Distributed power supply-based control method, system, equipment and medium

Abstract

The invention discloses a distributed power supply-based control method, system, equipment and medium, and relates to the technical field of distributed power supply control, and the method comprises the steps: collecting and preprocessing the real-time data of a distributed power supply, training a prediction model to optimize parameters, obtaining a load demand and power generation capability estimation, and adjusting the power based on a prediction result. Distributing active power according to a total load and a power contribution factor; setting a target power factor to calculate reactive power; monitoring voltage and load changes of a power grid in real time; dynamically selecting and adjusting a reactive power control strategy; and visualizing the running state and backing up the data to a database. Through closed-loop cooperation of data prediction, dynamic power control, value-oriented intelligent operation and maintenance and visual management, the stability, electric energy quality, operation and maintenance economy and management efficiency of the distributed power supply system are comprehensively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distributed power supply control, and particularly relates to a distributed power supply control method, system, device and medium. BACKGROUND

[0002] With the development of renewable energy technology and the transformation of energy structure, distributed power supply has gradually become an important part of modern power systems. Traditional centralized power systems cannot fully adapt to the large-scale access of new distributed energy due to their single distributed power supply mode. To cope with this change, distributed power supply control methods have gradually emerged, which collect and analyze real-time data to reasonably allocate active power and reactive power of distributed power supply to balance load demand.

[0003] Modern distributed power supply control methods can not only achieve accurate prediction of load demand by introducing intelligent sensors, data communication and deep learning algorithms, but also adjust power flow according to the output characteristics of distributed power supply to improve the response speed and stability of power systems. However, existing control methods mostly rely on traditional prediction and regulation models, which are difficult to fully meet the complex operating environment and real-time requirements of distributed power supply. In addition, with the increase of distributed energy access scale, the stability and power quality of power grid face greater challenges, and existing technologies still need to be further optimized in terms of real-time processing of multi-source data, prediction accuracy and fault detection. SUMMARY

[0004] In view of the above existing problems, the present application provides a distributed power supply control method, system, device and medium to solve the problems of slow response speed and poor stability of power systems in the prior art.

[0005] To solve the above technical problems, a distributed power supply control method is proposed, which includes, Collecting real-time operating data of distributed power supply and preprocessing the data, based on the preprocessed data, performing load prediction and power generation capacity prediction, and obtaining load demand and power generation capacity estimation by training prediction model optimization parameters; based on the prediction results, adjusting power, allocating active power according to total load demand and power contribution factor, setting target power factor and calculating reactive power, real-time monitoring of power grid voltage and load change, selecting reactive power control strategy according to load fluctuation degree, and dynamically adjusting reactive power output based on the selected strategy; real-time monitoring of the operating state of each distributed power supply, detecting abnormal data points, classifying abnormal data points, generating maintenance priority based on the fault classification results and reactive power benefits, and formulating maintenance measures based on the maintenance priority; visualizing the operating state of distributed power supply and storing the operating data in the database for backup.

[0006] As a preferred scheme of the distributed power supply control method, the data preprocessing comprises collecting multi-source real-time data from the distributed power supply and the load nodes through a sensor network; the collected data is quality evaluated to identify and eliminate abnormal values and missing values; and data in different formats is converted into a unified standard format. The converted data is normalized, and the processed data is arranged and stored in time series.

[0007] As a preferred scheme of the distributed power supply control method, the load prediction and power generation capacity prediction comprise constructing a prediction model architecture, using historical operation data as a training sample set, adjusting internal parameters of the model through an iterative optimization algorithm, setting a model convergence condition, stopping training when the condition is met, inputting real-time data into the trained model to obtain a prediction result, and performing credibility evaluation and error analysis on the prediction result.

[0008] As a preferred scheme of the distributed power supply control method, the power adjustment comprises extracting total load demand and power generation capacity of each power supply from the prediction result, calculating power distribution weights based on power supply capacity and power generation capacity, distributing active power output of each power supply according to the weight proportion, comparing the distributed power with the actual power generation capacity, and performing power limiting processing; calculating the required reactive power according to the target power factor and the active power, monitoring the power grid operation state, identifying the load fluctuation level, selecting a matched control mode according to the fluctuation level, and performing reactive power regulation operation of the selected control mode.

[0009] As a preferred scheme of the distributed power supply control method, the detection of abnormal data points comprises establishing control charts of voltage, current, temperature, active power and reactive power, setting a 3σ control limit, calculating Z-score values of data points exceeding the limit, and setting Z-score absolute values greater than 3 as abnormal; calculating the probability density of the abnormal points using a Gaussian distribution model, setting a probability density threshold H, and determining as abnormal if the threshold is lower; and using a random forest algorithm for fault classification, inputting abnormal data features, and outputting fault levels.

[0010] As a preferred scheme of the distributed power supply control method, the reactive power benefit comprises calculating the reactive power cost of each power supply using a quadratic cost function, obtaining the marginal cost and the local marginal price and benefit value by taking the partial derivative of the cost function with respect to the reactive power; The reactive power cost formula is calculated as: wherein, is the economic benefit obtained by the ith distributed power supply by providing reactive power, is the reactive power of the ith distributed power supply, is the local marginal price, that is, the incremental cost of meeting the unit reactive power demand in the power system, is the total number of distributed power supplies, is the cost function of the ith distributed power supply to meet the reactive power , and is the rated reactive power capacity of the ith distributed power supply. The generation of the maintenance priority includes arranging the fault power supply in descending order of fault level, arranging the fault power supply in descending order of reactive power benefit in the same fault level, the high priority includes all serious faults and high benefit medium faults, the medium priority includes medium-low benefit medium faults and high benefit slight faults, and the low priority includes medium-low benefit slight faults.

[0011] As a preferred scheme of the distributed power supply control method provided by the application, the operation state of the distributed power supply is displayed through visualization, including generating real-time graphs of voltage, current and power factor using an ECharts library, drawing a stacked column chart of active power and reactive power using a D3.js library, displaying power contribution of each power supply, and marking abnormal equipment with a red warning icon on an abnormal monitoring panel, and displaying specific parameter values and abnormal reasons on hovering. The data report supports filtering according to a time range, and when exported as PDF, includes all running parameter statistical charts, and when exported as Excel, includes an original data table; the data storage adopts a time series database, is sorted according to collection time and is marked with an equipment ID and a parameter type label; the cloud backup adopts an incremental backup strategy, and MD5 verification is performed daily to ensure data integrity.

[0012] The preferred technical scheme has the beneficial effects of realizing sensitive and reliable identification of early equipment failure by monitoring the operation state in real time and detecting abnormal data points through a control chart and a Gaussian distribution model double mechanism, reducing the false negative and false positive rates, classifying the failure into levels by using a random forest algorithm, generating a maintenance priority in combination with reactive power benefit calculation, enabling operation and maintenance resources to be preferentially invested in equipment that has the greatest impact on system safety and economy, shortening the failure processing time, reducing operation and maintenance costs, and ensuring long-term safe and economic operation of the system.

[0013] As a preferred scheme of the distributed power supply control system provided by the application, the system comprises a data collection module, a prediction module, a power adjustment module, a failure detection module and a data storage module.

[0014] The data collection module is configured to collect real-time data from each distributed power supply and load node through a sensor and to pre-process the data.

[0015] The prediction module is configured to perform load prediction using an LSTM model and to perform power generation capacity prediction using a convolutional neural network.

[0016] The power adjustment module is configured to perform power distribution and to adjust a power factor in real time based on the load and power generation capacity prediction results, to collect power grid voltage and load change information in real time, and to dynamically adjust active and reactive power distribution.

[0017] The fault detection module is configured to monitor real-time operation states of each distributed power supply, to perform fault detection and classification, to generate a priority ranking and to develop maintenance measures according to fault severity and reactive power benefits.

[0018] The data storage module is configured to display real-time operation states of the distributed power supply through a chart and to store all collected and analyzed distributed power supply operation data in a local database.

[0019] A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method for controlling distributed power supplies when executing the computer program.

[0020] A computer readable storage medium stores a computer program, and the computer program implements the steps of the method for controlling distributed power supplies when executed by a processor.

[0021] The present application has the following advantages: The present application changes the control from passive response to active prediction through high-quality data preprocessing and accurate prediction, lays a foundation for optimal scheduling, performs fine power distribution and dynamic reactive power strategy adjustment based on prediction results, actively maintains power quality while ensuring supply and demand balance, realizes the change from regular maintenance to value-oriented condition-based maintenance through double abnormality detection and fault classification combined with economic benefits, optimizes operation and maintenance resource allocation, intuitively visualizes and safely stores and manages data, improves human-computer interaction efficiency and decision-making level, provides data asset support for continuous iteration of the system, and jointly constitutes an efficient, intelligent and economic distributed power supply control system. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0023] Figure 1 This is a general flowchart of a distributed power supply control method provided in one embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of a fault detection process based on a distributed power control method, provided as an embodiment of the present invention.

[0025] Figure 3 The present invention provides a system scheme flowchart based on a distributed power supply control system according to one embodiment of the present invention. Detailed Implementation

[0026] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0027] Example 1, referring to Figure 1 As one embodiment of the present invention, a distributed power supply control method is provided, comprising: S100: Collects real-time operating data of distributed power sources, preprocesses the data, performs load forecasting and power generation capacity forecasting based on the preprocessed data, and obtains load demand and power generation capacity estimates by training the forecasting model and optimizing the parameters.

[0028] S200: Based on the prediction results, it adjusts the power output, allocates active power according to the total load demand and power contribution factor, sets the target power factor and calculates reactive power, monitors grid voltage and load changes in real time, selects reactive power control strategies according to the degree of load fluctuation, and dynamically adjusts reactive power output based on the selected strategies.

[0029] S300: Real-time monitoring of the operating status of each distributed power source, detection of abnormal data points, fault classification of abnormal data points, generation of maintenance priorities based on fault classification results and reactive power gains, and formulation of maintenance measures based on maintenance priorities.

[0030] S400: Visualizes the operating status of distributed power sources and stores the operating data in a database for backup.

[0031] It should be noted that by introducing an adaptive reactive power control strategy, this invention can dynamically select the control mode based on the significance of load fluctuations, thereby improving the reliability of power grid operation and power quality. Random forest is used to classify distributed power source data for faults, enabling accurate differentiation between minor, moderate and severe faults. Furthermore, reactive power revenue ranking is combined to generate maintenance priorities for faulty distributed power sources, thus optimizing the rational allocation of maintenance resources.

[0032] Example 2, refer to Figure 1 and Figure 2 This is a second embodiment of the present invention, which provides a distributed power supply control method, including: In step S100, collecting real-time operating data of distributed power sources includes collecting multi-source real-time data from distributed power sources and load nodes through a sensor network, wherein the multi-source real-time data includes voltage, current, load, power data and meteorological data.

[0033] In this embodiment of the application, the data preprocessing in step S100 includes steps S101 to S104: S101: Perform data cleaning by setting a data range threshold, using a sliding window filtering algorithm to remove sensor noise, and using interpolation to fill in missing data.

[0034] S102: Perform data format conversion, unifying data from different communication protocols such as Modbus and IEC61850 into JSON format to ensure data compatibility.

[0035] S103: Normalize the converted data by using the minimum-maximum normalization method to map all data to the [0,1] interval and eliminate dimensional differences.

[0036] S104: Organize and store the processed data according to the time series to provide input for the subsequent prediction module.

[0037] In an optional implementation, in step S100, the data preprocessing further includes: performing multi-resolution analysis on the data using wavelet transform to identify and remove high-frequency noise components; retaining effective signals using threshold processing; performing data format conversion to uniformly convert data from different protocols into XML format; storing metadata using the hierarchical structure of XML; standardizing the data by using the Z-score standardization method to adjust the data to a distribution with a mean of 0 and a variance of 1; and indexing the processed data by timestamp and storing it in a time-series database.

[0038] In another optional implementation, in step S100, the data preprocessing may further include: using the isolated forest algorithm for anomaly detection, training a model to identify and remove outliers, and using a moving average method to smooth the data; performing data format conversion, serializing the data into Avro binary format, and using Avro's schema evolution function to support dynamic data structures; performing feature scaling on the data, using the RobustScaler method to reduce the impact of outliers, and classifying and storing the data in a distributed file system according to device ID and timestamp.

[0039] Furthermore, in step S100, the load forecasting and power generation capacity forecasting include steps S111~S113: S111: Use the LSTM model to predict distributed generation load. Input historical load data as the training set into the LSTM model for model training. Define the loss function and Adam optimizer to iteratively optimize the model parameters. Stop the iteration when the loss of the LSTM model no longer decreases during continuous iteration. Output the model parameters to update the LSTM model. Input real-time load data into the LSTM model to obtain the predicted load demand.

[0040] S112: A power generation capacity prediction model is built using a convolutional neural network, including an input layer, a convolutional layer, a pooling layer, and an output layer.

[0041] S113: The collected meteorological data and historical power generation data are used as the training set to input into the power generation capacity prediction model for iterative training. The loss function and Adam optimizer are defined to iteratively optimize the model parameters. When the loss of the power generation capacity prediction model no longer decreases during continuous iteration, the iteration is stopped, and the model parameters are output to update the power generation capacity prediction model.

[0042] It should be noted that, in the embodiments of this application, load forecasting in step S111 includes steps A1 to A4: A1: Collect historical load data as a training set, including time series load values.

[0043] A2: Construct an LSTM model architecture. The input layer receives historical load data, the hidden layer contains 128 neurons, and the output layer outputs the load prediction value for the next 24 hours.

[0044] A3: Use the Adam optimizer to train the model, define the loss function and iteratively optimize the model parameters. Stop training when the loss function decreases by less than 0.001 after 10 consecutive iterations.

[0045] A4: Input real-time load data into the trained LSTM model to obtain future load demand prediction results, and perform credibility assessment and error analysis on the prediction values.

[0046] In an optional implementation, step S111, obtaining the predicted load demand further includes collecting historical load data and performing a stationarity test, making the time series stationary through differencing, fitting an ARIMA model, identifying autocorrelation and partial autocorrelation functions to determine the model order, optimizing parameters using maximum likelihood estimation, performing model validation, ensuring the model fit effect through residual analysis, inputting real-time data into the ARIMA model, outputting future load forecast values, and updating the results using a rolling forecasting method.

[0047] In another optional implementation, in step S111, obtaining the predicted load demand may further include: collecting historical load data and related features (temperature, date type), selecting a radial basis function (RBF) as the kernel function, training an SVM regression model, optimizing hyperparameters using grid search; performing model evaluation, ensuring generalization ability through cross-validation, inputting real-time feature data into the SVM model, outputting the load prediction value, and providing an uncertainty estimate in conjunction with confidence intervals.

[0048] Furthermore, in this embodiment of the application, step S113, the power generation capacity prediction includes steps B1 to B4: B1: Collect meteorological data (temperature, light intensity) and historical power generation data as the training set.

[0049] B2: Construct a CNN model architecture. The input layer receives multi-dimensional data, extracts spatial features through 3 convolutional layers and 2 pooling layers, and the output layer outputs the predicted power generation capacity of each distributed power source.

[0050] B3: Use the Adam optimizer to train the model, define the loss function and iteratively optimize it, and stop training when the loss function no longer decreases significantly in consecutive iterations.

[0051] B4: Input real-time meteorological data and power generation data into the trained CNN model to obtain predicted power generation capacity, and perform error analysis and calibration on the results.

[0052] In an optional implementation, step S113 further includes collecting meteorological data and historical power generation data, extracting features (wind speed, sunshine duration), training a random forest model, constructing multiple decision trees, reducing overfitting through Bootstrap aggregation, performing feature importance analysis, optimizing feature selection, inputting real-time feature data into the model, outputting predicted power generation capacity values, and improving robustness by utilizing the integrated results.

[0053] In another optional implementation, step S113 may further include collecting historical data and performing feature engineering, including encoding time features and meteorological features, training an XGBoost model, minimizing the loss function by iteratively adding a decision tree, and using an early stopping method to prevent overfitting; performing model tuning, selecting the best parameters through cross-validation, inputting real-time data into the model, outputting a power generation capacity prediction, and interpreting the prediction results in conjunction with Shapley values.

[0054] In this embodiment of the application, step S200, the power adjustment includes steps S201 to S206: S201: Extract the total load demand and the predicted generation capacity of each distributed power source from the forecast results, and collect the rated capacity.

[0055] S202: Calculate the power contribution factor for each distributed power source, based on the sum of the product of rated capacity and predicted generation capacity, expressed by the formula: in, Let be the power contribution factor of the i-th distributed power source. Let be the rated capacity, i.e., the maximum active power output capability of the i-th distributed power source. Let be the predicted power generation capacity of the i-th distributed power source. This represents the total number of distributed power sources. Let be the rated capacity, i.e., the maximum active power output capability of the j-th distributed power source.

[0056] S203: Active power is allocated based on the power contribution factor and total load demand. The allocated active power is compared with the predicted generation capacity of each distributed power source. When the allocated power of a distributed power source exceeds the predicted generation capacity, the predicted generation capacity value is used as the final power value, and the excess power is recovered. The formula is as follows: in, The active power allocated to the i-th distributed power source This refers to the total active power required to meet future total load demand, which must be met by all distributed power sources.

[0057] S204: Set the target power factor, calculate reactive power based on active power, and derive the formula for calculating reactive power using trigonometric functions: in, For the target power factor, Let be the reactive power of the i-th distributed power source.

[0058] S205: Real-time acquisition of load changes in grid voltage, setting load thresholds using the quantile threshold method and optimizing them based on actual application scenarios; if the load is greater than or equal to the load threshold, it is considered a significant fluctuation; otherwise, it is considered a slight fluctuation. When the load fluctuates slightly, the Q(V) control strategy, i.e., the reactive power control strategy based on voltage response, is selected; when the load fluctuates significantly, the PF(P) control strategy, i.e., the reactive power control strategy based on power factor regulation, is selected.

[0059] S206: Select the control strategy according to the degree of fluctuation. When there is slight fluctuation, the voltage control strategy is adopted, and the reactive power compensation is calculated by the voltage deviation. When there is significant fluctuation, the PF(P) power factor control strategy is adopted, and the reactive power output is adjusted by the power factor deviation. When the Q(V) control strategy is selected, the formula for calculating the reactive power compensation is expressed as: in, This is the reactive power compensation amount. The target voltage set for the system, The current grid voltage, As the reference voltage, The reactive power regulation coefficient is determined through regression analysis using historical reactive power regulation data. When the PF(P) control strategy is selected, the current power factor is calculated using the active and reactive power of the distributed power source, expressed by the formula: in, Let be the active power of the i-th distributed power source. Let be the reactive power of the i-th distributed power source. Here, is the power factor, and i is the index of the distributed power source.

[0060] In an optional implementation, step S200 further includes extracting load and generation data from the prediction results, defining a fuzzy rule base, with input variables being voltage deviation and load change rate, and output variable being reactive power adjustment amount, performing fuzzy inference, calculating the output membership degree using the Mamdani method, defuzzifying to obtain accurate reactive power compensation value, adjusting the output of each distributed power source according to the results, and monitoring system stability in real time.

[0061] In another optional implementation, in step S200, the power adjustment may further include: establishing a power grid system model, including state-space equations to represent power flow, obtaining future load and generation sequences from prediction results, solving an optimization problem, minimizing the objective function (power deviation and cost), calculating the optimal control input, and applying the control sequence in real time to adjust active and reactive power, updating the model through feedback correction, and handling uncertainties.

[0062] In step S300, the detection of abnormal data points includes steps S301 to S304: S301: Establish control diagrams for voltage, current, temperature, active power, and reactive power, and set 3σ control limits.

[0063] S302: Calculate the Z-score value for data points that exceed the limit, and set the absolute value of the Z-score to be greater than 3 as an anomaly; S303: Calculate the probability density of outliers using a Gaussian distribution model, setting a probability density threshold H=0.01. A value below the probability density threshold is considered abnormal; the formula for calculating the probability density is as follows: in, The values ​​of the data points to be evaluated. For probability density, The mean of the dataset. Let be the standard deviation of the dataset. Pi It is an exponential function.

[0064] S304: Fault classification uses the random forest algorithm. Input abnormal data features and output fault level: 1 for minor fault, 2 for moderate fault, and 3 for severe fault.

[0065] Furthermore, in step S300, the reactive power gain includes steps S311~S313: S311: Calculate the reactive power cost of each power source using the quadratic cost function, obtain the marginal cost using the partial derivative of the cost function with respect to reactive power, and obtain the local marginal price and revenue value. The formula for calculating reactive power cost is expressed as follows: in, Let be the economic benefit obtained by the i-th distributed power source by providing reactive power. Let be the reactive power of the i-th distributed power source. The local marginal price is the incremental cost of satisfying the unit reactive power demand in the power system. This represents the total number of distributed power sources. To satisfy the reactive power requirement of the i-th distributed power source The cost function, Let be the rated reactive power capacity of the i-th distributed power source.

[0066] S312: Based on the calculated reactive power revenue, the distributed power sources are sorted in descending order. The revenue data is divided into three intervals: high, medium, and low using the quantile method. The high-revenue interval includes the distributed power sources with the highest revenue in the top 1 / 3 of the data (i.e., the top 33.3%); the medium-revenue interval includes the distributed power sources with the middle 1 / 3 of the data (i.e., the data from 33.3% to 66.6%); and the low-revenue interval includes the distributed power sources with the lowest 1 / 3 of the data (i.e., the bottom 33.3%). S313: When the returns are unevenly distributed or the data is concentrated, adjust the boundaries according to the data distribution. The percentile method includes defining high returns as returns at or above the 75th percentile, medium returns as returns between the 25th and 75th percentiles, and low returns as returns at or below the 25th percentile.

[0067] It should be noted that, in step S304, the fault classification also includes steps C1 to C4: C1: Random forest is used to classify abnormal data points into minor, moderate and severe faults, generating fault levels (severe fault is 3, moderate fault is 2, minor fault is 1). Distributed power sources are classified according to fault levels. The maintenance priority of the faulty distributed power sources is generated by combining the fault level classification results and the reactive power gain ranking results. Based on the maintenance priority, matching maintenance measures are formulated.

[0068] C2: When the faulty distributed power supply is of high priority, the fault type and severity will be confirmed immediately, emergency maintenance resources (such as spare parts and technicians) will be automatically allocated, and the task will be assigned to the maintenance team first. When the fault allows for remote repair, an adjustment command will be sent immediately to restore the operating status.

[0069] C3: When the faulty distributed power supply is of medium priority, it shall be included in the priority maintenance task within the current maintenance cycle or the day. However, after the high priority equipment has been dealt with, the maintenance team shall be arranged to carry out on-site inspection as soon as possible during the off-peak hours.

[0070] C4: When the faulty distributed power supply is of low priority, the low-priority device is marked as a "planned maintenance" task and scheduled for processing in the next round of regular maintenance cycle. The fault status is continuously monitored before the planned maintenance period arrives.

[0071] Furthermore, in step S300, the maintenance measures based on maintenance priority include classifying and sorting all distributed power sources according to the fault level from high to low, sorting distributed power sources with the same fault level according to the reactive power gain from high to low, and dividing the maintenance priority into high priority, medium priority and low priority based on the sorting results. Among them, high priority includes distributed power sources with severe faults in all benefit ranges and medium faults with only high benefit; medium priority refers to medium faults with both medium and low benefit ranges and minor faults with only high benefit; and low priority refers to minor faults with both medium and low benefit ranges.

[0072] In step S400, the visualization of the operating status of the distributed power source includes steps S401 to S405: S401: Uses the ECharts library to generate real-time curves of voltage, current, and power factor, and uses the D3.js library to draw stacked bar charts of active and reactive power, showing the power contribution of each power source.

[0073] S402: The anomaly monitoring panel uses a red warning icon to mark abnormal devices, and hovering over it displays the specific parameter values ​​and the cause of the anomaly.

[0074] S403: Data reports support filtering by time range. When exported as PDF, they include statistical charts of all running parameters. When exported as Excel, they include the original data table.

[0075] S404: Data storage uses a time-series database, sorted by acquisition time and labeled with device ID and parameter type tags.

[0076] S405: Cloud backup adopts an incremental backup strategy and performs MD5 verification daily to ensure data integrity.

[0077] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0078] Example 3, referring to Figure 3 This is the third embodiment of the present invention, which provides a distributed power control system, including a data collection module, a prediction module, a power adjustment module, a fault detection module, and a data storage module.

[0079] The data collection module is used to collect real-time data from each distributed power source and load node through sensors and to perform preprocessing.

[0080] The prediction module is used to perform load forecasting using an LSTM model and to perform power generation capacity forecasting using a convolutional neural network.

[0081] The power adjustment module is used to allocate power and adjust the power factor in real time based on load and power generation capacity prediction results, and dynamically adjust the active and reactive power allocation by collecting grid voltage and load changes in real time.

[0082] The fault detection module is used to monitor the real-time operating status of each distributed power source, perform fault detection and classification, generate priority ranking and formulate maintenance measures based on fault severity and reactive power gain.

[0083] The data storage module is used to display the real-time operating status of the distributed power source through charts and to store all collected and analyzed distributed power source operating data in a local database.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0085] Example 4, the fourth embodiment of the present invention, differs from the previous three embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0087] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0088] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

Claims

1. A method for controlling distributed power sources, characterized by: The application relates to a distributed power supply system and a power supply method thereof. Real-time operation data of distributed power supplies are collected and preprocessed, based on the preprocessed data, load prediction and power generation capacity prediction are carried out, and load demand and power generation capacity estimation are obtained by training a prediction model and optimizing parameters; Based on the prediction results, power adjustment is carried out, active power is distributed according to total load demand and power contribution factors, a target power factor is set and reactive power is calculated, grid voltage and load changes are monitored in real time, reactive power control strategies are selected according to the load fluctuation level, and reactive power output is dynamically adjusted based on the selected strategies; The operation states of the distributed power supplies are monitored in real time, abnormal data points are detected, fault classification is carried out on the abnormal data points, maintenance priorities are generated based on the fault classification results and reactive power benefits, and maintenance measures are formulated based on the maintenance priorities; The operation states of the distributed power supplies are visualized and displayed, and the operation data are stored in a database for backup.

2. The method of claim 1, wherein: The data preprocessing includes collecting multi-source real-time data from distributed power supplies and load nodes through a sensor network, quality evaluation of the collected data, identification and elimination of abnormal values and missing values, and conversion of data in different formats into a unified standard format. The converted data are normalized, and the processed data are arranged and stored in time series.

3. The method of claim 2, wherein: The load prediction and power generation capacity prediction include constructing a prediction model architecture, using historical operation data as a training sample set, adjusting model internal parameters through an iterative optimization algorithm, setting model convergence conditions, stopping training when the conditions are met, inputting real-time data into the trained model to obtain prediction results, and performing credibility evaluation and error analysis on the prediction results.

4. The method of claim 3, wherein: The power adjustment includes extracting total load demand and power generation capacity of each power supply from the prediction results, calculating power distribution weights based on power supply capacity and power generation capacity, distributing active power output of each power supply according to the weight proportion, comparing the distributed power with the actual power generation capacity, and performing power limiting processing; The required reactive power is calculated according to the target power factor and active power, the grid operation state is monitored, the load fluctuation level is identified, a matched control mode is selected according to the fluctuation level, and reactive power adjustment operation of the selected control mode is executed.

5. The method for controlling distributed power sources according to claim 4, wherein: The detection of abnormal data points includes establishing control charts of voltage, current, temperature, active power and reactive power, setting a 3sigma control limit, and calculating Z-score values of data points exceeding the limit; The probability density of the abnormal points is calculated using a Gaussian distribution model, a probability density threshold H is set, and the abnormality is determined if the threshold is lower than the threshold; the fault classification uses a random forest algorithm, inputs the abnormal data features, and outputs the fault level.

6. The method for controlling distributed power sources according to Claim 5, wherein: The reactive power benefit includes calculating the reactive power cost of each power supply using a quadratic cost function, obtaining the marginal cost by taking the partial derivative of the cost function with respect to the reactive power, and obtaining the local marginal price and benefit value; The formula for calculating the reactive power cost is as follows: wherein, is the economic benefit obtained by the ith distributed generator by providing reactive power, is the reactive power of the ith distributed generator, is the local marginal price, i.e. the incremental cost of meeting a unit demand for reactive power in the power system, is the total number of distributed generators, is the cost function of the ith distributed generator to meet the reactive power demand, is the rated reactive power capacity of the ith distributed generator; The generation of maintenance priorities includes arranging the faulty power supplies in descending order of fault level, and arranging the same fault level in descending order of reactive power benefit, the high priority includes all serious faults and high benefit medium faults, the medium priority includes medium-low benefit medium faults and high benefit slight faults, and the low priority includes medium-low benefit slight faults.

7. The method for controlling distributed power sources according to Claim 6, wherein: The visualization of the operation state of the distributed power supply includes generating real-time graphs of voltage, current and power factor using the ECharts library, drawing a stacked column chart of active power and reactive power using the D3.js library, showing the power contribution of each power supply, and using a red warning icon to mark abnormal devices on the abnormal monitoring panel, and hovering to display specific parameter values and abnormal reasons; The data report supports filtering by time range, and when exported as PDF, it includes all running parameter statistical charts, and when exported as Excel, it includes raw data tables; the data storage uses a time series database, which is sorted by collection time and labeled with device ID and parameter type tags; The cloud backup adopts an incremental backup strategy, and performs MD5 verification daily to ensure data integrity.

8. A distributed power source control system applying the distributed power source control method according to any one of claims 1 to 7. The method comprises a data collection module, a prediction module, a power adjustment module, a fault detection module, and a data storage module. The data collection module is configured to collect real-time data from each distributed power supply and load node through sensors and perform preprocessing; The prediction module is configured to use an LSTM model for load prediction and a convolutional neural network for power generation capacity prediction; The power adjustment module is configured to perform power distribution and real-time adjustment of power factor based on the prediction results of load and power generation capacity, and to dynamically adjust active and reactive power distribution by real-time collection of power grid voltage and load change conditions; The fault detection module is configured to monitor the real-time operation state of each distributed power supply, perform fault detection and classification, generate priority ranking according to fault severity and reactive power benefit, and develop maintenance measures; The data storage module is configured to display the real-time operation state of the distributed power supply through charts, and store all collected and analyzed distributed power supply operation data in a local database. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to realize the steps of the distributed power supply control method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the distributed power supply control method in any one of claims 1 to 7.

Images