An intelligent data processing system for accessing new energy sources to the power grid

Abstract

The present invention discloses an intelligent data processing system for accessing new energy to a power grid, and relates to the technical field of power data processing. The intelligent data processing system for accessing new energy to a power grid includes: a data acquisition and preprocessing unit, a regularity prediction module, an anomaly detection module, a geographic location analysis and processing module, an equipment health management and predictive maintenance unit, and an equipment input power compensation unit; by detecting the data of the access of new energy power generation equipment to the power grid, and analyzing the regularity and abnormal data of its fluctuations, on the one hand, it is convenient to better control the safety of the power grid, and through abnormality analysis, it can be analyzed whether there is abnormal operation of new energy power generation, and its health can be evaluated, so as to facilitate maintenance in advance and avoid its failure rate; the application of the equipment input power compensation unit can complement the power of different power generation equipment connected to the power grid, and avoid the problem of instability of the local power grid caused by fluctuations.

Description

Technical Field

[0001] The present invention relates to the technical field of power data processing, and in particular to an intelligent data processing system for accessing new energy sources to a power grid. Background Art

[0002] The access of new energy power generation equipment to the power grid needs to meet certain conditions, including the performance, safety, stability of the equipment and the technical requirements for grid connection. In addition, the carrying capacity of the power grid needs to be considered to ensure that the access of new energy power generation equipment will not have an adverse impact on the stable operation of the power grid. There are two main access methods for new energy power generation equipment: distributed access and centralized access. Distributed access refers to the decentralized access of new energy power generation equipment to the distribution network, which is suitable for small and decentralized new energy power generation projects. Centralized access refers to the centralized access of new energy power generation equipment to the high-voltage power grid, which is suitable for large and centralized new energy power generation projects.

[0003] Publication No. CN117713209A discloses a hierarchical coordination control system for accessing a new energy source to a power grid, which belongs to the technical field of power grid control, and includes: a power grid data acquisition module, a power grid data processing module, a power grid data analysis module, a hierarchical coordination control module, and a power grid data storage module. The present invention solves the problem that the existing access of new energy sources to the power grid cannot effectively perform hierarchical coordination control on the new energy sources accessed to the power grid, resulting in poor use effect of the access of new energy sources to the power grid. The present invention collects and processes power grid load data, determines power grid load characterization data, analyzes and evaluates the power grid load characterization data based on power grid energy storage standard data, determines the power grid load analysis results, formulates a hierarchical coordination control method based on the power grid load analysis results, and performs hierarchical coordination control on the access of new energy sources to the power grid based on the hierarchical coordination control method, which can effectively perform hierarchical coordination control on the new energy sources accessed to the power grid, and can improve the use effect of the access of new energy sources to the power grid.

[0004] As shown in the above technology, the existing technology only controls the power of renewable energy connected to the power grid, so as to achieve better use effect. It is based on real-time detection and processing of power input by new energy equipment. However, since new energy equipment relies on natural resources to generate electricity, although there are certain natural laws, there is still great instability, which leads to the input power being not stable enough. Even if the current stabilization and voltage stabilization equipment are set for adjustment, the total power generation per unit time still fluctuates, especially when the power generation equipment fails after long-term use, it is more difficult to control. If a power generation equipment suddenly fails to operate normally, it will have a greater impact on the stability of the power grid. If it can be predicted and prepared in advance, many losses will be reduced. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an intelligent data processing system for accessing new energy to a power grid, which solves the above-mentioned problems existing in the existing management and control system for accessing new energy to a power grid.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent data processing system for accessing new energy to a power grid, comprising:

[0007] Data collection and preprocessing unit: deploy sensors on new energy power generation equipment and power grids to collect data in real time; perform data cleaning, data normalization and geographic location tagging preprocessing on the collected data;

[0008] Regularity prediction module: The regularity prediction module uses the long short-term memory network in time series analysis as the core prediction model. By analyzing the historical data of new energy power generation equipment and key components of the power grid, it identifies and predicts the future change trend of the operating parameters of new energy power generation equipment;

[0009] Anomaly detection module: Set anomaly thresholds based on the normal fluctuation range obtained from historical data statistics; use LSTM networks to process time series data to predict future data points; data points that exceed the preset anomaly thresholds are defined as outliers;

[0010] Geographic location analysis and processing module: Use GIS software or online map services to visualize geographic location information, combine anomaly detection results, and mark the location and type of abnormal events on the map;

[0011] Equipment health management and predictive maintenance unit: Use machine learning algorithms to establish and train health assessment models, and judge the health status of new energy power generation equipment by analyzing the anomalies of data input into the power grid by new energy power generation equipment; and set predictive maintenance strategies to perform predictive maintenance on new energy power generation equipment;

[0012] Equipment input power compensation unit: Based on the analysis results of the comprehensive rule prediction module, the anomaly detection module, the geographic location analysis and processing module, and the equipment health management and predictive maintenance unit, it dynamically compensates the input power of new energy power generation equipment connected to the same power grid in the same area to maintain regional power stability.

[0013] Preferably, the data acquisition of the data acquisition and preprocessing unit includes: collecting power data of new energy power generation equipment from the power grid monitoring system, including active power, reactive power, voltage, and current time series data;

[0014] Data preprocessing: clean the data, remove noise, outliers and missing values, and normalize the data to ensure that data of different dimensions are comparable;

[0015] Geographic location tag: Add geographic location information such as longitude, latitude and altitude to each piece of data, or mark specific line and equipment code information.

[0016] Preferably, the work content of the rule prediction module includes:

[0017] Parameter definition:

[0018] Define time series data:

[0019] The time step T represents the time interval for setting the time series data;

[0020] The historical data length L is the number of historical data points used to train the LSTM model, which is adjusted based on data availability and model performance;

[0021] The feature dimension T represents the number of features at each time point, including temperature, current, and voltage;

[0022] Define LSTM model parameters:

[0023] The number of layers F indicates the number of layers in the LSTM network, which affects the complexity and learning ability of the model;

[0024] The number of hidden units H indicates the number of hidden units in each layer of LSTM and determines the capacity of the model;

[0025] The learning rate η represents the learning rate during model training, which affects the training speed and convergence;

[0026] Batch size B, which indicates the number of samples input into the network during each training, affects memory usage and training efficiency;

[0027] The number of training rounds E represents the number of iterations of model training, which is adjusted according to the performance of the validation set;

[0028] Prediction output:

[0029] The prediction step size S represents the number of future time points predicted by the model, that is, the predicted time range;

[0030] Use LSTM network to build the core prediction model:

[0031] The LSTM network consists of an input gate i t 、Forget gate t , output gate o t and cell state, where the cell state is divided into cell state update and the hidden state h t , the expression is as follows:

[0032] Input gate: i t =σ(W i ·[ht-1 , x t ]+b i );

[0033] Forget gate: f t =σ(W f ·[h t-1 , x t ]+b f );

[0034] Cell status update:

[0035]

[0036] Output gate: o t =σ(W o ·[h t-1 , x t ]+b o );

[0037] Hidden state: h t =o t tanh(C t );

[0038] Among them, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W and b are weight and bias parameters, x t is the input feature at the current time point, h t-1 is the hidden state at the previous time point, C t-1 and C t are the cell states at the previous time point and the current time point, respectively;

[0039] The prediction formula is expressed as:

[0040] Given a historical data sequence (x1, x2, ..., x L ), the LSTM model will output the predicted sequence in, The feature value at time point t predicted by the model; the prediction process is achieved by iterating the hidden state and cell state of the LSTM network.

[0041] Preferably, the abnormal contents detected by the abnormality detection module include abnormal fluctuation value detection, identification of fluctuations beyond the regularity and abnormal fluctuation frequency detection:

[0042] Abnormal fluctuation value detection:

[0043] Define abnormal fluctuations: Set a threshold to distinguish normal fluctuations from abnormal fluctuations. The threshold is determined based on the mean, standard deviation, or expert experience of historical data.

[0044] Calculation formula: Abnormal fluctuation value detection is achieved by calculating the Z-score of time series data. The Z-score formula is expressed as:

[0045]

[0046] Where X is the data point, μ is the mean, and σ is the standard deviation;

[0047] Abnormality judgment: When the Z-score of a data point exceeds the set threshold, it is considered that the data point has abnormal fluctuations;

[0048] Identification of irregular fluctuations: Based on historical data, the normal fluctuation range or pattern of the equipment is identified, and the real-time data is compared with the normal fluctuation model. If the data point deviates significantly from the model prediction value, it is considered that there is an irregular fluctuation.

[0049] Abnormal fluctuation frequency detection:

[0050] Calculate frequency: Count the number of abnormal fluctuations in a given time period and calculate the frequency of abnormal fluctuations. The calculation formula is:

[0051]

[0052] Among them, P represents the abnormal fluctuation frequency, N represents the number of abnormal fluctuations, and N 总 Represents the total number of data points.

[0053] Preferably, the parameter design content of the geographic location analysis and processing module includes:

[0054] Map accuracy: set the zoom level and resolution of the map;

[0055] Visual elements: including equipment location markers, status indicators, and abnormal event markers;

[0056] Geographic coordinates: including latitude and longitude coordinates and altitude, calibrating the coordinates of each new energy power generation equipment;

[0057] Grid line classification: according to voltage level, line unit, line type and specific line code.

[0058] Preferably, in the equipment health management and predictive maintenance unit, the establishment of the health assessment model extracts statistical features and time-frequency features from the time series of key parameters of the equipment; the health assessment model is trained using a machine learning algorithm, a feature vector is input, and a health status score or a remaining life prediction value is output;

[0059] Development of a predictive maintenance strategy includes:

[0060] Maintenance threshold setting: Based on the output of the health assessment model, set the threshold for triggering maintenance operations;

[0061] Maintenance plan formulation: Combine geographic location information to optimize the scheduling and route planning of maintenance personnel and formulate maintenance plans.

[0062] Preferably, the input feature vector includes four input features:

[0063] The abnormal fluctuation detection index, i.e., the number of times the Z-score exceeds the threshold, is denoted as A;

[0064] The standard deviation of the residuals beyond the regular fluctuation index, that is, the deviation from the normal ARIMA model, is denoted as B;

[0065] The abnormal fluctuation frequency index, that is, the number of abnormal fluctuations per unit time, is denoted by C;

[0066] Other relevant characteristics, denoted as D, include equipment operating time, ambient temperature, wind speed / light intensity, which are specifically determined according to the type of new energy;

[0067] Apply the trained model to real-time data, input feature vector X real , the model outputs the health status score or grade y of the device pred , according to the score or grade, the threshold is set to determine whether the equipment is in a normal, minor fault or serious fault state.

[0068] Preferably, the health assessment model is constructed by selecting a support vector machine machine learning algorithm to construct the model, and the input feature vector is X=[A, B, C, D];

[0069] Model training: Using feature vector X from historical data train and the corresponding health status label y train Train the model;

[0070] Use support vector machine SVM to optimize the objective function:

[0071]

[0072] Where C is the regularization parameter, ξ i is the slack variable; w is the normal vector of the hyperplane, which determines the direction of the hyperplane; b is the displacement term of the hyperplane, which determines the distance between the hyperplane and the origin; ξ i is a slack variable used to deal with noise or indistinguishability in data points; C is a regularization parameter used to control the complexity of the model and the degree of penalty for misclassification; m is the number of data points;

[0073] The optimization goal of SVM is to find a set of parameters w, b and ξ i , so that the objective function reaches the minimum value.

[0074] Preferably, the design parameters of the device input electrical compensation unit include the following:

[0075] Enter the data collection parameters:

[0076] Data collection frequency: set the time interval for data collection to ensure the real-time and accuracy of data;

[0077] Data accuracy: Set the accuracy of data collection to meet the system's needs for data processing and analysis;

[0078] Geographic location information accuracy: Set the accuracy of geographic location information to ensure the accuracy of device location;

[0079] Enter electroanalysis parameters:

[0080] Rated input power setting: set the rated input power value for each renewable energy power generation equipment as a basis for judging whether the input power is excessive or insufficient;

[0081] Input power monitoring threshold: Set the threshold for input power monitoring. When the actual input power exceeds or falls below the input power monitoring threshold, input power analysis is triggered.

[0082] Dynamic compensation adjusts model parameters:

[0083] Compensation strategy priority: set the priority of different compensation strategies;

[0084] Compensation efficiency coefficient: Set the efficiency coefficient during the compensation process to reflect the losses during power transmission, storage, and conversion;

[0085] Model iteration optimization cycle: Set the model iteration optimization cycle to ensure that the model can adapt to changes in power grid operation;

[0086] Assume that the actual input power of a new energy power generation equipment is P actual , rated input power is P rated , then the calculation formula for excess or insufficient input power is:

[0087] ΔP=P actual -P rated ;

[0088] Among them, ΔP represents the excess or shortage of input electricity. When ΔP>0, it means excess, and when ΔP<0, it means shortage;

[0089] Regional input electrical balance analysis:

[0090] Assume that the number of power generation equipment connected to the same power grid in the same area is M, and the excess or shortage of input power of each equipment is ΔP j , and j∈1, 2, ..., M, then the analytical formula for the regional input electrical balance is:

[0091]

[0092] In addition to analyzing the actual collected data to compensate for each new energy power generation equipment, the equipment input power compensation unit also inputs the values ​​predicted by the regularity prediction module to predict the values ​​that need to be compensated within a specified time period in the future.

[0093] Preferably, the analytical formula of the electric balance indicates that in the same area, the excess and shortage of input electricity of all power generation equipment should offset each other to achieve a balanced state; however, in actual situations, due to the influence of various factors, it is impossible to achieve a complete balance, and at this time, it is necessary to optimize through a dynamic compensation adjustment model;

[0094] Compensation strategy formulation: When formulating the compensation strategy, an optimization model is constructed based on the surplus and shortage of input electricity, the geographical location of the equipment, and power demand factors; the objective function of the optimization model is set to minimize power transmission losses or maximize power utilization efficiency;

[0095] Let x j represents the compensation amount of the jth device, and the optimization model is expressed as:

[0096]

[0097] Among them, c j Represents the unit cost of compensation for the jth device; the goal of the optimization model is to minimize the total cost of compensation while satisfying the regional input power balance; an iterative algorithm is used to solve it. In each iteration, the model parameters are adjusted and the compensation amount is recalculated according to the current execution of the compensation strategy; through multiple iterations, the optimal solution is gradually approached.

[0098] The present invention provides an intelligent data processing system for accessing new energy sources to a power grid. Compared with the prior art, it has the following beneficial effects:

[0099] 1. The regularity prediction and anomaly detection modules of the intelligent data processing system for new energy access to the power grid can accurately predict the operating status of the equipment and detect anomalies in a timely manner. By detecting the data of the new energy power generation equipment accessing the power grid and analyzing its fluctuation regularity and abnormal data, it is convenient to better control the security of the power grid. Through anomaly analysis, it can be analyzed whether there is abnormal operation of new energy power generation and its health level can be evaluated, which is convenient for early maintenance and avoids its failure rate. The geographic location analysis and processing module realizes the visualization of data, which is convenient for operation and maintenance personnel to quickly locate and handle abnormal events. The application of the equipment input power compensation unit can complement the power of different power generation equipment connected to the power grid to avoid the problem of instability of local power grid caused by fluctuations. The cooperation of multiple modules effectively improves the stability and safety of power grid operation.

[0100] 2. This module is used for the intelligent data processing system for the access of new energy to the power grid. By finely defining the time series data and LSTM model parameters, this module can more accurately capture the historical change trends of the operating parameters of new energy power generation equipment and provide strong support for model training. As the core prediction model, the LSTM network, with its unique cell state and gating mechanism, enables the model to memorize the key information in the time series for a long time, effectively solving the shortcomings of traditional time series analysis methods in long-term prediction. In addition, this module achieves accurate prediction of future time points by iterating the hidden state and cell state of the LSTM network, and the setting of the prediction step enables the model to flexibly respond to prediction needs in different time ranges. This not only improves the accuracy and reliability of the prediction, but also provides a scientific basis for the operation and management of power grid dispatching and new energy power generation equipment.

[0101] 3. The intelligent data processing system for new energy access to the power grid realizes comprehensive and accurate identification of abnormal operating parameters of new energy power generation equipment through comprehensive abnormal fluctuation value detection, identification of fluctuations beyond the rules and abnormal fluctuation frequency detection. Abnormal fluctuation value detection effectively distinguishes normal fluctuations from abnormal fluctuations by setting thresholds and calculating Z-scores, thereby improving the accuracy of detection. Identification of fluctuations beyond the rules establishes a normal fluctuation model based on historical data, and compares and identifies significantly deviated data points in real time, further enhancing the sensitivity of detection. At the same time, abnormal fluctuation frequency detection provides an important basis for evaluating the operating status of equipment and formulating maintenance strategies by counting the number of abnormal fluctuations and calculating the frequency.

[0102] 4. This intelligent data processing system for new energy access to the power grid extracts statistical features and time-frequency features, combines support vector machine machine learning algorithms, and constructs a high-precision health assessment model, which can accurately assess the health status of new energy power generation equipment, predict the remaining life, and provide a scientific basis for predictive maintenance. The formulation of predictive maintenance strategies combines geographic location information, optimizes the scheduling and path planning of maintenance personnel, and improves maintenance efficiency. In addition, the input feature vectors such as abnormal fluctuation detection indicators, out-of-regular fluctuation indicators, and abnormal fluctuation frequency indicators introduced by the unit fully reflect the operating status of the equipment and improve the accuracy and reliability of health assessment. By applying the trained model in real time, hidden dangers of equipment failure can be discovered in time, the expansion of failures can be avoided, and downtime and maintenance costs can be reduced.

[0103] 5. This intelligent data processing system for new energy access to the power grid realizes accurate dynamic compensation for the input electricity of new energy power generation equipment by integrating the analysis results of multiple modules, and effectively maintains the stability of regional power. The design parameters include data collection frequency, data accuracy, geographic location information accuracy, etc., which ensure the real-time and accuracy of the data and the accuracy of the location information. By setting parameters such as rated input electricity and monitoring thresholds, and building a dynamic compensation adjustment model, the unit can flexibly select compensation strategies according to actual conditions, optimize power transmission, storage and conversion processes, and reduce losses. In addition, the unit also introduces regional input power balance analysis. By building an optimization model and using an iterative algorithm to solve it, the goal of minimizing compensation costs while meeting power balance is achieved. This not only improves the efficiency of power utilization, but also reduces operation and maintenance costs, providing a strong guarantee for the stable operation of new energy power generation equipment and the safety and reliability of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] Figure 1 is a system module block diagram of the present invention;

[0105] Figure 2 It is a schematic diagram of the main process of the present invention. DETAILED DESCRIPTION

[0106] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0107] See also Figure 1-Figure 2 , the present invention provides the following five technical solutions:

[0108] The first implementation mode: an intelligent data processing system for accessing new energy to a power grid, comprising:

[0109] Data acquisition and preprocessing unit: deploy sensors on renewable energy power generation equipment and power grids to collect data in real time; perform data cleaning, data normalization and geographic location tagging preprocessing on the collected data; data acquisition includes: collecting power data of renewable energy power generation equipment from the power grid monitoring system, including active power, reactive power, voltage, and current time series data; data preprocessing: clean the data, remove noise, outliers (such as extreme values ​​caused by sensor failures) and missing values, and normalize the data to ensure that data of different dimensions are comparable; geographic location tagging: add geographic location information of latitude, longitude, and altitude to each piece of data, or mark specific line and equipment coding information;

[0110] Regularity prediction module: The regularity prediction module uses the long short-term memory network in time series analysis as the core prediction model. By analyzing the historical data of new energy power generation equipment and key components of the power grid, it identifies and predicts the future change trend of the operating parameters of new energy power generation equipment;

[0111] Anomaly detection module: Set anomaly thresholds based on the normal fluctuation range obtained from historical data statistics; use LSTM networks to process time series data to predict future data points; data points that exceed the preset anomaly thresholds are defined as outliers;

[0112] Geographic location analysis and processing module: Use GIS software or online map services to visualize geographic location information, combine anomaly detection results, and mark the location and type of abnormal events on the map; parameter design content includes: map accuracy: set the zoom level and resolution of the map; visualization elements: including equipment location markers, status indicators (such as color coding to indicate health status), and abnormal event markers; geographic coordinates: including latitude and longitude coordinates and altitude, calibrate the coordinates of each new energy power generation equipment; grid line classification: according to voltage level, line unit, line type, and specific line code;

[0113] As shown in Table 1 below, the geographical coordinates of the equipment and lines are marked and classified in the following way:

[0114]

[0115]

[0116] Equipment health management and predictive maintenance unit: Utilizes data collected through data collection and preprocessing, and uses machine learning algorithms to establish and train health assessment models to determine the health status of new energy power generation equipment by analyzing the anomalies of power data input into the power grid from new energy power generation equipment; and sets predictive maintenance strategies to perform predictive maintenance on new energy power generation equipment;

[0117] Equipment input power compensation unit: Based on the analysis results of the comprehensive rule prediction module, the anomaly detection module, the geographic location analysis and processing module, and the equipment health management and predictive maintenance unit, it dynamically compensates the input power of new energy power generation equipment connected to the same power grid in the same area to maintain regional power stability.

[0118] The regularity prediction and anomaly detection module can accurately predict the operating status of the equipment and detect anomalies in a timely manner. By detecting the data of the new energy power generation equipment connected to the power grid and analyzing its fluctuation patterns and abnormal data, it is convenient to better control the security of the power grid. Through abnormal analysis, it can analyze whether there is abnormal operation of new energy power generation and evaluate its health level, which is convenient for early maintenance to avoid its failure rate. The geographic location analysis and processing module realizes the visualization of data, which is convenient for operation and maintenance personnel to quickly locate and handle abnormal events. The application of the equipment input power compensation unit can complement the power of different power generation equipment connected to the power grid to avoid the problem of instability of local power grids caused by fluctuations. The cooperation of multiple modules effectively improves the stability and safety of power grid operation.

[0119] The second implementation mode is mainly different from the first implementation mode in that the work contents of the rule prediction module include:

[0120] Parameter definition:

[0121] Define time series data:

[0122] The time step T represents the time interval for setting the time series data, such as every hour, every day, etc.;

[0123] The historical data length L is the number of historical data points used to train the LSTM model, which is adjusted based on data availability and model performance;

[0124] The feature dimension T represents the number of features at each time point, including temperature, current, and voltage;

[0125] Define LSTM model parameters:

[0126] The number of layers F indicates the number of layers in the LSTM network, which affects the complexity and learning ability of the model;

[0127] The number of hidden units H indicates the number of hidden units in each layer of LSTM and determines the capacity of the model;

[0128] The learning rate η represents the learning rate during model training, which affects the training speed and convergence;

[0129] Batch size B, which indicates the number of samples input into the network during each training, affects memory usage and training efficiency;

[0130] The number of training rounds E represents the number of iterations of model training, which is adjusted according to the performance of the validation set;

[0131] Prediction output:

[0132] The prediction step size S represents the number of future time points predicted by the model, that is, the predicted time range;

[0133] Use LSTM network to build the core prediction model:

[0134] The LSTM network consists of an input gate i t 、Forget gate t , output gate o t and cell state, where the cell state is divided into cell state update and the hidden state h t , the expression is as follows:

[0135] Input gate: i t =σ(W i ·[h t-1 , x t ]+b i );

[0136] Forget gate: f t =σ(W f ·[h t-1 , x t ]+b f );

[0137] Cell status update:

[0138]

[0139] Output gate: o t =σ(W o ·[h t-1 , x t ]+b o );

[0140] Hidden state: h t =o t tanh(C t );

[0141] Among them, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W and b are weight and bias parameters, x t is the input feature at the current time point, h t-1 is the hidden state at the previous time point, C t-1 and C t are the cell states at the previous time point and the current time point, respectively;

[0142] The prediction formula is expressed as:

[0143] Given a historical data sequence (x1, x2, ..., x L ), the LSTM model will output the predicted sequence in, The feature value at time point t predicted by the model; the prediction process is achieved by iterating the hidden state and cell state of the LSTM network.

[0144] By finely defining time series data and LSTM model parameters, this module can more accurately capture the historical change trends of the operating parameters of new energy power generation equipment and provide strong support for model training. As the core prediction model, the LSTM network's unique cell state and gating mechanism enable the model to memorize key information in the time series for a long time, effectively solving the shortcomings of traditional time series analysis methods in long-term prediction. In addition, this module achieves accurate prediction of future time points by iterating the hidden state and cell state of the LSTM network. The setting of the prediction step enables the model to flexibly respond to prediction needs in different time ranges. This not only improves the accuracy and reliability of the prediction, but also provides a scientific basis for power grid dispatching and the operation and management of new energy power generation equipment.

[0145] The third implementation mode is mainly different from the first implementation mode in that: the abnormal contents detected by the abnormality detection module include abnormal fluctuation value detection, identification of fluctuations beyond the rules and abnormal fluctuation frequency detection:

[0146] Abnormal fluctuation value detection:

[0147] Define abnormal fluctuations: Set a threshold to distinguish normal fluctuations from abnormal fluctuations. The threshold is determined based on the mean, standard deviation, or expert experience of historical data.

[0148] Calculation formula: Abnormal fluctuation value detection is achieved by calculating the Z-score of time series data. The Z-score formula is expressed as:

[0149]

[0150] Where X is the data point, μ is the mean, and σ is the standard deviation;

[0151] Abnormality judgment: When the Z-score of a data point exceeds the set threshold, it is considered that the data point has abnormal fluctuations;

[0152] Identification of irregular fluctuations: Based on historical data, the normal fluctuation range or pattern of the equipment is identified, and the real-time data is compared with the normal fluctuation model. If the data point deviates significantly from the model prediction value, it is considered that there is an irregular fluctuation.

[0153] Abnormal fluctuation frequency detection:

[0154] Calculate frequency: Count the number of abnormal fluctuations in a given time period and calculate the frequency of abnormal fluctuations. The calculation formula is:

[0155]

[0156] Among them, P represents the abnormal fluctuation frequency, N represents the number of abnormal fluctuations, and N 总 Represents the total number of data points.

[0157] Through comprehensive abnormal fluctuation value detection, identification of fluctuations beyond the rules, and abnormal fluctuation frequency detection, comprehensive and accurate identification of abnormal operating parameters of new energy power generation equipment is achieved. Abnormal fluctuation value detection effectively distinguishes normal fluctuations from abnormal fluctuations by setting thresholds and calculating Z-scores, thereby improving the accuracy of detection. Identification of fluctuations beyond the rules establishes a normal fluctuation model based on historical data, and compares and identifies significantly deviated data points in real time, further enhancing the sensitivity of detection. At the same time, abnormal fluctuation frequency detection provides an important basis for evaluating the operating status of equipment and formulating maintenance strategies by counting the number of abnormal fluctuations and calculating the frequency.

[0158] The fourth implementation mode is mainly different from the third implementation mode in that: in the equipment health management and predictive maintenance unit, the establishment of the health assessment model extracts statistical features and time-frequency features from the time series of key parameters of the equipment; the health assessment model is trained by using a machine learning algorithm, inputs a feature vector, and outputs a health status score or a remaining life prediction value;

[0159] Development of a predictive maintenance strategy includes:

[0160] Maintenance threshold setting: Based on the output of the health assessment model, set the threshold for triggering maintenance operations;

[0161] Maintenance plan formulation: Combine geographic location information to optimize the scheduling and route planning of maintenance personnel and formulate maintenance plans.

[0162] The input feature vector includes four input features:

[0163] The abnormal fluctuation detection index, i.e., the number of times the Z-score exceeds the threshold, is denoted as A;

[0164] The standard deviation of the residuals beyond the regular fluctuation index, that is, the deviation from the normal ARIMA model, is denoted as B;

[0165] The abnormal fluctuation frequency index, that is, the number of abnormal fluctuations per unit time, is denoted by C;

[0166] Other relevant characteristics, denoted as D, include equipment operating time, ambient temperature, wind speed / light intensity, which are specifically determined according to the type of new energy;

[0167] Apply the trained model to real-time data, input feature vector X real , the model outputs the health status score or grade y of the device pred , according to the score or grade, the threshold is set to determine whether the equipment is in a normal, minor fault or serious fault state.

[0168] The health assessment model is constructed by selecting a support vector machine machine learning algorithm to construct the model, and the input feature vector is X=[A, B, C, D];

[0169] Model training: Using feature vector X from historical data train and the corresponding health status label y train Train the model;

[0170] Use support vector machine SVM to optimize the objective function:

[0171]

[0172] Where C is the regularization parameter, ξ i is the slack variable; w is the normal vector of the hyperplane, which determines the direction of the hyperplane; in SVM, our goal is to find a hyperplane that can separate data points of different categories as well as possible; the equation of the hyperplane can be expressed as wTx+b=0, where x is the input feature vector; b is the displacement term of the hyperplane, which determines the distance between the hyperplane and the origin; in SVM, b and w together define the position of the hyperplane; ξ i is a slack variable used to deal with noise or indistinguishability in data points; in hard margin SVM, all data points must be correctly classified and outside the interval; however, in practice, this may not be possible due to noise or overlap of data points; therefore, a slack variable ξ is introduced i , allowing some data points to be within the interval or even misclassified, but at a certain cost (i.e. ξ iThe value of will be added to the objective function); C is the regularization parameter, which is used to control the complexity of the model and the degree of penalty for misclassification; the larger the value of C, the heavier the penalty for misclassification, and the model tends to find a more complex model to classify all data points as correctly as possible; on the contrary, the smaller the value of C, the lighter the penalty for misclassification, and the model tends to find a simpler model, even if it means that some data points will be misclassified; m is the number of data points; in SVM, it is necessary to calculate the distance from each data point to the hyperplane, and determine the slack variable ξ based on these distances i The value of

[0173] The optimization goal of SVM is to find a set of parameters w, b and ξ i , so that the objective function reaches the minimum value.

[0174] By extracting statistical features and time-frequency features and combining them with the support vector machine machine learning algorithm, a high-precision health assessment model was constructed, which can accurately assess the health status of new energy power generation equipment and predict the remaining life, providing a scientific basis for predictive maintenance. The formulation of predictive maintenance strategies combines geographic location information, optimizes the scheduling and path planning of maintenance personnel, and improves maintenance efficiency. In addition, the input feature vectors such as abnormal fluctuation detection indicators, out-of-regular fluctuation indicators, and abnormal fluctuation frequency indicators introduced by the unit fully reflect the operating status of the equipment and improve the accuracy and reliability of health assessment. By applying the trained model in real time, hidden dangers of equipment failure can be discovered in a timely manner, the expansion of failures can be avoided, and downtime and maintenance costs can be reduced.

[0175] The fifth implementation mode is mainly different from the first implementation mode in that the design parameters of the device input electrical compensation unit include the following:

[0176] Enter the data collection parameters:

[0177] Data collection frequency: set the time interval for data collection, such as every minute, every hour, etc., to ensure the real-time and accuracy of the data;

[0178] Data accuracy: Set the accuracy of data collection, such as voltage and current accuracy requirements, to meet the system's needs for data processing and analysis;

[0179] Geographic location information accuracy: Set the accuracy of geographic location information, such as latitude, longitude, and altitude, to ensure the accuracy of the device location;

[0180] Enter electroanalysis parameters:

[0181] Rated input power setting: set the rated input power value for each renewable energy power generation equipment as a basis for judging whether the input power is excessive or insufficient;

[0182] Input power monitoring threshold: Set the threshold for input power monitoring. When the actual input power exceeds or falls below the input power monitoring threshold, input power analysis is triggered.

[0183] Dynamic compensation adjusts model parameters:

[0184] Compensation strategy priority: Setting the priority of different compensation strategies, such as direct power transmission, power storage and release, power conversion, etc., to ensure that the best strategy can be selected when multiple compensation strategies are available;

[0185] Compensation efficiency coefficient: Set the efficiency coefficient during the compensation process to reflect the losses during power transmission, storage, and conversion;

[0186] Model iteration and optimization cycle: set the model iteration and optimization cycle, such as daily, weekly, etc., to ensure that the model can adapt to changes in power grid operation;

[0187] Assume that the actual input power of a new energy power generation equipment is P actual , rated input power is P rated , then the calculation formula for excess or insufficient input power is:

[0188] ΔP=P actual -P rated ;

[0189] Among them, ΔP represents the excess or shortage of input electricity. When ΔP>0, it means excess, and when ΔP<0, it means shortage;

[0190] Regional input electrical balance analysis:

[0191] Assume that the number of power generation equipment connected to the same power grid in the same area is M, and the excess or shortage of input power of each equipment is ΔP j , and j∈1, 2, ..., M, then the analytical formula for the regional input electrical balance is:

[0192]

[0193] In addition to analyzing the actual collected data to compensate for each new energy power generation equipment, the equipment input power compensation unit also inputs the values ​​predicted by the regularity prediction module to predict the values ​​that need to be compensated within a specified time period in the future.

[0194] The analytical formula of power balance indicates that in the same area, the excess and shortage of input power of all power generation equipment should offset each other to achieve a balanced state; however, in actual situations, due to the influence of various factors, it is impossible to achieve a complete balance. In this case, it is necessary to optimize through a dynamic compensation adjustment model;

[0195] Compensation strategy formulation: When formulating the compensation strategy, an optimization model is constructed based on the surplus and shortage of input electricity, the geographical location of the equipment, and power demand factors; the objective function of the optimization model is set to minimize power transmission losses or maximize power utilization efficiency;

[0196] Let x j represents the compensation amount of the jth device (it can be positive for output compensation or negative for acceptance compensation), then the optimization model is expressed as:

[0197]

[0198] Among them, c j represents the unit cost of compensation for the jth device (including power transmission loss, equipment maintenance costs, etc.); the goal of the optimization model is to minimize the total cost of compensation while satisfying the regional input power balance; an iterative algorithm is used to solve the problem. In each iteration, the model parameters (such as compensation efficiency coefficient, priority, etc.) are adjusted according to the current execution of the compensation strategy, and the compensation amount is recalculated; through multiple iterations, the optimal solution is gradually approached.

[0199] By integrating the analysis results of multiple modules, accurate dynamic compensation for the input electricity of new energy power generation equipment is achieved, effectively maintaining the stability of regional power. The design parameters include data collection frequency, data accuracy, geographic location information accuracy, etc., to ensure the real-time and accuracy of the data and the accuracy of the location information. By setting parameters such as rated input electricity and monitoring thresholds, and building a dynamic compensation adjustment model, the unit can flexibly select compensation strategies according to actual conditions, optimize power transmission, storage and conversion processes, and reduce losses. In addition, the unit also introduces regional input power balance analysis. By building an optimization model and using an iterative algorithm to solve it, the goal of minimizing compensation costs while meeting power balance is achieved. This not only improves the efficiency of power utilization, but also reduces operation and maintenance costs, providing a strong guarantee for the stable operation of new energy power generation equipment and the safety and reliability of the power grid.

[0200] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0201] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0202] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent data processing system for accessing new energy to a power grid, characterized in that: include: Data acquisition and preprocessing unit: Deploy sensors on new energy power generation equipment and power grids to collect data in real time; Pre-process the collected data by cleaning, normalizing and marking the geographic location; Regularity prediction module: The regularity prediction module uses the long short-term memory network in time series analysis as the core prediction model. By analyzing the historical data of new energy power generation equipment and key components of the power grid, it identifies and predicts the future change trend of the operating parameters of new energy power generation equipment; Anomaly detection module: sets anomaly thresholds based on the normal fluctuation range obtained from historical data statistics; And use LSTM network to process time series data to predict future data points; Data points exceeding the preset anomaly threshold are defined as outliers; Geographic location analysis and processing module: Use GIS software or online map services to visualize geographic location information, combine anomaly detection results, and mark the location and type of abnormal events on the map; Equipment health management and predictive maintenance unit: Use machine learning algorithms to establish and train health assessment models, and judge the health status of new energy power generation equipment by analyzing the anomalies of data input into the power grid by new energy power generation equipment; and set predictive maintenance strategies to perform predictive maintenance on new energy power generation equipment; Equipment input power compensation unit: Based on the analysis results of the comprehensive rule prediction module, the anomaly detection module, the geographic location analysis and processing module, and the equipment health management and predictive maintenance unit, it dynamically compensates the input power of new energy power generation equipment connected to the same power grid in the same area to maintain regional power stability.

2. According to claim 1, an intelligent data processing system for accessing new energy to a power grid is characterized by: The data acquisition of the data acquisition and preprocessing unit includes: collecting power data of new energy power generation equipment from the power grid monitoring system, including active power, reactive power, voltage, and current time series data; Data preprocessing: clean the data, remove noise, outliers and missing values, and normalize the data to ensure that data of different dimensions are comparable; Geographic location tag: Add geographic location information such as longitude, latitude and altitude to each piece of data, or mark specific line and equipment code information.

3. According to claim 1, the intelligent data processing system for accessing new energy to the power grid is characterized by: The work content of the rule prediction module includes: Parameter definition: Define time series data: The time step T represents the time interval for setting the time series data; The historical data length L is the number of historical data points used to train the LSTM model, which is adjusted based on data availability and model performance; The feature dimension T represents the number of features at each time point, including temperature, current, and voltage; Define LSTM model parameters: The number of layers F indicates the number of layers in the LSTM network, which affects the complexity and learning ability of the model; The number of hidden units H indicates the number of hidden units in each layer of LSTM and determines the capacity of the model; The learning rate η represents the learning rate during model training, which affects the training speed and convergence; Batch size B, which indicates the number of samples input into the network during each training, affects memory usage and training efficiency; The number of training rounds E represents the number of iterations of model training, which is adjusted according to the performance of the validation set; Prediction output: The prediction step size S represents the number of future time points predicted by the model, that is, the predicted time range; Use LSTM network to build the core prediction model: The LSTM network consists of an input gate i t 、Forget gate t , output gate o t and cell state, where the cell state is divided into cell state update and the hidden state h t , the expression is as follows: Input gate: i t =σ(W i ·[h t-1 , x t ]+b i ); Forget gate: f t =σ(W f ·[h t-1 , x t ]+b f ); Cell status update: Output gate: o t =σ(W o ·[h t-1 , x t ]+b o ); Hidden state: h t =o t tanh(C t ); Among them, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W and b are weight and bias parameters, x t is the input feature at the current time point, h t-1 is the hidden state at the previous time point, C t-1 and C t are the cell states at the previous time point and the current time point, respectively; The prediction formula is expressed as: Given a historical data sequence (x1, x2, ..., x L ), the LSTM model will output the predicted sequence in, The feature value at time point t predicted by the model; the prediction process is achieved by iterating the hidden state and cell state of the LSTM network.

4. The intelligent data processing system for accessing new energy to a power grid according to claim 1 is characterized in that: The abnormal contents detected by the abnormal detection module include abnormal fluctuation value detection, identification of fluctuations beyond the rules and abnormal fluctuation frequency detection: Abnormal fluctuation value detection: Define abnormal fluctuations: Set a threshold to distinguish normal fluctuations from abnormal fluctuations. The threshold is determined based on the mean, standard deviation, or expert experience of historical data. Calculation formula: Abnormal fluctuation value detection is achieved by calculating the Z-score of time series data. The Z-score formula is expressed as: Where X is the data point, μ is the mean, and σ is the standard deviation; Abnormality judgment: When the Z-score of a data point exceeds the set threshold, it is considered that the data point has abnormal fluctuations; Identification of irregular fluctuations: Based on historical data, the normal fluctuation range or pattern of the equipment is identified, and the real-time data is compared with the normal fluctuation model. If the data point deviates significantly from the model prediction value, it is considered that there is an irregular fluctuation. Abnormal fluctuation frequency detection: Calculate frequency: Count the number of abnormal fluctuations in a given time period and calculate the frequency of abnormal fluctuations. The calculation formula is: Among them, P represents the abnormal fluctuation frequency, N represents the number of abnormal fluctuations, and N 总 Represents the total number of data points.

5. The intelligent data processing system for accessing new energy to a power grid according to claim 1 is characterized in that: The parameter design content of the geographic location analysis and processing module includes: Map accuracy: set the zoom level and resolution of the map; Visual elements: including equipment location markers, status indicators, and abnormal event markers; Geographic coordinates: including latitude and longitude coordinates and altitude, calibrating the coordinates of each new energy power generation equipment; Grid line classification: according to voltage level, line unit, line type and specific line code.

6. The intelligent data processing system for accessing new energy to a power grid according to claim 1 is characterized by: In the equipment health management and predictive maintenance unit, the establishment of a health assessment model extracts statistical features and time-frequency features from the time series of key parameters of the equipment; Use machine learning algorithms to train health assessment models, input feature vectors, and output health status scores or remaining life expectancy prediction values; Development of a predictive maintenance strategy includes: Maintenance threshold setting: Based on the output of the health assessment model, set the threshold for triggering maintenance operations; Maintenance plan formulation: Combine geographic location information to optimize the scheduling and route planning of maintenance personnel and formulate maintenance plans.

7. The intelligent data processing system for accessing new energy to a power grid according to claim 6 is characterized by: The input feature vector includes four input features: The abnormal fluctuation detection index, i.e., the number of times the Z-score exceeds the threshold, is denoted as A; The standard deviation of the residuals beyond the regular fluctuation index, that is, the deviation from the normal ARIMA model, is denoted as B; The abnormal fluctuation frequency index, that is, the number of abnormal fluctuations per unit time, is denoted by C; Other relevant characteristics, denoted as D, include equipment operating time, ambient temperature, wind speed / light intensity, which are determined according to the type of new energy; Apply the trained model to real-time data, input feature vector X real , the model outputs the health status score or grade y of the device pred , according to the score or grade, the threshold is set to determine whether the equipment is in a normal, minor fault or serious fault state.

8. The intelligent data processing system for accessing new energy to a power grid according to claim 7 is characterized by: The health assessment model is constructed by selecting a support vector machine machine learning algorithm to construct the model, and the input feature vector is X=[A, B, C, D]; Model training: Using feature vector X from historical data train and the corresponding health status label y train Train the model; Use support vector machine SVM to optimize the objective function: Where C is the regularization parameter, ξ i is the slack variable; w is the normal vector of the hyperplane, which determines the direction of the hyperplane; b is the displacement term of the hyperplane, which determines the distance between the hyperplane and the origin; ξ i is a slack variable used to deal with noise or indistinguishability in data points; C is a regularization parameter used to control the complexity of the model and the degree of penalty for misclassification; m is the number of data points; The optimization goal of SVM is to find a set of parameters w, b and ξ i , so that the objective function reaches the minimum value.

9. The intelligent data processing system for accessing new energy to a power grid according to claim 1 is characterized by: The design parameters of the device input electrical compensation unit include the following: Enter the data collection parameters: Data collection frequency: set the time interval for data collection to ensure the real-time and accuracy of data; Data accuracy: Set the accuracy of data collection to meet the system's needs for data processing and analysis; Geographic location information accuracy: Set the accuracy of geographic location information to ensure the accuracy of device location; Enter electroanalysis parameters: Rated input power setting: set the rated input power value for each renewable energy power generation equipment as a basis for judging whether the input power is excessive or insufficient; Input power monitoring threshold: Set the threshold for input power monitoring. When the actual input power exceeds or falls below the input power monitoring threshold, input power analysis is triggered. Dynamic compensation adjusts model parameters: Compensation strategy priority: set the priority of different compensation strategies; Compensation efficiency coefficient: Set the efficiency coefficient during the compensation process to reflect the losses during power transmission, storage, and conversion; Model iteration optimization cycle: Set the model iteration optimization cycle to ensure that the model can adapt to changes in power grid operation; Assume that the actual input power of a new energy power generation equipment is P actual , rated input power is P rated , then the calculation formula for excess or insufficient input power is: ΔP=P actual -P rated ; Among them, ΔP represents the excess or shortage of input electricity. When ΔP>0, it means excess, and when ΔP<0, it means shortage; Regional input electrical balance analysis: Assume that the number of power generation equipment connected to the same power grid in the same area is M, and the excess or shortage of input power of each equipment is ΔP j , and j∈1, 2, ..., M, then the analytical formula for the regional input electrical balance is: In addition to analyzing the actual collected data to compensate for each new energy power generation equipment, the equipment input power compensation unit also inputs the values ​​predicted by the regularity prediction module to predict the values ​​that need to be compensated within a specified time period in the future.

10. The intelligent data processing system for accessing new energy to a power grid according to claim 9, characterized in that: The analytical formula of power balance indicates that in the same area, the excess and shortage of input power of all power generation equipment should offset each other to achieve a balanced state; however, in actual situations, due to the influence of various factors, it is impossible to achieve a complete balance. In this case, it is necessary to optimize through a dynamic compensation adjustment model; Compensation strategy formulation: When formulating the compensation strategy, an optimization model is constructed based on the surplus and shortage of input electricity, the geographical location of the equipment, and power demand factors; the objective function of the optimization model is set to minimize power transmission losses or maximize power utilization efficiency; Let x j represents the compensation amount of the jth device, and the optimization model is expressed as: Among them, c j Represents the unit cost of compensation for the jth device; the goal of the optimization model is to minimize the total cost of compensation while satisfying the regional input power balance; an iterative algorithm is used to solve it. In each iteration, the model parameters are adjusted and the compensation amount is recalculated according to the current execution of the compensation strategy; through multiple iterations, the optimal solution is gradually approached.

Images