A direct current power supply system based on intelligent center control

Abstract

The application discloses a direct-current power supply system based on intelligent calculation center control and relates to the technical field of power grids, which comprises a data acquisition module, which is used for collecting load training data and line decision data through an intelligent power grid background; an intelligent power grid background acquires the line decision data and generates a training four-tuple set; a model training module is used for training a first machine learning model for predicting the maximum load value through the load training data acquired by the intelligent power grid background and training a second machine learning model for deciding whether to use a replacement standby line load when an abnormal transmission state is detected based on the training four-tuple set; the application can realize the automatic isolation and recovery of the intelligent power grid by monitoring whether the topological line of the intelligent power grid is faulty in real time, deciding whether to replace the standby line to restore power supply if there is a fault, preventing the power cut range from expanding, and restoring the power supply of the non-fault area.

Description

Technical Field

[0001] The present invention relates to the field of power grid technology, and more specifically, to a direct current power supply system based on intelligent computing center control. Background Art

[0002] A smart grid distribution network refers to a complex network consisting of multiple distribution devices, lines, and terminals, responsible for transmitting electricity from the transmission grid to users. Fault diagnosis and self-healing refer to the ability to automatically detect, diagnose, and recover from a fault in the distribution network to ensure the reliability of power supply. Traditional fault diagnosis methods rely primarily on manual experience and are difficult to accurately diagnose complex faults. When a distribution fault is identified, fault self-healing methods still require manual intervention, making it difficult to quickly recover from the fault. The distribution network lacks intelligence. To address this problem, existing methods use grid monitoring datasets and historical grid monitoring datasets for each distribution area, combined with GCN and GRU networks, to quickly analyze the fault risks of the distribution area. This accelerates the discovery of fault risks and improves the diagnosis rate of distribution faults. For example, Chinese patent application number CN117498561B discloses a method for monitoring abnormal grid operation based on edge computing. Although this method can improve the intelligence of distribution network fault diagnosis, research and application of this method and existing technologies have revealed that this method and existing technologies have at least the following shortcomings:

[0003] When a fault is detected, there is a lack of a decision-making mechanism for the backup line, resulting in an inability to determine whether to use the backup line for switching after the fault is detected, affecting the speed and flexibility of power supply restoration.

[0004] To this end, the present invention provides a DC power supply system based on intelligent computing center control. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a DC power supply system based on intelligent computing center control to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above objectives, in a first aspect, the present invention provides the following technical solutions: a DC power supply system based on intelligent computing center control, comprising:

[0007] The data acquisition module is used to collect load training data and line decision data through the smart grid background; the smart grid background obtains the line decision data and generates a training quadruple set;

[0008] A model training module is configured to train a first machine learning model for predicting maximum load values ​​using load training data obtained from the smart grid backend, and to train a second machine learning model based on a training quadruple set to determine whether to use a replacement backup line load when an abnormal transmission state is detected; and to load the trained first and second machine learning models into the smart grid backend for the topology line to be diagnosed for a fault.

[0009] The decision module generates a replacement backup line load in the smart grid background, and outputs a decision on whether to use the replacement backup line load based on the replacement backup line load, the predicted maximum load value, and the second machine learning model.

[0010] Furthermore, the load training data includes load characteristic data and a maximum load value corresponding to the load characteristic data; the load characteristic data includes line current, line voltage and line temperature;

[0011] The method for obtaining the maximum load value in the load training data is:

[0012] The maximum load value is the maximum value of the load value of the topology line corresponding to each set of load characteristic data; the maximum load value is the load value obtained by the load sensor when the load is continuously added to the corresponding topology line until the topology line fails.

[0013] Furthermore, the line decision data includes the maximum load value, transmission evaluation coefficient, active power, replacement standby line load and line change decision of the i-th topology line;

[0014] The active power is the actual power consumed by all loads on the i-th topology line in the test smart grid;

[0015] Replacing the backup line load means avoiding the starting and ending intersections of the topological routes with abnormal transmission status and selecting a backup line to continue transmitting the power load;

[0016] The line change decision is to select or not select one of the replacement backup line loads when the i-th topology line faces the grid transmission state during each smart grid test;

[0017] Analyze the power grid transmission state of the i-th topology line according to the transmission evaluation coefficient, where the power grid transmission state includes abnormal transmission state and normal transmission state;

[0018] The transmission evaluation coefficient generation process includes:

[0019] The transmission status data of the i-th topology line is obtained, and the transmission status data is evaluated to obtain a transmission evaluation coefficient.

[0020] Furthermore, the method for analyzing the power grid transmission state of the i-th topology line according to the transmission evaluation coefficient includes:

[0021] Preset a coefficient threshold, and compare the transmission evaluation coefficient with the preset coefficient threshold;

[0022] If the transmission evaluation coefficient is greater than a preset coefficient threshold, the power grid transmission state is generated as an abnormal transmission state;

[0023] If the transmission evaluation coefficient is less than or equal to the preset coefficient threshold, the power grid transmission state is generated as a normal transmission state.

[0024] Furthermore, the training quad-tuple set refers to generating a set of training quad-tuples from each set of line decision data; generating a training quad-tuple set based on all line decision data;

[0025] The method of generating a set of training quads from each set of route decision data includes:

[0026] The maximum load value, active power and replacement standby line load in each set of line decision data are used as the current state;

[0027] A route change decision based on the route decision data is used as a selected action;

[0028] Calculate the reward value Q of each set of line decision data after selecting the action;

[0029] The training quadruple includes the current state, selected action, reward value Q, and next state corresponding to each set of line decision data;

[0030] The reward value is calculated as follows:

[0031] The reward value is calculated based on the maximum load value, active power, replacement backup line load, and transmission evaluation coefficient.

[0032] Furthermore, the method of training a first machine learning model for predicting the maximum load value using load training data obtained from the smart grid backend includes:

[0033] Each set of load characteristic data is taken as the form of a first eigenvector, and the line current, line voltage and line temperature in the first eigenvector are taken as the input of a first machine learning model. The first machine learning model takes the maximum load value predicted for each set of load characteristic data as the output, the actual maximum load value corresponding to each set of load characteristic data as the prediction target, and the minimization of the sum of the first prediction accuracies of all predicted maximum load values ​​as the training target; the first machine learning model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the first machine learning model is one of a polynomial regression model or an SVM model.

[0034] Furthermore, a second machine learning model is trained to decide whether to use a replacement backup line load when an abnormal transmission state is detected as follows:

[0035] The current state and reward value Q in the training quadruple set are used as the input of the second machine learning model. The second machine learning model is trained by randomly extracting multiple groups of quadruple sets from the training quadruple set to learn whether to choose to use the replacement backup line load to obtain the maximum reward value Q under different initial states; the second machine learning model is a deep Q network model.

[0036] Furthermore, the maximum load value of the i-th topology line is predicted based on the load characteristic data and the first machine learning model as follows:

[0037] The load characteristic data collected by the smart grid background is input into the first machine learning model to obtain the output predicted maximum load value.

[0038] Furthermore, the decision of whether to use the replacement backup line load is output as follows:

[0039] The maximum load value obtained from the smart grid background, the generated replacement backup line load, and the active power of the smart grid to be controlled are input into the second machine learning model to obtain an output decision on whether to use the replacement backup line load.

[0040] Technical effects and advantages of the present invention:

[0041] This invention monitors the topological status of smart grid lines in real time, enabling timely fault detection. When a fault is detected, the system intelligently decides, based on a trained model, whether to replace backup lines to restore power to non-faulty areas, effectively preventing the spread of outages and enabling automated isolation and recovery of smart grid faults.

[0042] With the collaborative work of the data acquisition module, model training module and decision-making module, the system can conduct in-depth analysis and learning of load data and line status, form accurate fault judgment and backup line switching strategies, and significantly enhance the power grid's ability to cope with various fault situations; through intelligent decision-making to replace backup lines, power supply to non-fault areas can be quickly restored, reducing the duration and scope of power outages caused by faults, ensuring the continuity and reliability of power supply, and providing users with more stable power services; using machine learning models to train and analyze load data and line decision data, the system realizes the automation and intelligence of fault diagnosis and backup line switching decisions, greatly improving the efficiency of fault diagnosis and processing, reducing manual intervention, and reducing labor costs and the risk of human error. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of a DC power supply method based on intelligent computing center control according to Example 1;

[0044] Figure 2 This is a schematic diagram of the backup line;

[0045] Figure 3 This is a schematic diagram of a DC power supply system based on intelligent computing center control in Example 2. DETAILED DESCRIPTION

[0046] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0047] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0048] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and a similar second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0049] Example 1:

[0050] See also Figure 1 As shown, this embodiment discloses a DC power supply method based on intelligent computing center control, including:

[0051] Step 1: The smart grid backend collects load training data and line decision data;

[0052] It should be noted that the smart grid is divided into i topological lines. The smart grid division methods include division according to geographical areas, division according to the load distribution of the smart grid, or division according to the power resource distribution of the smart grid. The specific division method is determined by technical personnel in this field based on the actual situation of the smart grid and is not limited here. The smart grid is divided into i topological lines in order to simplify management, optimize scheduling, and achieve fault isolation and distributed control.

[0053] The load training data is collected from different topology lines of the test smart grid, and the load training data includes load characteristic data and a maximum load value corresponding to the load characteristic data; the load characteristic data includes line current, line voltage and line temperature.

[0054] Here, the line current is the current intensity of each topology line in the smart grid. It is understandable that since the current intensity of the topology lines is different, in order to prevent overload and short circuit and improve the safety and stability of the smart grid, it is necessary to monitor the current of different topology lines of the smart grid. The line current is measured by a current transformer to measure the current intensity of i topology lines in the smart grid. The line current is obtained by calculating the average value of the current values ​​of multiple current transformers on the same topology line.

[0055] The line voltage is acquired in real time through a voltage transformer. It is understandable that in several tested smart grids, by installing voltage transformers on i topology lines, the average value of multiple voltage transformers on the same topology line is calculated as the line voltage.

[0056] The line temperature is acquired in real time by temperature sensors on the same topology line. It is understandable that when measuring the line temperature, the average value of multiple temperature sensors on the same topology line is used as the line temperature.

[0057] The method for obtaining the maximum load value in the load training data is:

[0058] The maximum load value is the maximum value of the load value of the topology line corresponding to each set of load characteristic data; the maximum load value is the load value obtained by the load sensor when the load is continuously added to the corresponding topology line until the topology line fails.

[0059] The line decision data includes the maximum load value, transmission evaluation coefficient, active power, replacement standby line load and line change decision of the i-th topology line.

[0060] Specifically, the transmission evaluation coefficient generation process includes:

[0061] The transmission status data of the i-th topology line is obtained, and the transmission status data is evaluated to obtain a transmission evaluation coefficient.

[0062] The transmission status data includes the harmonic content of the i-th topology line, the switch state value of the circuit breaker, and the power factor. The harmonic content is the proportion of harmonics in the i-th topology line, and the power factor is the utilization efficiency of electric energy. The harmonic content and power factor are obtained through a power quality analyzer.

[0063] It should be noted that the switch status value of the circuit breaker is pre-set, and different values ​​are set according to the open state or closed state of the circuit breaker. The switch status value of the circuit breaker corresponding to the circuit breaker in the open state is set to 1, and the switch status value of the circuit breaker corresponding to the circuit breaker in the closed state is set to 0.

[0064] The harmonic content and power factor of the i-th topology line are marked as and ,in, Indicates the phase difference angle between voltage and current;

[0065] The harmonic content of the i-th topology line , power factor Perform formula calculations to obtain the transmission evaluation coefficient , the transmission evaluation coefficient The following expression is met:

[0066] ;

[0067] Where, represents the weight factor of the harmonic content of the i-th topology line, Indicates the weight factor of the backup line load, harmonic content The larger the value, the worse the power quality; the power factor is between 0 and 1, and the larger the value, the higher the power utilization rate, which is a positive indicator; , which is revised by staff based on experience. That is, the weight factor reflects the influence of the value contained in the transmission status data on the transmission evaluation coefficient of the topology line. The greater the influence, the greater the weight factor of the corresponding value.

[0068] It should be noted that the transmission assessment coefficient is only used to determine whether a fault has occurred in the power grid transmission status, that is, to reflect the health of the line. Therefore, the calculation of the transmission assessment coefficient is a dimensionless calculation.

[0069] In the implementation, the power grid transmission state of the i-th topology line is analyzed according to the transmission evaluation coefficient, and the power grid transmission state includes abnormal transmission state and normal transmission state;

[0070] The method for analyzing the power grid transmission state of the i-th topology line according to the transmission evaluation coefficient includes:

[0071] Preset a coefficient threshold, and compare the transmission evaluation coefficient with the preset coefficient threshold;

[0072] If the transmission evaluation coefficient is greater than the preset coefficient threshold, the power grid transmission state is generated as an abnormal transmission state, indicating that when the circuit breaker is open, the harmonic content of the i-th topology line becomes larger and the power factor becomes larger;

[0073] If the transmission evaluation coefficient is less than or equal to the preset coefficient threshold, the grid transmission state is generated as a normal transmission state, indicating that when the circuit breaker is open, the harmonic content of the i-th topology line becomes smaller and the power factor becomes smaller.

[0074] It should be noted that the preset coefficient threshold is determined by technical personnel in this field based on actual needs and a large amount of experimental data. The larger the transmission evaluation coefficient, the worse the grid transmission state of the corresponding i-th topology line is, and the more likely it is to fail. When the load of the topology line cannot continue to be transmitted, a backup topology line is selected to continue transmitting the load.

[0075] The active power is the actual power consumed by all loads (such as electrical appliances, equipment, etc.) on the i-th topology line in the test smart grid; it can be understood that the active power of the i-th topology line is measured by a power quality analyzer or a power meter.

[0076] The replacement standby line load is a load that bypasses the topological route with abnormal transmission status;

[0077] The replacement route is generated as follows:

[0078] Obtain the topology line from the power supply area to the user area pre-generated by the smart grid, and calculate the starting intersection point and ending intersection point of the topology line with the topology line in abnormal transmission state; the starting intersection point and ending intersection point are the circuit breaker opening position of the topology line in normal transmission state and the circuit breaker opening position of the next topology line in normal transmission state, respectively. Figure 2 As shown, the dotted line is the backup line.

[0079] Replacing the backup line load means avoiding the starting and ending intersections of the topological routes with abnormal transmission status and selecting a backup line to continue transmitting power load.

[0080] The line change decision is to select or not select one of the replacement backup line loads when the i-th topology line faces the grid transmission state during each smart grid test;

[0081] Specifically, in the line change decision, when the transmission state of the i-th topology line facing the power grid is normal, the decision is not to use the replacement standby line load; conversely, when the transmission state of the i-th topology line facing the power grid is abnormal, the decision is to use the replacement standby line load.

[0082] Step 2: The smart grid backend obtains line decision data and generates a training quadruple set;

[0083] It should be noted that: the training quadruple set refers to generating a set of training quadruple from each set of line decision data; generating a training quadruple set based on all line decision data;

[0084] Specifically, the method of generating a set of training quads from each set of line decision data includes:

[0085] The maximum load value, active power and replacement standby line load in each set of line decision data are used as the current state;

[0086] A route change decision based on the route decision data is used as a selected action;

[0087] Calculate the reward value Q of each set of line decision data after selecting the action;

[0088] The training quadruple includes the current state, selected action, reward value Q, and next state corresponding to each set of line decision data;

[0089] The reward value Q is calculated as follows:

[0090] Mark the maximum load value as , mark the active power as , mark the replacement backup line load as ; The calculation formula of reward value Q is:

[0091] ;

[0092] in, is the capacity margin benefit weight; is the capacity margin penalty weight; To switch risk weights; For action, ,when Select to switch to the backup line when Choose not to switch to the backup line; is a smaller value; is the preset coefficient threshold of the transmission evaluation coefficient; positive benefit term , encourage switching when the backup line has sufficient reserve margin; negative cost item , penalizes the switching risk of high power load; penalty term , quantify the potential loss of not switching the backup line.

[0093] Switching decision trigger conditions:

[0094] when ,and > When the backup margin benefit is greater than the risk cost of switching and the current line is in an abnormal transmission state, the backup line switching is triggered;

[0095] when or When ≈0, that is, the line is healthy or the load is extremely low, and the potential loss of not switching the backup line is close to 0, there is no risk in maintaining the status quo and no need to switch.

[0096] The maximum load value of the following set of line decision data is the next state.

[0097] Step 3: The smart grid backend obtains load training data to train a first machine learning model for predicting the maximum load value, and trains a second machine learning model based on the training quadruple set to decide whether to use a replacement backup line load when an abnormal transmission state is detected;

[0098] In implementation, the method for obtaining load training data from the smart grid backend to train a first machine learning model for predicting the maximum load value includes:

[0099] Each set of load characteristic data is taken as a first eigenvector, and the line current, line voltage, and line temperature in all the first eigenvectors are taken as inputs of a first machine learning model. The first machine learning model uses the maximum load value predicted for each set of load characteristic data as output, the actual maximum load value corresponding to each set of load characteristic data as a prediction target, and minimizing the sum of the first prediction accuracies of all predicted maximum load values ​​as a training objective. The calculation formula for the first prediction accuracy is: ; Train the first machine learning model with the goal of minimizing the sum of the first prediction accuracies, where is the group number of the load characteristic data, is the number of load characteristic data sets, is the first prediction accuracy, For the The predicted maximum load value corresponding to the group load characteristic data, For the The actual maximum load value corresponding to the group load characteristic data; training the first machine learning model until the sum of the first prediction accuracies reaches convergence and then stops training; the first machine learning model is one of a polynomial regression model or an SVM model.

[0100] In implementation, the second machine learning model for deciding whether to use a replacement backup line load when an abnormal transmission state is detected is trained as follows:

[0101] The current state and reward value Q in the training quadruple set are used as inputs of the second machine learning model. The second machine learning model is trained by randomly selecting a number of quadruple sets from the training quadruple set to learn whether to choose to use the replacement backup line load to obtain the maximum reward value Q under different initial states; the second machine learning model is a deep Q network model.

[0102] Step 4: Load the first machine learning model and the second machine learning model into the smart grid backend of the topology line to be fault diagnosed;

[0103] When the topology line to be diagnosed for fault transmits power according to the topology line, the smart grid background monitors the topology line in real time to see whether there is an abnormal transmission state; if there is an abnormal transmission state, go to step 5; if there is no abnormal transmission state, continue monitoring.

[0104] Step 5: The smart grid backend collects load characteristic data and predicts the maximum load value of the i-th topology line based on the load characteristic data and the first machine learning model;

[0105] In practice, the smart grid backend collects load characteristic data in the following ways:

[0106] The smart grid backend obtains the monitored line current, line voltage and line temperature in real time.

[0107] In implementation, the maximum load value of the i-th topology line is predicted based on the load characteristic data and the first machine learning model as follows:

[0108] The load characteristic data collected by the smart grid background is input into the first machine learning model to obtain the output predicted maximum load value.

[0109] Step 6: The smart grid backend generates a replacement backup line load and outputs a decision on whether to use the replacement backup line load based on the replacement backup line load, the predicted maximum load value, and the second machine learning model;

[0110] In implementation, the decision of whether to use the replacement backup line load is output as follows:

[0111] The maximum load value obtained from the smart grid background, the generated replacement backup line load, and the active power of the smart grid to be controlled are input into the second machine learning model to obtain an output decision on whether to use the replacement backup line load.

[0112] In this embodiment, load training data and line decision data are collected through a smart grid backend. The smart grid backend obtains the line decision data and generates a training set of four-tuples. The smart grid backend uses the load training data to train a first machine learning model for predicting maximum load values, and based on the training set of four-tuples, trains a second machine learning model for deciding whether to use a replacement backup line load when an abnormal transmission state is detected. The trained first and second machine learning models are loaded into the smart grid backend for the topological line to be diagnosed for fault. The smart grid backend collects load characteristic data and predicts the maximum load value of the i-th topological line based on the load characteristic data and the first machine learning model. The smart grid backend generates a replacement backup line load and outputs a decision on whether to use the replacement backup line load based on the replacement backup line load, the predicted maximum load value, and the second machine learning model. This embodiment monitors the topological lines of the smart grid in real time for faults. If a fault occurs, it intelligently determines whether to replace the backup line to restore power supply, thereby preventing the spread of the power outage and restoring power to non-fault areas. This achieves automatic isolation and restoration of the smart grid, enhances the grid's ability to respond to faults, and ensures the continuity and reliability of power supply.

[0113] Example 2:

[0114] See also Figure 3 As shown, this embodiment provides a DC power supply system based on intelligent computing center control, including:

[0115] The data acquisition module is used to collect load training data and line decision data through the smart grid background; the smart grid background obtains the line decision data and generates a training quadruple set;

[0116] A model training module is configured to train a first machine learning model for predicting maximum load values ​​using load training data obtained from the smart grid backend, and to train a second machine learning model based on a training quadruple set to determine whether to use a replacement backup line load when an abnormal transmission state is detected; and to load the trained first and second machine learning models into the smart grid backend for the topology line to be diagnosed for a fault.

[0117] The decision module generates a replacement backup line load in the smart grid background, and outputs a decision on whether to use the replacement backup line load based on the replacement backup line load, the predicted maximum load value, and the second machine learning model.

[0118] The formulas involved in the above are all calculated by removing the dimensions and taking their numerical values. They are a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The weight factors in the formula and the various preset thresholds in the analysis process are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data; the size of the weight factor is to quantify each parameter to obtain a specific value, which is convenient for subsequent comparison. The size of the weight factor depends on the amount of sample data and the corresponding processing coefficient initially set by technical personnel in this field for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0119] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0120] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A DC power supply system based on intelligent computing center control, characterized in that: include: Data acquisition module, used to collect load training data and line decision data through the smart grid background; The smart grid backend obtains line decision data and generates a training quadruple set; The line decision data includes the maximum load value, transmission evaluation coefficient, active power, replacement backup line load and line change decision of the i-th topology line; The active power is the actual power consumed by all loads on the i-th topology line in the test smart grid; Replacing the backup line load means avoiding the starting and ending intersections of the topological route with abnormal transmission status and selecting a backup line to continue transmitting power load; The line change decision is to select or not select one of the replacement backup line loads when the i-th topology line faces the grid transmission state during each smart grid test; Analyze the power grid transmission state of the i-th topology line according to the transmission evaluation coefficient, where the power grid transmission state includes abnormal transmission state and normal transmission state; The transmission evaluation coefficient generation process includes: Acquire transmission status data of the i-th topology line, and evaluate the transmission status data to obtain a transmission evaluation coefficient; The training quadruple set refers to generating a set of training quadruple from each set of line decision data; the training quadruple set is generated based on all line decision data; The method of generating a set of training quads from each set of route decision data includes: The maximum load value, active power and replacement standby line load in each set of line decision data are used as the current state; A route change decision based on the route decision data is used as a selected action; Calculate the reward value Q of each set of line decision data after selecting the action; The training quadruple includes the current state, selected action, reward value Q, and next state corresponding to each set of line decision data; The reward value Q is calculated as follows: The reward value is calculated based on the maximum load value, active power, replacement standby line load, and transmission evaluation coefficient; A model training module is configured to train a first machine learning model for predicting maximum load values ​​using load training data obtained from the smart grid backend, and to train a second machine learning model based on a training quadruple set to determine whether to use a replacement backup line load when an abnormal transmission state is detected; and to load the trained first and second machine learning models into the smart grid backend for the topology line to be diagnosed for a fault. The decision module generates a replacement backup line load in the smart grid background, and outputs a decision on whether to use the replacement backup line load based on the replacement backup line load, the predicted maximum load value, and the second machine learning model.

2. A DC power supply system based on intelligent computing center control according to claim 1, characterized in that: The load training data includes load characteristic data and a maximum load value corresponding to the load characteristic data; the load characteristic data includes line current, line voltage and line temperature; The method for obtaining the maximum load value in the load training data is: The maximum load value is the maximum value of the load value of the topology line corresponding to each set of load characteristic data; the maximum load value is the load value obtained by the load sensor when the load is continuously added to the corresponding topology line until the topology line fails.

3. The DC power supply system based on intelligent computing center control according to claim 1, characterized in that: The method for analyzing the power grid transmission state of the i-th topology line according to the transmission evaluation coefficient includes: Preset a coefficient threshold, and compare the transmission evaluation coefficient with the preset coefficient threshold; If the transmission evaluation coefficient is greater than a preset coefficient threshold, the power grid transmission state is generated as an abnormal transmission state; If the transmission evaluation coefficient is less than or equal to the preset coefficient threshold, the power grid transmission state is generated as a normal transmission state.

4. The DC power supply system based on intelligent computing center control according to claim 1, characterized in that: The method for training a first machine learning model for predicting a maximum load value using load training data obtained from a smart grid backend includes: Each set of load characteristic data is taken as the form of a first eigenvector, and the line current, line voltage and line temperature in the first eigenvector are taken as the input of a first machine learning model. The first machine learning model takes the maximum load value predicted for each set of load characteristic data as the output, and the actual maximum load value corresponding to each set of load characteristic data as the prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted maximum load values ​​as the training target; the first machine learning model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the first machine learning model is a polynomial regression model or an SVM model.

5. The DC power supply system based on intelligent computing center control according to claim 4 is characterized in that: The method for training the second machine learning model for deciding whether to use a replacement backup line load when an abnormal transmission state is detected is as follows: The current state and reward value Q in the training quadruple set are used as the input of the second machine learning model. The second machine learning model is trained by randomly extracting multiple groups of quadruple sets from the training quadruple set to learn whether to choose to use the replacement backup line load to obtain the maximum reward value Q under different initial states; the second machine learning model is a deep Q network model.

6. The DC power supply system based on intelligent computing center control according to claim 5, characterized in that: The method for predicting the maximum load value of the i-th topology line based on the load characteristic data and the first machine learning model is: The load characteristic data collected by the smart grid background is input into the first machine learning model to obtain the output predicted maximum load value.

7. The DC power supply system based on intelligent computing center control according to claim 6, characterized in that: The decision on whether to use the replacement backup line load is output as follows: The maximum load value obtained from the smart grid background, the generated replacement backup line load, and the active power of the smart grid to be controlled are input into the second machine learning model to obtain an output decision on whether to use the replacement backup line load.

Images