance strategy feedback and response results after governance, a neural network model with governance perception, structure dependent modeling and strategy adaptation evaluation ability is established, through multi-dimensional perception of multi-device component operation data, sub-area governance behavior, regional cooperative strategy and global station governance strategy, a training data set integrating structure time series characteristics is constructed; a multi-task label set is constructed based on governance history data; and joint training is carried out through a unified objective function, the comprehensive modeling ability of the model for different level governance characteristics is improved, and finally the trained model is used as an AI dynamic multi-level governance model, which outputs photovoltaic power station governance state vector, structure dependent feature map and multi-dimensional response evaluation matrix, and outputs the auxiliary generation and dynamic adaptation of intelligent governance strategy.

5. The multi-level governance method of real-time data of a new energy station power generation equipment according to claim 1, characterized in that, Step four also includes: S42, by presetting a first threshold Q1 in advance, and comparing and analyzing the fluctuation abnormality coefficient BDYC with the first threshold Q1, a first evaluation result is obtained, including: When the fluctuation abnormality coefficient BDYC is less than the first threshold Q1, it indicates that the component fluctuation in the sub-area is normal, and there is no deviation from the historical stable state, and continuous monitoring is carried out; When the fluctuation abnormality coefficient BDYC is greater than or equal to the first threshold Q1, it indicates that the component fluctuation in the sub-area is abnormal, and there is deviation from the historical stable state, a first warning instruction is triggered, and a first strategy is generated: marking the abnormal occurrence time, component number and physical coordinates; comparing with the historical database, identifying repeated abnormal points, and taking heat control and voltage adjustment governance measures for the repeated abnormal points; starting the photovoltaic field area governance response consistency analysis mechanism.

6. The multi-level governance method of real-time data of a new energy station power generation equipment according to claim 1, characterized in that, Step five also includes: S52, by presetting a second threshold Q2 in advance, and comparing and analyzing the sub-area governance response consistency coefficient XYYZ with the second threshold Q2, a second evaluation result is obtained, including: When the sub-area governance response consistency coefficient XYYZ is less than or equal to the second threshold Q2, it indicates that the governance response consistency of the sub-area is qualified, the strategy remains unchanged, and continuous monitoring is carried out; When the sub-area governance response consistency coefficient XYYZ is greater than the second threshold Q2, it indicates that the governance response consistency of the sub-area is unqualified, a second warning instruction is triggered, and a second strategy is generated: starting the governance sub-area redistribution and strategy redeployment mechanism.

7. The multi-level governance method of real-time data of a new energy station power generation equipment according to claim 1, characterized in that, Step six also includes: S62, by presetting a third threshold Q3 in advance, and comparing and analyzing the governance synergy efficiency coefficient ZLXN with the third threshold Q3, a third evaluation result is obtained, including: When the governance synergy effectiveness coefficient ZLXN is less than or equal to the third threshold Q3, it indicates that the cross-region governance synergy effect is qualified, no adjustment is made, and continuous monitoring is performed. When the governance synergy effectiveness coefficient ZLXN is greater than the third threshold Q3, it indicates that the cross-region governance synergy effect is unqualified, a third early warning instruction is triggered, and a third strategy is generated: starting a global governance optimization process at the photovoltaic power station level to make the multi-level strategy synergistic and unified. S63, based on the response characteristics of the device component level fluctuation anomaly coefficient BDYC, the sub-region level sub-region governance response consistency coefficient XYYZ, and the cross-region level governance synergy effectiveness coefficient ZLXN, reconstructing the device-sub-region-cross-region-station four-level control link, and deploying a data collaboration hub. S64, based on the hth pair of sub-region interaction data and the evolution trajectory of the governance synergy coefficient ZLXN, construct The dynamic correction model uses the least mean square error (MSE) criterion to regress and approximate the residual error of the migration fitting. The sensitivity threshold of the constraint control system is used to adjust the output power, voltage, current and temperature control strategy, and the dynamic control fine tuning is implemented on the time scale of 10 seconds through the linkage of EMS and PCS.

Images