An intelligent bidirectional power supply stable output control system and method
Through an intelligent two-way power stable output control system, the control parameters of the two-way DC-DC converter are automatically adjusted through the intelligent two-way power supply stable output control system in complex wind power generation environments. The problem of difficulty in ensuring efficiency achieves efficient and stable power output and adaptability.
Patent Information
- Application Number
- CN202410778108.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-17
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-06-17
AI Technical Summary
The traditional two-way DC-DC converter control system is difficult to ensure the stability and efficiency of power output in complex and changing wind power generation environments, and the fixed parameter settings cannot adapt to wind speed fluctuations, temperature changes and grid load changes.
An intelligent two-way power supply stable output control system is designed, including data perception layer, data processing layer, intelligent decision-making layer and control execution layer. The environment perception module and power state perception module are used to monitor data in real time, and through data cleaning and machine learning modules Perform prediction and optimization, and automatically adjust the control parameters of the bidirectional DC-DC converter.
It improves the stability and efficiency of power output, enhances the adaptability of the system, reduces the need for manual intervention, improves the reliability and operating efficiency of the system, and extends the service life of the equipment.
Smart Images

Figure CN118399743B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power supply output control, and in particular to an intelligent bidirectional power supply stable output control system and method. Background Art
[0002] With the increasing global attention and utilization of renewable energy, wind power generation, as a clean and sustainable form of energy, has been widely promoted and applied worldwide. In wind power generation systems, power management systems play a vital role, especially in power conversion and storage. Bidirectional DC-DC converters, as power conversion devices that can operate in both Buck (step-down) and Boost (boost) modes, are widely used in the power conversion process of wind power generation systems to meet different power requirements.
[0003] However, wind power generation systems face a complex and changeable operating environment, including wind speed fluctuations, temperature changes, grid load changes, etc. These changes directly affect the working efficiency and stability of the bidirectional DC-DC converter. The traditional bidirectional DC-DC converter control system mainly relies on manually set fixed control parameters, such as switching frequency, duty cycle, inductor current limit, output voltage reference, etc., to achieve stable power output. However, in the complex and changeable wind power generation environment, fixed parameter settings are often difficult to cope with various situations and cannot guarantee the stability and efficiency of power output. To this end, we propose an intelligent bidirectional power supply stable output control system and method. Summary of the invention
[0004] The object of the present invention is to provide an intelligent bidirectional power supply stable output control system and method to solve the problems raised in the above background technology.
[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an intelligent bidirectional power supply stable output control system, comprising a data perception layer, a data processing layer, an intelligent decision-making layer and a control execution layer;
[0006] The data perception layer includes an environment perception module and a power state perception module; the environment perception module is used to monitor the environmental parameters of the wind power generation system in real time, and send the environmental parameters to the data preprocessing layer; the power state perception module is used to monitor the working state parameters of the bidirectional DC-DC converter in real time, and send the working state parameters to the data preprocessing layer;
[0007] The data preprocessing layer includes a data cleaning module, which is responsible for cleaning the raw data collected by the data perception layer, removing outliers and duplicate data, and sending the cleaned data to the intelligent decision-making layer;
[0008] The intelligent decision-making layer includes a machine learning module and a decision optimization module; the machine learning module is used to use the received data to make real-time predictions, automatically optimize and adjust the control parameters of the bidirectional DC-DC converter, and the machine learning module includes a model training submodule and a model prediction submodule, wherein the model training submodule uses historical data for training, and the model prediction submodule makes predictions based on real-time data; the decision optimization module is used to generate the final control instructions based on the prediction results of the machine learning module, combined with the actual needs and safety constraints of the power supply system, and send the control instructions to the control execution layer;
[0009] The control execution layer includes a control instruction parsing module and a bidirectional DC-DC converter control module; the control instruction parsing module is used to receive and parse the control instruction, and convert the parsed control instruction into a specific control signal; the bidirectional DC-DC converter control module is used to adjust the working parameters of the bidirectional DC-DC converter according to the received control signal to achieve stable output of the power supply, and the bidirectional DC-DC converter control module includes a Buck mode control submodule and a Boost mode control submodule, which are used to adjust the control parameters in the buck mode and the boost mode respectively.
[0010] Preferably, the environmental parameters monitored in real time by the environmental perception module include wind speed, wind direction and temperature data;
[0011] The working status parameters monitored in real time by the power status sensing module include working mode, input voltage, output voltage, input current, output current, power and switch state; wherein the input voltage refers to the DC power supply voltage connected to the converter, and the output voltage is the DC voltage output by the converter; the input current is the current flowing into the converter, and the output current is the current flowing out of the converter; the power is the ratio of the electric energy input or output of the converter to time; the switch state refers to the on and off state of the switch element in the converter.
[0012] Preferably, the operating parameters adjusted by the bidirectional DC-DC converter include switching frequency, duty cycle, inductor current limit, current loop bandwidth, voltage loop bandwidth, synchronous rectification control and phase shift control.
[0013] Preferably, the model training submodule constructs and trains a parameter adjustment prediction model, and the specific steps include:
[0014] Step 1: Collecting the operating data of the bidirectional DC-DC converter in the wind power generation system, including environmental parameters, working state parameters and corresponding control parameters;
[0015] Step 2: Preprocess the collected data, including removing outliers, duplicate data or invalid data through data cleaning, eliminating the impact of dimensional differences on model training through data standardization and normalization, and forming a regular data set; divide the processed data set into training set, validation set and test set, where the training set is used for model training, the validation set is used for model effect verification and optimization, and the test set is used to verify the generalization ability of the model on unknown data sets;
[0016] Step 3: Select a suitable machine learning model that can predict and optimize the control parameters of the bidirectional DC-DC converter based on the input environmental parameters and working state parameters;
[0017] Step 4: Set initial parameters for the selected prediction model and prepare for training;
[0018] Step 5: Use the training set to train the parameter adjustment prediction model, and continuously adjust the model parameters through the optimization algorithm to minimize the gap between the predicted control parameters and the actual control parameters; use the validation set to verify the trained model, and adjust the model parameters or structure according to the verification results; use the test set to comprehensively evaluate the optimized model through the set evaluation indicators to ensure that the model performance meets the expected standards;
[0019] Step 6: Save the trained and verified parameter adjustment prediction model for subsequent use in real-time prediction and optimization of control parameters of the bidirectional DC-DC converter in the intelligent decision-making layer.
[0020] Preferably, a linear regression model is selected to construct a parameter adjustment prediction model, and the algorithm formula of the linear regression model is as follows:
[0021]
[0022] Where y is the target variable, representing the control parameters of the bidirectional DC-DC converter, are input variables, representing environmental parameters and power state parameters; is the regression coefficient, determined by model training; is the intercept term, which represents the basic value of the control parameter of the bidirectional DC-DC converter when there is no environmental influence and power state parameter; is the error term; the purpose of model training is to estimate the regression coefficient , so that the error between the predicted value and the true value is minimized; when the model training submodule initializes the model parameters, for each feature in the linear regression model Corresponding parameters Assign an initial value, which is set to a random number close to 0 and is also the intercept term An initial value is also given; these initial parameter values will serve as the starting point for model training and will be continuously optimized and adjusted through subsequent training processes;
[0023] The model training submodule uses the data in the training set to continuously iterate and update the parameters of the linear regression model through the gradient descent algorithm to minimize the loss function value between the true value and the predicted value; in each iteration, the value of the model parameter is adjusted according to the gradient information of the loss function to the model parameter until the loss function converges or the preset number of iterations is reached.
[0024] Preferably, the model training submodule uses the data in the test set to calculate the error index between the model prediction value and the actual value; and comprehensively evaluates the prediction accuracy, stability and generalization ability of the model through the error index, so as to determine the performance of the linear regression model in the prediction task.
[0025] Preferably, the error indicators include mean square error, mean absolute error and coefficient of determination R².
[0026] Preferably, the data cleaning module receives the raw data collected by the data perception layer, sets a reasonable value range for each type of parameter, traverses the raw data, marks the values beyond the set range as outliers, and removes the data marked as outliers from the raw data.
[0027] Preferably, the data cleaning module sorts the data after cleaning the outliers, and traverses the sorted data to compare whether adjacent data are exactly the same; if the adjacent data are exactly the same, they are marked as duplicate data; and the data marked as duplicate is removed from the data.
[0028] Preferably, a method for using an intelligent bidirectional power supply stable output control system comprises:
[0029] S1: Real-time monitoring and collection of environmental parameters of the wind power generation system and working state parameters of the bidirectional DC-DC converter through the data perception layer; the environmental parameters include wind speed, wind direction and temperature, and the working state parameters include input voltage, output voltage, input current, output current, power and switch state;
[0030] S2: The data preprocessing layer receives the raw data from the data perception layer and cleans the data through the data cleaning module to remove outliers and duplicate data to ensure the validity and accuracy of the data;
[0031] S3: The cleaned data is transmitted to the intelligent decision-making layer, where the machine learning module uses the data to make real-time predictions. The machine learning module first uses the model training submodule to train the model using historical data and optimize the model parameters. Then the model prediction submodule makes predictions based on the real-time data and automatically optimizes and adjusts the control parameters of the bidirectional DC-DC converter.
[0032] S4: The decision optimization module generates the final control instructions based on the prediction results of the machine learning module and comprehensively considers the actual needs and safety constraints of the power system;
[0033] S5: The control execution layer receives the control instructions from the decision optimization module and parses the instructions into specific control signals through the control instruction parsing module;
[0034] S6: The bidirectional DC-DC converter control module adjusts the operating parameters of the converter according to the received control signal, including the step-down control parameters in the Buck mode and the step-up control parameters in the Boost mode.
[0035] Compared with the prior art, the present invention has the following beneficial effects:
[0036] 1. Improve the stability and efficiency of power output: By introducing machine learning algorithms, the system can automatically optimize and adjust the control parameters of the bidirectional DC-DC converter, so that it can achieve efficient and stable power output in complex and changeable wind power generation environments. Compared with the traditional method of manually setting fixed parameters, this intelligent parameter adjustment method can better adapt to various environmental changes and significantly improve the stability and efficiency of power output.
[0037] 2. Enhance the system's adaptive capabilities: Since the system can monitor and respond to the environmental parameters of the wind power system and the working state parameters of the bidirectional DC-DC converter in real time, it has strong adaptive capabilities. Regardless of how the wind speed fluctuates, how the temperature changes, or how the grid load changes, the system can make real-time predictions and adjustments through the machine learning module to ensure stable power output.
[0038] 3. Improve the intelligence level of the system: The control system of the present invention realizes the comprehensive intelligence of power management through the coordinated work of the data perception layer, data processing layer, intelligent decision-making layer and control execution layer. This not only reduces the need for manual intervention, but also improves the reliability and operation efficiency of the entire system.
[0039] 4. Extend the service life of the equipment: Since the system can adjust the control parameters of the bidirectional DC-DC converter in real time, it avoids the damage that may be caused by long-term operation of the equipment in harsh environments, thereby effectively extending the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 It is the overall structure diagram of the present invention;
[0041] Figure 2 Diagram of the steps to build and train parameters for a predictive model using a linear regression model. DETAILED DESCRIPTION
[0042] 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.
[0043] See also Figure 1-2 The present invention provides a technical solution: an intelligent bidirectional power supply stable output control system, including a data perception layer, a data processing layer, an intelligent decision-making layer and a control execution layer.
[0044] The data perception layer includes an environment perception module and a power state perception module.
[0045] The environmental perception module is used to monitor the environmental parameters of the wind power generation system in real time and send the environmental parameters to the data preprocessing layer; the environmental perception module is responsible for monitoring the environmental parameters of the wind power generation system in real time, such as wind speed, wind direction, temperature, etc. These environmental parameters are key factors affecting the efficiency of wind power generation and the stability of the power management system; the environmental perception module collects data through various sensors (such as anemometers, thermometers, etc.) and sends these data to the data preprocessing layer in real time.
[0046] The power state sensing module is used to monitor the working state parameters of the bidirectional DC-DC converter in real time, and send the working state parameters to the data preprocessing layer; the power state sensing module mainly monitors the working state of the bidirectional DC-DC converter, including key parameters such as voltage, current, temperature, etc. These state parameters reflect the real-time performance and health status of the converter; the power state sensing module collects data through corresponding sensors and instruments, and also transmits these data to the data preprocessing layer in real time.
[0047] The data preprocessing layer includes a data cleaning module, which is responsible for cleaning the raw data collected by the data perception layer, removing abnormal values and duplicate data, and sending the cleaned data to the intelligent decision-making layer; after receiving the raw data from the data perception layer, the data cleaning module will perform data preprocessing, including removing abnormal values (such as erroneous data caused by sensor failure), deleting duplicate data, and properly processing missing data. The cleaned data is more accurate and reliable, providing a high-quality data foundation for subsequent intelligent decision-making.
[0048] The intelligent decision-making layer includes a machine learning module and a decision optimization module; the machine learning module is used to use the received data to make real-time predictions, automatically optimize and adjust the control parameters of the bidirectional DC-DC converter, and the machine learning module includes a model training submodule and a model prediction submodule, wherein the model training submodule uses historical data for training to build a machine learning model that can predict the optimal control parameters of the bidirectional DC-DC converter, and the model will continue to learn and optimize to improve its prediction accuracy and efficiency; the model prediction submodule uses the trained model to make predictions based on the real-time collected environmental and power status data, and automatically optimizes and adjusts the control parameters of the bidirectional DC-DC converter, such as switching frequency, duty cycle, etc.
[0049] The decision optimization module generates the final control instructions based on the prediction results of the machine learning module, combined with the actual needs and safety constraints of the power system (such as maximum current limit, voltage fluctuation range, etc.), which are designed to ensure that the bidirectional DC-DC converter can achieve efficient and stable power output in various complex and changing environments. The generated control instructions are then sent to the control execution layer.
[0050] The control execution layer includes a control instruction parsing module and a bidirectional DC-DC converter control module;
[0051] The control instruction parsing module: after receiving the control instructions from the intelligent decision-making layer, this module is responsible for parsing these instructions and converting them into specific control signals. These control signals are electrical signals that the bidirectional DC-DC converter can understand and execute.
[0052] The bidirectional DC-DC converter control module is used to adjust the working parameters of the bidirectional DC-DC converter according to the received control signal to achieve stable output of the power supply. The bidirectional DC-DC converter control module includes a Buck mode control submodule and a Boost mode control submodule.
[0053] Buck mode control submodule: When the system needs to step down, this submodule will adjust the operating parameters of the converter in Buck mode according to the control signal, such as the on and off time of the switch tube, to ensure stable and efficient buck output.
[0054] Boost mode control submodule: When the system needs to boost, this submodule will adjust the operating parameters of the converter in Boost mode according to the control signal to achieve stable and efficient boost output.
[0055] The system first monitors the environmental parameters and power supply status of the wind power generation system in real time through the data perception layer. Specifically, the environmental perception module uses various sensors (such as anemometers and thermometers) to collect key environmental parameters such as wind speed and temperature. At the same time, the power supply status perception module uses corresponding sensors and instruments to monitor the working status of the bidirectional DC-DC converter, such as voltage and current. Subsequently, the data preprocessing layer cleans the collected raw data. The data cleaning module is responsible for removing outliers and duplicate data, and appropriately filling in missing data to ensure the accuracy and reliability of the data. The cleaned data will serve as the basis for subsequent intelligent decision-making. In the intelligent decision-making layer, the present invention uses a machine learning module to predict and optimize the control parameters of the bidirectional DC-DC converter. The model training submodule uses historical data for training to build an accurate prediction model. The model prediction submodule predicts based on real-time data and provides optimal control parameter recommendations for the bidirectional DC-DC converter. The decision optimization module will comprehensively consider the actual needs and safety constraints of the power system, such as current and voltage limitations, and generate the final control instructions. These instructions are intended to ensure that the bidirectional DC-DC converter can achieve stable and efficient power output in various complex environments. Finally, in the control execution layer, the control instruction parsing module parses the received control instructions into specific control signals. The bidirectional DC-DC converter control module adjusts the converter's operating parameters in Buck mode and Boost mode, such as switching frequency, duty cycle, etc., based on these signals to achieve stable output of the power supply.
[0056] Through the above-mentioned specific implementation methods, the bidirectional power supply stable output control system based on machine learning of the present invention can perceive and adapt to the environmental changes of the wind power generation system in real time, automatically optimize and adjust the control parameters of the bidirectional DC-DC converter, which not only improves the efficiency and stability of power management, but also provides a strong guarantee for the sustainable operation of the wind power generation system.
[0057] The present invention will be further described below in conjunction with Examples 1 to 3:
[0058] In Example 1, the environmental parameters monitored in real time by the environmental perception module include wind speed, wind direction and temperature data.
[0059] Wind speed: Wind speed directly affects the power generation efficiency and output power of wind turbines. The machine learning model predicts the optimal working state of the converter based on wind speed data. For example, at high wind speeds, the converter may need to be adjusted to handle higher current and voltage.
[0060] Wind direction: Although the direct impact of wind direction on wind turbines may not be as significant as wind speed, it can affect the windward area and power generation efficiency of the wind turbine. The machine learning model will take into account changes in wind direction to optimize the control parameters of the converter, especially when the wind direction changes frequently.
[0061] Temperature: Temperature affects the heat dissipation and performance of electrical components. Higher temperatures may cause component performance degradation, while lower temperatures may affect the conductivity of materials. The machine learning model adjusts the operating parameters of the converter based on temperature data to ensure efficient and stable operation under various temperature conditions.
[0062] The working status parameters monitored in real time by the power status sensing module include working mode, input voltage, output voltage, input current, output current, power and switch state; wherein the input voltage refers to the DC power supply voltage connected to the converter, and the output voltage is the DC voltage output by the converter; the input current is the current flowing into the converter, and the output current is the current flowing out of the converter; the power is the ratio of the electric energy input or output of the converter to time; the switch state refers to the on and off state of the switch element in the converter.
[0063] Working mode (Buck mode or Boost mode): The working mode determines whether the converter is in a buck or boost state. The machine learning model predicts and adjusts the corresponding control parameters based on the current working mode to ensure stable power output in different modes.
[0064] Input voltage / output voltage: These voltage parameters are key indicators of converter performance. The machine learning model analyzes the changes in input and output voltages to predict and adjust control parameters such as switching frequency and duty cycle to maintain a stable voltage output.
[0065] Input current / output current: The current data reflects the load condition of the converter. The machine learning model will predict and adjust the converter control strategy based on the current changes to adapt to different load requirements.
[0066] Power: Power data provides information about the electrical energy input or output of the converter. The machine learning model uses power data to optimize the efficiency of the converter to ensure efficient operation under different power requirements.
[0067] Switching state: The on and off states of the switching elements directly affect the efficiency and stability of the converter. The machine learning model monitors the switching state and adjusts the control strategy accordingly to reduce switching losses and improve power conversion efficiency.
[0068] The operating parameters adjusted by the bidirectional DC-DC converter include switching frequency, duty cycle, inductor current limit, current loop bandwidth, voltage loop bandwidth, synchronous rectification control and phase shift control.
[0069] The switching frequency determines the on-off speed of the power switching devices (such as transistors or thyristors) in the converter. A higher switching frequency can reduce the size of the filter, but it may also increase switching losses and electromagnetic interference. This system achieves a balance between efficiency and electromagnetic compatibility by optimizing the switching frequency.
[0070] The duty cycle refers to the ratio of the conduction time of the power switch device in one cycle to the cycle time. By adjusting the duty cycle, the magnitude and direction of the output voltage can be controlled. When the duty cycle is greater than 50%, the electric energy flows from the input side to the output side; conversely, the electric energy flows from the output side to the input side. This system uses the flexible adjustment of the duty cycle to achieve efficient bidirectional transmission of electric energy.
[0071] The inductor current limit is the current upper limit set to protect the converter. Limiting the inductor current can effectively prevent equipment damage caused by overcurrent. This system ensures the stable operation of the converter by real-time monitoring and controlling the inductor current within a safe range.
[0072] Current loop bandwidth reflects the response speed of the current control system. A wider current loop bandwidth can improve the dynamic response performance of the system and enable the converter to adapt to load changes more quickly. This system achieves precise control of the converter output current by optimizing the current loop bandwidth.
[0073] The voltage loop bandwidth determines the regulation speed of the voltage control system. An appropriate voltage loop bandwidth can ensure the stability of the output voltage. This system effectively suppresses the fluctuation of the output voltage and provides stable power output by adjusting the voltage loop bandwidth.
[0074] Synchronous rectification control technology is used to improve the efficiency of the converter. By accurately controlling the on-off timing of the rectifier switch, the rectification loss can be reduced. This system adopts a synchronous rectification control strategy, which significantly improves the overall efficiency of the converter.
[0075] Phase-shift control is a method of controlling the output of a converter by adjusting the conduction phase of a switching device. Phase-shift control can achieve zero voltage switching (ZVS) or zero current switching (ZCS), thereby reducing switching losses.
[0076] In Example 2, the model training submodule constructs and trains a parameter adjustment prediction model, and the specific steps include:
[0077] Step 1: Collect the operating data of the bidirectional DC-DC converter in the wind power generation system, including environmental parameters, working state parameters and corresponding control parameters; these data will serve as the basis for machine learning model training.
[0078] The operating data of the bidirectional DC-DC converter is collected from the data monitoring points of the wind power generation system, including environmental parameters such as wind speed and temperature, as well as the working status parameters of the converter (such as working mode, voltage, current, etc.), and the collected data is stored in a database for subsequent processing and analysis.
[0079] Step 2: Preprocess the collected data, including removing outliers, duplicate data or invalid data through data cleaning, eliminating the impact of dimensional differences on model training through data standardization and normalization, and forming a regular data set to ensure data quality and consistency and improve the accuracy and efficiency of model training; use data processing libraries (such as NumPy, Pandas, etc.) to standardize and normalize the data to ensure that all data are in the same numerical range.
[0080] The processed data set is divided into training set, validation set and test set, where the training set is used for model training, the validation set is used to verify and optimize the model effect, and the test set is used to verify the generalization ability of the model on unknown data sets; usually, the training set occupies most of the data (such as 70%), and the validation set and test set each occupy a part (such as 15% and 15%).
[0081] Step 3: Select the linear regression model as the prediction model. The linear regression model can establish a linear relationship between input parameters (environmental and working state parameters) and output parameters (control parameters). The model can predict and optimize the control parameters of the bidirectional DC-DC converter based on the input environmental parameters and working state parameters.
[0082] Step 4: Initialize the model parameters, such as weights and biases, to prepare for model training.
[0083] Step 5: Use the training set to train the parameter adjustment prediction model, and continuously adjust the model parameters through the optimization algorithm to minimize the gap between the predicted control parameters and the actual control parameters; use the validation set to verify the trained model, and adjust the model parameters or structure according to the verification results; use the test set to comprehensively evaluate the optimized model through the set evaluation indicators to ensure that the model performance meets the expected standards;
[0084] Step 6: Save the trained and verified parameter adjustment prediction model as an executable model file or format (such as a pickle file, TensorFlow's SavedModel, etc.), and deploy the saved model to the intelligent decision-making layer so that it can subsequently receive data transmitted by the data perception layer in real time and predict and optimize the control parameters of the bidirectional DC-DC converter.
[0085] Below is a code example of building and training a parameter tuning prediction model using Python as the programming language:
[0086]
[0087] In the above code, we first import the necessary libraries. Numpy is used for numerical calculations. LinearRegression is the linear regression model in the sklearn library. Train_test_split is used to divide the training set and the test set. StandardScaler is used for data standardization. Mean_squared_error and r2_score are used to evaluate model performance. Joblib is used to save and load the model.
[0088] Select the linear regression model to build a parameter adjustment prediction model. The algorithm formula of the linear regression model is as follows:
[0089]
[0090] Where y is the target variable, representing the control parameters of the bidirectional DC-DC converter, are input variables, representing environmental parameters and power state parameters; is the regression coefficient, determined by model training; is the intercept term, which represents the basic value of the control parameter of the bidirectional DC-DC converter when there is no environmental influence and power state parameter; is the error term; the purpose of model training is to estimate the regression coefficient , so that the error between the predicted value and the true value is minimized; when the model training submodule initializes the model parameters, for each feature in the linear regression model Corresponding parameters Assign an initial value, which is set to a random number close to 0 and is also the intercept term An initial value is also given; these initial parameter values will serve as the starting point for model training and will be continuously optimized and adjusted through subsequent training processes;
[0091] The model training submodule uses the data in the training set to continuously iterate and update the parameters of the linear regression model through the gradient descent algorithm to minimize the loss function value between the true value and the predicted value; in each iteration, the value of the model parameter is adjusted according to the gradient information of the loss function to the model parameter until the loss function converges or the preset number of iterations is reached.
[0092] The workflow of the model training submodule is as follows:
[0093] S1. Initialize model parameters: Before training begins, the model training submodule initializes each model parameter (including the intercept term and the regression coefficient ) is assigned a random initial value close to 0. These initial values will serve as the starting point for the gradient descent algorithm.
[0094] S2. Calculate predicted values: In each iteration, the model training submodule uses the current model parameters and the input data in the training set to calculate the predicted values according to the linear regression formula.
[0095] S3. Calculate the loss function value: Next, the model training submodule calculates the loss function value between the true value and the predicted value. In linear regression, the commonly used loss function is the mean squared error (MSE), and its calculation formula is:
[0096]
[0097] in, is the number of samples in the training set, It is The true value of the samples, The model is The predicted value of a sample.
[0098] S4. Calculate gradients and update parameters: The model training submodule calculates the gradient of the loss function with respect to the current model parameters. The gradient indicates the direction and magnitude of the model parameters that should be adjusted in order to reduce the value of the loss function. Based on this gradient information, the model training submodule uses the gradient descent algorithm to update the value of the model parameters. Specifically, for each parameter (including the intercept term ), and its update formula is as follows:
[0099]
[0100] in, is the learning rate, which controls the step size of parameter update; is the loss function over parameters The partial derivative of , that is, the gradient.
[0101] S5, Iterative Optimization: The model training submodule will repeat the above steps and continuously iteratively update the model parameters until the loss function converges to a stable minimum value or reaches a preset number of iterations. Through this process, the model training submodule can gradually optimize the parameters of the linear regression model and improve the prediction accuracy of the model.
[0102] The model training submodule uses the data in the test set to calculate the error index between the model prediction value and the actual value, and comprehensively evaluates the prediction accuracy, stability and generalization ability of the model through the error index, so as to determine the performance of the linear regression model in the prediction task. The error index includes Mean Squared Error, MSE, Mean Absolute Error (MAE) and Coefficient of Determination R² (Coefficient of Determination).
[0103] Mean square error is a common metric for measuring the difference between the model's predictions and the actual observed values. It calculates the average of the squares of the differences between the predicted values and the actual values. The larger the MSE, the worse the model's predictions, because large squared errors are magnified, highlighting inaccurate predictions. Conversely, the smaller the MSE, the closer the model's predictions are to the actual values, and the better the predictions. MSE is very sensitive to large errors, so it can effectively reflect the model's prediction performance in extreme cases.
[0104] In Example 3, the data cleaning module receives the raw data collected by the data perception layer, sets a reasonable value range for each type of parameter, traverses the raw data, marks the values that exceed the set range as abnormal values, and removes the data marked as abnormal values from the raw data. For each type of parameter, a reasonable value range is set, which can be set based on historical data, industry standards or expert advice. For example, for temperature sensor data, a reasonable range may be -50°C to 100°C. The specific steps for abnormal value processing are as follows:
[0105] S1. Traverse original data: traverse the received original data.
[0106] S2. Check the value range: For each data point, check whether its value is within the set reasonable range.
[0107] S3. Mark outliers: If the value of a data point exceeds the set range, it is marked as an outlier.
[0108] S4. Remove outliers: Remove all data points marked as outliers from the original dataset. This step can be done by creating a new dataset that only contains data points that are not marked as outliers.
[0109] The data cleaning module sorts the data after cleaning the outliers, and traverses the sorted data to compare whether the adjacent data are exactly the same; if the adjacent data are exactly the same, they are marked as duplicate data; and the data marked as duplicate is removed from the data. The specific steps of duplicate data processing are as follows:
[0110] S1. Sort the cleaned data: After removing the outliers, sort the remaining data. Sorting can be done based on the specific type and requirements of the data, such as sorting by timestamp, numerical value, etc.
[0111] S2. Traverse the sorted data: traverse the sorted data.
[0112] S3. Compare adjacent data: During the traversal process, compare whether the current data point is exactly the same as its adjacent data points (including all relevant fields).
[0113] S4. Mark duplicate data: If adjacent data points are found to be exactly the same, these data points are marked as duplicate data.
[0114] S5. Remove duplicate data: Remove all data points marked as duplicates from the dataset, keeping only unique data points. Similar to removing outliers, this can be done by creating a new dataset.
[0115] The data cleaning module can ensure that the data entering the subsequent processing flow is accurate, valid and free of redundancy, thereby improving the performance and accuracy of the entire data analysis or machine learning system.
[0116] A method for using an intelligent bidirectional power supply stable output control system, the method comprising:
[0117] S1: The data perception layer monitors and collects the environmental parameters of the wind power generation system and the working status parameters of the bidirectional DC-DC converter in real time; the environmental parameters include wind speed, wind direction and temperature, and the working status parameters include input voltage, output voltage, input current, output current, power and switch status; it provides a real-time and accurate data basis for subsequent data processing, model training and prediction, and helps the system better understand the current environment and equipment status.
[0118] S2: The data preprocessing layer receives the raw data from the data perception layer, and cleans the data through the data cleaning module to remove outliers and duplicate data, improve the quality and accuracy of the data, and ensure that the input data of the subsequent machine learning model is reliable and valid, thereby improving the prediction performance of the model.
[0119] S3: The cleaned data is transmitted to the intelligent decision-making layer, where the machine learning module uses this data for real-time prediction; the machine learning module first uses the model training submodule to train the model using historical data and optimize the model parameters. Then the model prediction submodule makes predictions based on real-time data, automatically optimizing and adjusting the control parameters of the bidirectional DC-DC converter to adapt to the ever-changing working environment.
[0120] S4: The decision optimization module generates the final control instructions based on the prediction results of the machine learning module and comprehensively considers the actual needs and safety constraints of the power system;
[0121] S5: The control execution layer receives the control instructions from the decision optimization module and parses the instructions into specific control signals through the control instruction parsing module;
[0122] S6: The bidirectional DC-DC converter control module adjusts the converter's operating parameters according to the received control signal, including the buck control parameters in Buck mode and the boost control parameters in Boost mode, to ensure that the power supply can achieve stable and efficient output in various complex and changeable wind power generation environments.
[0123] 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.
[0124] 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 bidirectional power supply stable output control system, characterized in that: It includes data perception layer, data processing layer, intelligent decision-making layer and control execution layer; The data perception layer includes an environment perception module and a power state perception module; the environment perception module is used to monitor the environmental parameters of the wind power generation system in real time, and send the environmental parameters to the data preprocessing layer; the power state perception module is used to monitor the working state parameters of the bidirectional DC-DC converter in real time, and send the working state parameters to the data preprocessing layer; The data preprocessing layer includes a data cleaning module, which is responsible for cleaning the raw data collected by the data perception layer, removing outliers and duplicate data, and sending the cleaned data to the intelligent decision-making layer; The intelligent decision-making layer includes a machine learning module and a decision optimization module; the machine learning module is used to use the received data to make real-time predictions, automatically optimize and adjust the control parameters of the bidirectional DC-DC converter, and the machine learning module includes a model training submodule and a model prediction submodule, wherein the model training submodule uses historical data for training, and the model prediction submodule makes predictions based on real-time data; The decision optimization module is used to generate the final control instruction according to the prediction results of the machine learning module, combined with the actual needs and safety constraints of the power supply system, and send the control instruction to the control execution layer; The control execution layer includes a control instruction parsing module and a bidirectional DC-DC converter control module; the control instruction parsing module is used to receive and parse the control instruction, and convert the parsed control instruction into a specific control signal; the bidirectional DC-DC converter control module is used to adjust the working parameters of the bidirectional DC-DC converter according to the received control signal to achieve stable output of the power supply, and the bidirectional DC-DC converter control module includes a Buck mode control submodule and a Boost mode control submodule, which are used to adjust the control parameters in the buck mode and the boost mode respectively; The model training submodule constructs and trains a parameter adjustment prediction model, and the specific steps include: Step 1: Collecting the operating data of the bidirectional DC-DC converter in the wind power generation system, including environmental parameters, working state parameters and corresponding control parameters; Step 2: Preprocess the collected data, including removing outliers, duplicate data or invalid data through data cleaning, eliminating the impact of dimensional differences on model training through data standardization and normalization, and forming a regular data set; divide the processed data set into training set, validation set and test set, where the training set is used for model training, the validation set is used for model effect verification and optimization, and the test set is used to verify the generalization ability of the model on unknown data sets; Step 3: Select a suitable machine learning model that can predict and optimize the control parameters of the bidirectional DC-DC converter based on the input environmental parameters and working state parameters; Step 4: Set initial parameters for the selected prediction model and prepare for training; Step 5: Use the training set to train the parameter adjustment prediction model, and continuously adjust the model parameters through the optimization algorithm to minimize the gap between the predicted control parameters and the actual control parameters; use the validation set to verify the trained model, and adjust the model parameters or structure according to the verification results; use the test set to comprehensively evaluate the optimized model through the set evaluation indicators to ensure that the model performance meets the expected standards; Step 6: Save the trained and verified parameter adjustment prediction model for subsequent real-time prediction and optimization of control parameters of the bidirectional DC-DC converter in the intelligent decision-making layer; Select the linear regression model to build a parameter adjustment prediction model. The algorithm formula of the linear regression model is as follows: Where y is the target variable, representing the control parameters of the bidirectional DC-DC converter, are input variables, representing environmental parameters and power state parameters; is the regression coefficient, determined by model training; is the intercept term, which represents the basic value of the control parameter of the bidirectional DC-DC converter when there is no environmental influence and power state parameter; is the error term; the purpose of model training is to estimate the regression coefficient , so that the error between the predicted value and the true value is minimized; when the model training submodule initializes the model parameters, for each feature in the linear regression model Corresponding parameters Assign an initial value, which is set to a random number close to 0 and is also the intercept term An initial value is also given; these initial parameter values will serve as the starting point for model training and will be continuously optimized and adjusted through subsequent training processes; The model training submodule uses the data in the training set to continuously iterate and update the parameters of the linear regression model through the gradient descent algorithm to minimize the loss function value between the true value and the predicted value; in each iteration, the value of the model parameter is adjusted according to the gradient information of the loss function to the model parameter until the loss function converges or the preset number of iterations is reached.
2. According to claim 1, the intelligent bidirectional power supply stable output control system is characterized by: The environmental parameters monitored in real time by the environmental perception module include wind speed, wind direction and temperature data; The working state parameters monitored in real time by the power state sensing module include working mode, input voltage, output voltage, input current, output current, power and switch state; wherein the input voltage refers to the DC power supply voltage connected to the converter, and the output voltage is the DC voltage output by the converter; the input current is the current flowing into the converter, and the output current is the current flowing out of the converter; the power is the ratio of the electric energy input or output of the converter to time; The switching state refers to the on and off state of the switching elements in the converter.
3. The intelligent bidirectional power supply stable output control system according to claim 2 is characterized in that: The operating parameters adjusted by the bidirectional DC-DC converter include switching frequency, duty cycle, inductor current limit, current loop bandwidth, voltage loop bandwidth, synchronous rectification control and phase shift control.
4. The intelligent bidirectional power supply stable output control system according to claim 3 is characterized in that: The model training submodule uses the data in the test set to calculate the error index between the model prediction value and the actual value; and comprehensively evaluates the prediction accuracy, stability and generalization ability of the model through the error index, so as to determine the performance of the linear regression model in the prediction task.
5. The intelligent bidirectional power supply stable output control system according to claim 4 is characterized in that: The error indicators include mean square error, mean absolute error and coefficient of determination R².
6. The intelligent bidirectional power supply stable output control system according to claim 5, characterized in that: The data cleaning module receives the raw data collected by the data perception layer, sets a reasonable value range for each type of parameter, traverses the raw data, marks the values that exceed the set range as outliers, and removes the data marked as outliers from the raw data.
7. The intelligent bidirectional power supply stable output control system according to claim 6, characterized in that: The data cleaning module sorts the data after cleaning the outliers, and traverses the sorted data to compare whether adjacent data are completely identical; if the adjacent data are completely identical, they are marked as duplicate data; and the data marked as duplicate is removed from the data.
8. The method for using the intelligent bidirectional power supply stable output control system according to claim 1 is characterized in that: The method of use includes: S1: Real-time monitoring and collection of environmental parameters of the wind power generation system and working state parameters of the bidirectional DC-DC converter through the data perception layer; the environmental parameters include wind speed, wind direction and temperature, and the working state parameters include input voltage, output voltage, input current, output current, power and switch state; S2: The data preprocessing layer receives the raw data from the data perception layer and cleans the data through the data cleaning module to remove outliers and duplicate data to ensure the validity and accuracy of the data; S3: The cleaned data is transmitted to the intelligent decision-making layer, where the machine learning module uses the data to make real-time predictions. The machine learning module first uses the model training submodule to train the model using historical data and optimize the model parameters. Then the model prediction submodule makes predictions based on the real-time data and automatically optimizes and adjusts the control parameters of the bidirectional DC-DC converter. S4: The decision optimization module generates the final control instructions based on the prediction results of the machine learning module and comprehensively considers the actual needs and safety constraints of the power system; S5: The control execution layer receives the control instructions from the decision optimization module and parses the instructions into specific control signals through the control instruction parsing module; S6: The bidirectional DC-DC converter control module adjusts the operating parameters of the converter according to the received control signal, including the step-down control parameters in the Buck mode and the step-up control parameters in the Boost mode.
