Intelligent control adjusting system applied to mobile transformer
Through the intelligent control and regulation system, using predicted load model and dynamic planning, the problem of excessive load of mobile transformers during large-scale activities is solved, and efficient utilization and stable supply of power resources are achieved.
Patent Information
- Application Number
- CN202411407124.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2024-11-08
AI Technical Summary
Traditional mobile transformers have a sharp increase in power consumption during large-scale activities, resulting in excessive load, waste of power resources and insufficient power supply, and unable to provide an effective regulation strategy.
An intelligent control and regulation system is designed, including a collection processing unit, a prediction optimization unit and a generation regulation unit. Through sensors, a predictive load model is established, a model parameter is optimized, a regulation strategy is generated, and a dynamic programming is used to reduce the calculation amount. , improve response speed and computing efficiency.
It realizes rapid load prediction and regulation of mobile transformers under high load states, reduces waste of power resources, improves computing efficiency and response speed, and ensures the stability of power supply.
Smart Images

Figure CN118915473A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent control and regulation, and in particular to an intelligent control and regulation system applied to a mobile transformer. Background Art
[0002] Intelligent control and regulation is an important technology that can monitor and adjust system parameters in real time, keep the system in the best condition, improve stability and accuracy, and use advanced algorithms and models to conduct complex data analysis, optimize decision-making, and reduce energy consumption, improve resource utilization, and reduce the impact on the environment by optimizing control strategies.
[0003] With the vigorous development of social economy in recent years, some large-scale commercial activities have also increased. The holding of large-scale activities cannot be separated from the support of electricity, and the electricity consumption of large-scale activities is very large within a certain period of time. In order to ensure that large-scale activities can be held perfectly, transformers are needed to regulate electricity. Traditional transformers may be too far away from the venue of large-scale activities, resulting in insufficient power supply. Therefore, mobile transformers are needed to complete the power supply task. However, due to the sudden increase in electricity consumption on the day of large-scale activities, the load of traditional mobile transformers will be too high, resulting in excessive power consumption of mobile transformers when powering certain electrical equipment or insufficient power supply to certain electrical equipment, resulting in a waste of power resources and no adjustment strategy. In order to solve this technical problem, we provide an intelligent control and adjustment system for mobile transformers. Summary of the invention
[0004] The object of the present invention is to provide an intelligent control and regulation system for a mobile transformer to solve the problems raised in the above background technology.
[0005] To achieve the above-mentioned purpose, an intelligent control and regulation system for mobile transformers is provided, comprising an acquisition and processing unit, a prediction and optimization unit, and a generation and regulation unit; The acquisition processing unit acquires the working current and working voltage of the mobile transformer through the sensor, and pre-processes the working current and working voltage of the mobile transformer to obtain the workload and workload data set of the mobile transformer; The prediction optimization unit is used to establish a prediction load model for the mobile transformer, and extract the initial model parameters of the prediction load model from the workload data set of the mobile transformer, and then input the initial model parameters into the hidden layer of the prediction load model of the mobile transformer for calculation, and obtain the output value of the initial model parameters in the hidden layer of the prediction load model of the mobile transformer, and then substitute the output value of the model parameters in the hidden layer of the prediction load model of the mobile transformer into the forward propagation algorithm to calculate the initial prediction load of the mobile transformer, and then define a loss function according to the mean of the square difference between the initial prediction load and the actual load of the mobile transformer, and update the initial model parameters according to the loss function to obtain the final model parameters, and finally adjust the prediction load model of the mobile transformer according to the final model parameters to obtain the prediction load of the mobile transformer; The generating and regulating unit dynamically plans the predicted load of the mobile transformer to generate a regulating strategy for the mobile transformer, and sends the regulating strategy for the mobile transformer to the executing mechanism, which issues a regulating instruction according to the regulating strategy for the mobile transformer to regulate the mobile transformer.
[0006] As a further improvement of the technical solution, the acquisition processing unit includes a sensor module and a processing generation module, the sensor module is used to collect the working current and working voltage of the mobile transformer, the processing generation module is used to pre-process the working current and working voltage of the mobile transformer, and obtain the workload and workload data set of the mobile transformer, the method of pre-processing the working current and working voltage of the mobile transformer specifically includes: The threshold range of the working current and working voltage of the mobile transformer is set, and the working current and working voltage of the mobile transformer that are not within the threshold range are eliminated. The working current and working voltage of the mobile transformer are denoised and calculated using the median filtering algorithm to obtain the workload of the mobile transformer. The workload of the mobile transformer is then sorted according to the time series to form a workload data set of the mobile transformer.
[0007] As a further improvement of the technical solution, the method of using the median filtering algorithm to perform denoising and calculation on the working current and working voltage of the mobile transformer to obtain the working load of the mobile transformer specifically includes: The working current and working voltage of the mobile transformer are sorted in chronological order to obtain the working current data sequence and working voltage data sequence of the mobile transformer, that is, and , Expressed as the working current of the mobile transformer, Expressed as The working current of the mobile transformer at a time point is It is the working voltage of the mobile transformer. Expressed as The working voltage of the mobile transformer at a time point; Extracting data points from the operating current data series of a mobile transformer , , set the window size to , for each data point Take the m data points around it, that is , and sort these m data points in ascending order to get the middle value , and recorded as the current output value after completing the median filter, recorded as , , and then perform the above operation on the mobile transformer working voltage data sequence to obtain , and and Substituting into the formula for calculation, we can get the working load of the mobile transformer: ; Said Expressed as the working load of the mobile transformer.
[0008] As a further improvement of the technical solution, the prediction optimization unit includes a model building module, and the model building module is used to build a prediction load model of the mobile transformer, specifically including: The workload dataset of the mobile transformer is divided into two groups with an average size of A subset of , and a long short-term memory network is selected as the architecture of the mobile transformer load prediction model, the architecture of the mobile transformer load prediction model includes an input layer, Hidden layer, A hidden layer and an output layer, wherein the input layer is used to receive the workload data of the mobile transformer, The hidden layer is used to train the load prediction model of the mobile transformer using the load data in the workload dataset subset to obtain initial model parameters. The hidden layer is used to calculate the initial predicted load of the mobile transformer through the forward propagation algorithm during the training process, and to update the initial model parameters of the mobile transformer predicted load model through the back propagation algorithm to obtain the final model parameters. Finally, the initial predicted load of the predicted load model is adjusted according to the final model parameters to obtain the predicted load of the mobile transformer. The output layer is used to output the predicted load of the mobile transformer.
[0009] As a further improvement of this technical solution, the The hidden layer uses a subset of the workload dataset to train the predictive load model for mobile transformers, including: The initial model parameters of the mobile transformer load prediction model are extracted from the workload dataset subset and are denoted as , Including the weight W and bias b of the predicted load model, a subset is randomly selected from the workload data set, and the size of the workload data set subset is , denoted as , randomly select the i-th sample from the workload dataset subset, denoted as ,but The training output value of the hidden layer is: ; ; in, for The output value of the hidden layer, is the output value after activation, For the input layer to The weight matrix of the hidden layer, For the input layer to The bias vector of the hidden layer, for Activation function for the hidden layer.
[0010] As a further improvement of the technical solution, the method of calculating the initial predicted load of the mobile transformer by the forward propagation algorithm and updating the initial model parameters of the mobile transformer predicted load model by the back propagation algorithm specifically includes: Will The output value of the hidden layer is input to In the hidden layer, The hidden layer calculates the initial predicted load of the mobile transformer through the forward propagation algorithm, and the calculation formula of the forward propagation algorithm is: ; ; in, for The output value of the hidden layer, for The output value after the hidden layer is activated, for Hidden layer to The weight matrix of the hidden layer, for The bias vector of the hidden layer, for The activation function of the hidden layer, the initial predicted load output value is: ; in, is the initial predicted load of the mobile transformer, for The weight matrix from the hidden layer to the output layer, is the bias vector of the output layer; The initial predicted load of the mobile transformer is compared with the actual load to obtain the mean square difference between the initial predicted load and the actual load of the mobile transformer. A loss function is defined by the mean square difference. The model parameter gradient of the mobile transformer load prediction model is calculated according to the loss function. Finally, the initial model parameters are updated according to the model parameter gradient.
[0011] As a further improvement of the technical solution, the prediction optimization unit includes a parameter updating module, the prediction optimization unit includes a parameter updating module, the parameter updating module is used to calculate the model parameter gradient of the mobile transformer prediction load model according to the loss function, and update the initial model parameters according to the model parameter gradient, and the specific method includes: Extract the actual load for a period of time ,in , by calculating the mean of the squared difference between the initial predicted load and the actual load, we get the mean square error, and let the mean square error be the loss function between the predicted load and the actual load , the loss function The calculation method is: ; in, is the workload dataset subset size; The mathematical expression of the initial predicted load model of the mobile transformer is recorded as: ; The loss function Substitute the formula to calculate the model parameter gradient of the mobile transformer load prediction model, namely: ; ; in, Model parameters for predicting load models for mobile transformers The gradient of Model parameters for predicting load models for mobile transformers The gradient of the model parameters is then updated according to the model parameter gradient, that is: ; ; in, Denote as the learning rate, and Substituting into the formula, we get the predicted load of the mobile transformer: ; in, is the activation function of the output layer.
[0012] As a further improvement of the technical solution, the generation and adjustment unit generates a mobile transformer adjustment strategy according to the predicted load of the mobile transformer, specifically including: Setting the load threshold , when the predicted load of the mobile transformer Exceeding the set load threshold When the value reaches 5%, the strategy calculation is triggered. The specific calculation steps are as follows: H1. Assume that the state variable of the mobile transformer is ,in, represents the predicted load of the mobile transformer at time t, represents the voltage level of the mobile transformer at time t; H2. Set the adjustment position of the mobile transformer tap to The number of capacitor banks switched on and off is recorded as , It represents the adjustment position of the mobile transformer at time t. It is expressed as the number of capacitor banks switched on and off at time t of the mobile transformer; H3. When taking a decision When adjusting the mobile transformer, the predicted load and voltage of the mobile transformer can be expressed as: ; ; in, is the predicted load of the mobile transformer after adjustment, It is expressed as the voltage of the mobile transformer after adjustment. Indicates that a decision is taken. and It is expressed as a function determined by physical laws and device characteristics; H4. Set an objective function, which is a comprehensive indicator of voltage fluctuation and minimum energy consumption, and is expressed by the formula: ; in, is the objective function, Expressed as voltage fluctuation, is the rated voltage, Indicates decision making The cost, is a weight parameter that weighs cost and voltage fluctuation, is the total number of time stages; H5. Calculating forward from the last stage, the recursive equation is: ; in, It is represented as the objective function value of each decision, and the objective function value of each decision is compared. The regulation strategy corresponding to the objective function of voltage fluctuation and minimizing the energy consumption is used as the final regulation strategy and sent to the actuator. The actuator issues an adjustment instruction according to the final regulation strategy to adjust the mobile transformer.
[0013] Compared with the prior art, the present invention has the following beneficial effects: In the intelligent control and regulation system applied to mobile transformers, the workload of the mobile transformer is obtained according to the current and voltage of the mobile transformer, and a predictive load model is established based on the workload of the mobile transformer. The predictive load model is optimized by optimizing the random gradient descent algorithm, which speeds up the update of model parameters and improves the calculation efficiency, so that the mobile transformer can still quickly predict the load under high load conditions. The forward propagation algorithm is introduced to quickly process the input data and calculate the predicted load of the mobile transformer, which speeds up the response time. The load and voltage of the mobile transformer are predicted by dynamic programming at each stage to reduce the amount of calculation. Finally, the reverse recursive algorithm is used to obtain the optimal decision in the initial stage, which improves the calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 It is an overall block diagram of the present invention.
[0015] The meaning of each number in the figure is: 1. Acquisition and processing unit; 11. Sensor module; 12. Processing and generation module; 2. Prediction and optimization unit; 21. Model building module; 22. Parameter updating module; 3. Generation and adjustment unit. DETAILED DESCRIPTION
[0016] 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.
[0017] Example The present invention provides an intelligent control and regulation system for mobile transformers. Figure 1 As shown, it includes a collection and processing unit 1, a prediction and optimization unit 2, and a generation and adjustment unit 3; The collection and processing unit 1 collects the working current and working voltage of the mobile transformer through sensors, and pre-processes the working current and working voltage of the mobile transformer to obtain the workload of the mobile transformer and generate a workload data set of the mobile transformer.
[0018] In order to realize comprehensive monitoring and analysis of the working state of the mobile transformer, it is necessary to accurately obtain its working current and working voltage, so a sensor module 11 is set to collect the working current and working voltage of the mobile transformer, so as to accurately obtain the working load of the transformer, which is helpful to optimize the operation and maintenance strategy of the transformer. Since the original collected current and voltage data may contain noise, it is necessary to pre-process to extract useful information, so a processing generation module 12 is set to pre-process the working current and working voltage of the mobile transformer, so as to obtain the working load and workload data set of the mobile transformer; The method of pre-processing the working current and working voltage of the mobile transformer specifically includes: Set the threshold range of the working current and working voltage of the mobile transformer, and eliminate the working current and working voltage of the mobile transformer that are not within the threshold range. These data may be caused by measurement errors, equipment failures or other abnormal conditions and should not be used for subsequent analysis and model training. The noise in the working current and working voltage will interfere with the judgment of the actual load conditions. De-noising processing can more accurately reflect the actual workload, and the median filtering algorithm is simple and efficient. It can quickly calculate and obtain the workload of the mobile transformer. The workload of the mobile transformer is then sorted according to the time series to form a workload data set for the mobile transformer, which helps to discover the law of load changes over time and provide a more organized data structure for subsequent analysis and prediction.
[0019] The working current and working voltage of the mobile transformer are de-noised and calculated by using the median filtering algorithm to obtain the working load of the mobile transformer, specifically including: The working current and working voltage of the mobile transformer are sorted in chronological order to obtain the working current data sequence and working voltage data sequence of the mobile transformer, that is, and , Expressed as the working current of the mobile transformer, Expressed as The working current of the mobile transformer at a time point is: Expressed as the working voltage of the mobile transformer, Expressed as The working voltage of the mobile transformer at a time point; Extracting data points from the operating current data series of a mobile transformer , , set the window size to , for each data point Take its surrounding data points, i.e. , and for this Sort the data points from small to large and get the middle value , and recorded as the current output value after completing the median filter, recorded as , , and then perform the above operation on the mobile transformer working voltage data sequence to obtain , and and Substituting into the formula for calculation, we can get the working load of the mobile transformer: ; Expressed as the working load of the mobile transformer.
[0020] The prediction optimization unit is used to establish a prediction load model for a mobile transformer, and extract the initial model parameters of the prediction load model from the workload data set of the mobile transformer, and then input the initial model parameters into the hidden layer of the prediction load model for calculation, and obtain the output value of the initial model parameters in the hidden layer of the prediction load model for the mobile transformer, and then substitute the output value of the model parameters in the hidden layer of the prediction load model for the forward propagation algorithm to calculate the initial prediction load of the mobile transformer, and then define a loss function according to the mean of the square difference between the initial prediction load and the actual load of the mobile transformer, and update the initial model parameters according to the loss function to obtain the final model parameters, and finally adjust the prediction load model of the mobile transformer according to the final model parameters to obtain the predicted load of the mobile transformer, and after multiple parameter updates and adjustments, the predicted value of the prediction load model is closer to the actual value, providing a data basis for subsequent strategy making.
[0021] The prediction optimization unit 2 includes a model building module 21, which is used to build a prediction load model of the mobile transformer, specifically including: The workload dataset of the mobile transformer is divided into two groups with an average size of The subset of can effectively reduce the amount of model calculations and improve work efficiency. The long short-term memory network is selected as the architecture of the mobile transformer load prediction model. The long short-term memory network is good at processing time series data with long-term dependencies, such as the workload data of the mobile transformer. It can remember and use early data information, avoiding the gradient vanishing and gradient exploding problems of traditional neural networks when processing long sequences. The architecture of the mobile transformer load prediction model includes the input layer, Hidden layer, Hidden layer and output layer, the input layer is used to receive the workload data of the mobile transformer, The hidden layer is used to train the predicted load model of the mobile transformer using the load data in the workload dataset subset to obtain the initial model parameters. The hidden layer is used in the training process. The forward propagation algorithm can quickly calculate the prediction results. Therefore, by calculating the initial predicted load of the mobile transformer through the forward propagation algorithm, a preliminary estimate of the transformer load can be quickly generated, providing timely reference for subsequent analysis and decision-making. The back propagation algorithm effectively adjusts the model parameters according to the prediction error. Updating the initial model parameters of the mobile transformer prediction load model through the back propagation algorithm helps to gradually optimize the model so that it can more accurately predict the transformer load. Finally, the prediction load model is adjusted according to the model parameters to obtain the predicted load of the mobile transformer. The output layer is used to output the predicted load of the mobile transformer.
[0022] The hidden layer uses a subset of the workload dataset to train the predictive load model for mobile transformers, including: Weights and biases determine the relationship between the input and output of the mobile transformer load prediction model. They play the role of regulating and converting input signals in the long short-term memory network, and directly affect the data processing and prediction results of the mobile transformer load prediction model. By adjusting the weights and biases, the mobile transformer load prediction model can learn patterns and rules from the training data.
[0023] The initial model parameters of the mobile transformer load prediction model are extracted from the workload dataset subset and are denoted as to ensure data quality and consistency, Including the weight W and bias b of the predicted load model, a subset is randomly selected from the workload dataset. The size of the workload dataset subset is , denoted as , randomly select the i-th sample from the workload dataset subset, denoted as ,but The training output value of the hidden layer is: ; ; in, for The output value of the hidden layer, is the output value after activation, For the input layer to The weight matrix of the hidden layer, For the input layer to The bias vector of the hidden layer, for Activation function for the hidden layer.
[0024] The method of calculating the initial predicted load of the mobile transformer by a forward propagation algorithm and updating the initial model parameters of the mobile transformer predicted load model by a back propagation algorithm specifically includes: Will The output value of the hidden layer is input to In the hidden layer, The hidden layer calculates the initial predicted load of the mobile transformer through the forward propagation algorithm. The calculation formula of the forward propagation algorithm is: ; ; in, for The output value of the hidden layer, for The output value after the hidden layer is activated, for Hidden layer to The weight matrix of the hidden layer, for The bias vector of the hidden layer, for The activation function of the hidden layer, the initial predicted load output value is: ; in, is the initial predicted load of the mobile transformer, for The weight matrix from the hidden layer to the output layer, is the bias vector of the output layer; The initial predicted load of the mobile transformer is compared with the actual load to obtain the mean square difference between the initial predicted load and the actual load of the mobile transformer. A loss function is defined by the mean square difference. The model parameter gradient of the mobile transformer load prediction model is calculated according to the loss function. Finally, the initial model parameters are updated according to the model parameter gradient.
[0025] The prediction optimization unit 2 includes a parameter updating module 22, which is used to calculate the model parameter gradient of the mobile transformer prediction load model according to the loss function, and update the model parameters according to the model parameter gradient, and the specific method includes: Extract the actual load for a period of time ,in , by calculating the mean of the squared difference between the initial predicted load and the actual load, we get the mean square error, and let the mean square error be the loss function between the predicted load and the actual load , loss function The calculation method is: ; in, is the workload dataset subset size; The mathematical expression of the initial predicted load model of the mobile transformer is recorded as: ; The loss function Substitute the formula to calculate the model parameter gradient of the mobile transformer load prediction model, namely: ; ; in, Model parameters for predicting load models for mobile transformers The gradient of Model parameters for predicting load models for mobile transformers The gradient of the model parameters is then updated according to the model parameter gradient, that is: ; ; in, Denote as the learning rate, and Substituting into the formula, we get the predicted load of the mobile transformer: ; in, is the activation function of the output layer.
[0026] The generating and regulating unit 3 performs dynamic planning according to the predicted load of the mobile transformer, generates a regulating strategy for the mobile transformer, and sends the regulating strategy to the executing agency. The executing agency issues a regulating instruction according to the regulating strategy to regulate the mobile transformer.
[0027] The generation and adjustment unit 3 generates a mobile transformer adjustment strategy according to the predicted load of the mobile transformer, specifically including: Setting the load threshold , when the predicted load of the mobile transformer Exceeding the set load threshold When the value reaches 5%, the strategy calculation is triggered. The specific calculation steps are as follows: H1. Assume that the state variable of the mobile transformer is , among which, among which, represents the predicted load of the mobile transformer at time t, It represents the voltage level of the mobile transformer at time t. By defining the state variables of the mobile transformer, the key characteristics of the system at each time point are clarified, and the state of the system at a specific time is fully described, covering the load and voltage, two factors that are crucial to power quality and regulation strategies. It provides a clear basis for subsequent analysis and decision-making, and can accurately track and evaluate the state changes of the system; H2. Set the adjustment position of the mobile transformer tap to The number of capacitor banks switched on and off is recorded as , It is expressed as the adjustment position of the mobile transformer at time t, It is expressed as the number of capacitor banks switched on and off at time t of the mobile transformer; H3. When taking a decision When adjusting the mobile transformer, the predicted load and voltage of the mobile transformer can be expressed as: ; ; in, is the predicted load of the mobile transformer after adjustment, It is expressed as the voltage of the mobile transformer after adjustment. Indicates that a decision is taken. and Expressed as functions determined according to physical laws and equipment characteristics, the above two stage transfer equations describe the state evolution law of the system from the current stage to the next stage. Using them is based on physical laws and equipment characteristics, and can accurately predict the load and voltage changes after taking a specific decision, which improves the scientificity and accuracy of decision-making and helps to achieve long-term optimization goals; H4. In order to balance the two important goals of optimizing voltage quality and controlling costs, an objective function is set. The objective function is a comprehensive indicator of voltage fluctuation and optimized energy consumption, which is expressed by the formula: ; in, is the objective function, Expressed as voltage fluctuation, is the rated voltage, Indicates decision making The cost, is a weight parameter that weighs cost and voltage fluctuation, is the total number of time stages. This objective function comprehensively considers voltage fluctuations and decision costs, and can comprehensively evaluate the comprehensive effects of different decision-making schemes, avoid one-sided decisions caused by focusing on a single factor, and find the optimal strategy to minimize the total cost while ensuring voltage quality; H5. Calculating forward from the last stage, the recursive equation is: ; in, It is expressed as the objective function value of each decision, and the objective function value of each decision is compared. The regulation strategy corresponding to the objective function of voltage fluctuation and minimizing energy consumption is used as the final regulation strategy. The actuator issues an adjustment instruction to adjust the mobile transformer according to the final regulation strategy. This dynamic programming recursive equation calculates the optimal value from the last stage forward, avoids repeated calculations, improves calculation efficiency, and can ensure that the global optimal solution is obtained. In complex multi-stage decision-making problems, the optimal regulation strategy can be found in an efficient and accurate manner.
[0028] The present invention acquires the working load of the mobile transformer by collecting the voltage and current of the mobile transformer, builds a predicted load model of the mobile transformer according to the working load of the mobile transformer, calculates the initial predicted load of the mobile transformer by substituting the output value of the initial model parameter in the hidden layer of the predicted load model of the mobile transformer into the forward propagation algorithm, thereby improving the calculation efficiency, and then updates the initial model parameters according to the loss function between the initial predicted load and the actual load to obtain the model parameters, and finally adjusts the predicted load model of the mobile transformer according to the model parameters to obtain the predicted load of the mobile transformer, and finally performs dynamic programming according to the predicted load of the mobile transformer to generate an adjustment strategy for the mobile transformer.
[0029] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.
Claims
1. An intelligent control and regulation system for mobile transformers, characterized in that: It includes a collection and processing unit (1), a prediction and optimization unit (2), and a generation and adjustment unit (3); The acquisition processing unit (1) acquires the working current and working voltage of the mobile transformer through a sensor, and pre-processes the working current and working voltage of the mobile transformer to obtain the working load and working load data set of the mobile transformer; The prediction optimization unit (2) is used to establish a prediction load model for the mobile transformer, extract initial model parameters of the prediction load model from the workload data set of the mobile transformer, input the initial model parameters into the hidden layer of the prediction load model for calculation, obtain the output value of the initial model parameters in the hidden layer of the prediction load model for the mobile transformer, substitute the output value of the model parameters in the hidden layer of the prediction load model for the mobile transformer into the forward propagation algorithm, calculate the initial prediction load of the mobile transformer, define a loss function according to the mean square difference between the initial prediction load and the actual load of the mobile transformer, update the initial model parameters according to the loss function, obtain the final model parameters, and finally adjust the prediction load model for the mobile transformer according to the final model parameters to obtain the prediction load of the mobile transformer; The generating and regulating unit (3) performs dynamic planning on the predicted load of the mobile transformer to generate a regulating strategy for the mobile transformer, and sends the regulating strategy for the mobile transformer to an actuator, which issues a regulating instruction according to the regulating strategy for the mobile transformer to regulate the mobile transformer.
2. The intelligent control and regulation system for mobile transformers according to claim 1 is characterized in that: The acquisition processing unit (1) comprises a sensor module (11) and a processing generation module (12), wherein the sensor module (11) is used to acquire the working current and working voltage of the mobile transformer, and the processing generation module (12) is used to pre-process the working current and working voltage of the mobile transformer and obtain the working load and working load data set of the mobile transformer, wherein the method of pre-processing the working current and working voltage of the mobile transformer specifically comprises: The threshold range of the working current and working voltage of the mobile transformer is set, and the working current and working voltage of the mobile transformer that are not within the threshold range are eliminated. The working current and working voltage of the mobile transformer are denoised and calculated using the median filtering algorithm to obtain the workload of the mobile transformer. The workload of the mobile transformer is then sorted according to the time series to form a workload data set of the mobile transformer.
3. The intelligent control and regulation system for mobile transformers according to claim 2 is characterized in that: The method of using the median filtering algorithm to perform denoising and calculation on the working current and working voltage of the mobile transformer to obtain the working load of the mobile transformer specifically includes: The working current and working voltage of the mobile transformer are sorted in chronological order to obtain the working current data sequence and working voltage data sequence of the mobile transformer, that is, and , Expressed as the working current of the mobile transformer, Expressed as The working current of the mobile transformer at a time point is It is the working voltage of the mobile transformer. Expressed as The working voltage of the mobile transformer at a time point; Extracting data points from the operating current data series of a mobile transformer , , set the window size to , for each data point Take its surrounding data points, i.e. , and for this Sort the data points from small to large and get the middle value , and recorded as the current output value after completing the median filter, recorded as , , and then perform the above operation on the mobile transformer working voltage data sequence to obtain , and and Substituting into the formula for calculation, we can get the working load of the mobile transformer: ; Said Expressed as the working load of the mobile transformer.
4. The intelligent control and regulation system for mobile transformers according to claim 1 is characterized in that: The prediction optimization unit (2) comprises a model building module (21), wherein the model building module (21) is used to build a prediction load model of a mobile transformer, and specifically comprises: The workload dataset of the mobile transformer is divided into two groups with an average size of A subset of , and a long short-term memory network is selected as the architecture of the mobile transformer load prediction model, the architecture of the mobile transformer load prediction model includes an input layer, Hidden layer, A hidden layer and an output layer, wherein the input layer is used to receive the workload data of the mobile transformer, The hidden layer is used to train the load prediction model of the mobile transformer using the load data in the workload dataset subset to obtain initial model parameters. The hidden layer is used to calculate the initial predicted load of the mobile transformer through the forward propagation algorithm during the training process, and to update the initial model parameters of the mobile transformer predicted load model through the back propagation algorithm to obtain the final model parameters. Finally, the initial predicted load of the predicted load model is adjusted according to the final model parameters to obtain the predicted load of the mobile transformer. The output layer is used to output the predicted load of the mobile transformer.
5. The intelligent control and regulation system for mobile transformers according to claim 4 is characterized in that: Said The hidden layer uses a subset of the workload dataset to train the predictive load model for mobile transformers, including: The initial model parameters of the mobile transformer load prediction model are extracted from the workload dataset subset and are denoted as , Including the weight W and bias b of the predicted load model, a subset is randomly selected from the workload data set, and the size of the workload data set subset is , denoted as , randomly select the i-th sample from the workload dataset subset, denoted as ,but The training output value of the hidden layer is: ; ; in, for The output value of the hidden layer, is the output value after activation, For the input layer to The weight matrix of the hidden layer, For the input layer to The bias vector of the hidden layer, for Activation function for the hidden layer.
6. The intelligent control and regulation system for mobile transformers according to claim 4 is characterized in that: The method of calculating the initial predicted load of the mobile transformer by the forward propagation algorithm and updating the initial model parameters of the mobile transformer predicted load model by the back propagation algorithm specifically includes: Will The output value of the hidden layer is input to In the hidden layer, The hidden layer calculates the initial predicted load of the mobile transformer through the forward propagation algorithm, and the calculation formula of the forward propagation algorithm is: ; ; in, for The output value of the hidden layer, for The output value after the hidden layer is activated, for Hidden layer to The weight matrix of the hidden layer, for The bias vector of the hidden layer, for The activation function of the hidden layer, the initial predicted load output value is: ; in, is the initial predicted load of the mobile transformer, for The weight matrix from the hidden layer to the output layer, is the bias vector of the output layer; The initial predicted load of the mobile transformer is compared with the actual load to obtain the mean square difference between the initial predicted load and the actual load of the mobile transformer. A loss function is defined by the mean square difference. The model parameter gradient of the mobile transformer load prediction model is calculated according to the loss function. Finally, the initial model parameters are updated according to the model parameter gradient.
7. The intelligent control and regulation system for mobile transformers according to claim 6 is characterized in that: The prediction optimization unit (2) comprises a parameter updating module (22), wherein the parameter updating module (22) is used to calculate a model parameter gradient of the mobile transformer load prediction model according to a loss function, and to update the initial model parameters according to the model parameter gradient, and the specific method includes: Extract the actual load for a period of time ,in , by calculating the mean of the squared difference between the initial predicted load and the actual load, we get the mean square error, and let the mean square error be the loss function between the predicted load and the actual load , the loss function The calculation method is: ; in, is the workload dataset subset size; The mathematical expression of the initial predicted load model of the mobile transformer is recorded as: ; The loss function Substitute the formula to calculate the model parameter gradient of the mobile transformer load prediction model, namely: ; ; in, Model parameters for predicting load models for mobile transformers The gradient of Model parameters for predicting load models for mobile transformers The gradient of the model parameters is then updated according to the model parameter gradient, that is: ; ; in, Denote as the learning rate, and Substituting into the formula, we get the predicted load of the mobile transformer: ; in, is the activation function of the output layer.
8. The intelligent control and regulation system for mobile transformers according to claim 3 is characterized in that: The generation and adjustment unit (3) generates a mobile transformer adjustment strategy according to the predicted load of the mobile transformer, specifically comprising: Setting the load threshold , when the predicted load of the mobile transformer Exceeding the set load threshold When the value reaches 5%, the strategy calculation is triggered. The specific calculation steps are as follows: H1. Assume that the state variable of the mobile transformer is ,in, represents the predicted load of the mobile transformer at time t, represents the voltage level of the mobile transformer at time t; H2. Set the adjustment position of the mobile transformer tap to The number of capacitor banks switched on and off is recorded as , It represents the adjustment gear position of the mobile transformer at time t. It is expressed as the number of capacitor banks switched on and off at time t of the mobile transformer; H3. When taking a decision When adjusting the mobile transformer, the predicted load and voltage of the mobile transformer can be expressed as: ; ; in, is the predicted load of the mobile transformer after adjustment, It is expressed as the voltage of the mobile transformer after adjustment. Indicates that a decision is taken. and It is expressed as a function determined by physical laws and device characteristics; H4. Set an objective function, which is a comprehensive indicator of voltage fluctuation and minimum energy consumption, and is expressed by the formula: ; in, is the objective function, Expressed as voltage fluctuation, is the rated voltage, Indicates decision making The cost, is a weight parameter that weighs cost and voltage fluctuation, is the total number of time stages; H5. Calculating forward from the last stage, the recursive equation is: ; in, It is represented as the objective function value of each decision, and the objective function value of each decision is compared. The regulation strategy corresponding to the objective function of voltage fluctuation and minimizing the energy consumption is used as the final regulation strategy and sent to the actuator. The actuator issues an adjustment instruction according to the final regulation strategy to adjust the mobile transformer.
Citation Information
Patent Citations
A method for short-term state prediction of a large-scale road network base on an improved k-nearest neighbor
CN109190797A
Transformer load optimization control system and method based on machine learning
CN117595244A
Early warning method and system for three-phase imbalance of transformer
CN117613950A
Collaborative optimization control method for economic operation of transformer
CN118100448A
Load prediction method and apparatus based on neural network
US20210133536A1
Cited By
New energy power transmission state monitoring method and system based on artificial intelligence algorithm
CN119944951A
