SVG low-voltage control system and method based on networking in photovoltaic power station

Abstract

The invention relates to the technical field of power supply systems, in particular to a photovoltaic power station-based networking SVG low-voltage control system and method.The photovoltaic power station-based networking SVG low-voltage control system comprises a data acquisition module, a decision module, an analogue simulation module, an execution module and a monitoring feedback module. Through simulation evaluation, instruction optimization, power converter control, safety monitoring, feedback and prediction and optimization, accurate control can be realized, and instruction reliability and equipment safety are improved.

Description

Technical Field

[0001] This invention relates to the field of power supply system technology, and in particular to a low-voltage control system and method based on SVG in a photovoltaic power plant. Background Technology

[0002] Against the backdrop of the transformation of the new power system towards a high proportion of new energy sources, the installed capacity of intermittent power sources such as wind power and photovoltaics continues to rise, leading to strong fluctuations and high disturbances in the power grid, which places higher demands on reactive power regulation and voltage support equipment. As a core control device, grid-based static var generators (SVG) need to adapt to changes in power grid operating conditions in real time. However, traditional grid-based SVGs have significant limitations: long data acquisition cycles make it difficult to capture dynamic parameters of equipment and the power grid in a timely manner; decision-making relies on fixed thresholds or static algorithms, which cannot be iteratively optimized according to operating conditions; control commands lack virtual and real pre-verification, and direct physical execution can easily cause power grid disturbances or equipment failures; moreover, the monitoring feedback and control links are disconnected, forming an "open-loop control" that makes it difficult to avoid safety risks in a timely manner. Therefore, it is urgent to build a full-process intelligent control system covering "sensing-decision-simulation-execution-feedback" to solve the problem of synergistic accuracy and safety of grid-based SVGs and support the stable operation of the new power system.

[0003] Chinese patent CN115051378A discloses a reactive power hierarchical control system for photovoltaic power plants. The system comprises three layers: a first layer (reactive power demand detection layer) that calculates the reactive power compensation required to maintain the grid connection voltage based on real-time monitoring of the grid connection voltage and power by an AVC system; a second layer (power plant-level reactive power coordination and allocation layer) that collects the total reactive power surplus of inverters in each collector line through the AVC system, compares it with the reactive power compensation obtained from the first layer, and allocates a portion of the reactive power compensation to each collector line according to its reactive power requirements, with the remaining reactive power being handled by the SVG (Static Var Generator); and a third layer (inter-inverter reactive power coordination and allocation layer) that distributes the reactive power allocated from each collector line to each inverter according to a distribution procedure, enabling the inverters to output the corresponding reactive power compensation according to the allocated amount. This addresses the issues of low utilization rate of photovoltaic inverters, underutilization of their reactive power support capabilities, and coordination and scheduling between the reactive power of photovoltaic inverters and the SVG (Static Var Generator) in the photovoltaic power plant. However, this solution still suffers from low control accuracy and low operational safety due to insufficient coordination among networked SVG devices. Summary of the Invention

[0004] Therefore, this invention provides a low-voltage control system and method based on in-grid SVG in photovoltaic power plants to overcome the problems of low control accuracy and low operational safety caused by insufficient coordination of in-grid SVG equipment in the prior art.

[0005] To achieve the above objectives, in one aspect, the present invention provides a low-voltage control system based on a photovoltaic power plant's in-grid SVG, comprising: The data acquisition module is used to acquire SVG network data; The decision module is used to analyze the current operating status of the device based on the network SVG data to obtain the current operating status of the device. It also constructs a decision model using the decision model construction method, generates control decisions based on the decision model and the current operating status of the device, and converts control commands based on the control decisions to obtain control commands. The simulation module is used to perform simulation evaluation on control commands, obtain simulation evaluation results, and score and adjust the conversion process of control commands based on the simulation evaluation results. The execution module is used to control the power converter according to control commands; The monitoring and feedback module is used to acquire the running data after execution, and to make security predictions for the network SVG devices based on the running data after execution, obtain security prediction results, and optimize the scoring adjustment process based on the security prediction results.

[0006] Furthermore, when the decision module analyzes the current device operating status based on the network SVG data, it compares the operating voltage Uy in the network SVG data with the preset operating voltage Uy0, judges the voltage situation based on the comparison result, and outputs the current device status based on the judgment result, wherein: When Uy≥Uy0, the decision module determines that the voltage is sufficient and outputs normal operation as the current device status. When Uy < Uy0, the decision module determines that the voltage is insufficient and outputs the fault operation as the current equipment status.

[0007] Furthermore, the decision model construction method includes: Step A1: Divide the decision dataset in the SVG network data into a 70% decision training set and a 30% decision validation set; Step A2: Select a recurrent neural network model as the neural network architecture for the decision model, select the Adam optimizer and cross-entropy loss function to train the recurrent neural network model, load the decision training set into the recurrent neural network model, perform forward propagation through the recurrent neural network model, and calculate the output value of the decision model. Step A3: Calculate the loss function value based on the output value and the true value of the recurrent neural network model, calculate the gradient through the backpropagation algorithm, and update the weights and biases of the recurrent neural network model. Repeat the forward propagation, loss calculation and backpropagation process until the preset training rounds are reached. Then, verify the accuracy of the recurrent neural network model through the decision validation set, and output the recurrent neural network model with an accuracy of 95%.

[0008] Furthermore, the decision module inputs the current device state into the decision model and obtains the control decision output by the decision model.

[0009] Furthermore, the decision module calculates the voltage difference Uc based on the target capacitor voltage Um in the control decision and the actual value Us of the DC-side capacitor voltage in the network SVG data, sets Uc=Um-Udc, and constructs a reactive current calculation model using the reactive current calculation model construction method to obtain the reactive current calculation model. The voltage difference Uc is input into the reactive current calculation model to obtain the reactive current output value output by the reactive current calculation model, and the reactive current output value is sent to the execution module as a control command.

[0010] Furthermore, the reactive current calculation model construction method includes: The reactive current dataset in the network SVG data is divided into a 70% training set, a 15% validation set, and a 15% test set. The training set is input into a decision tree model to train the model, resulting in a trained decision tree model. The validation set is then input into the trained decision tree model for hyperparameter iterative optimization, resulting in a hyperparameter-iterated decision tree model. An analysis test set is then input into the hyperparameter-iterated decision tree model to perform conflict analysis testing, yielding the test results. The total number of samples in the test set is set as q0, the number of correctly calculated test samples is set as q, and the test accuracy is set as Q, where Q = q / q0. The test accuracy Q is compared with a preset test accuracy Q0, with a setting of 85% ≤ Q0 ≤ 92%. Based on the comparison results, the training performance of the hyperparameter-iterated decision tree model is assessed, and the reactive current calculation model is output based on the assessment results. When Q≥Q0, the decision module determines that the training of the decision tree model after hyperparameter iteration has reached the target, and outputs the decision tree model after hyperparameter iteration as the reactive current calculation model. When Q < Q0, the decision module determines that the training of the decision tree model after hyperparameter iteration is not up to standard, updates the reactive current dataset to obtain an updated reactive current dataset, and retrains, iterates, optimizes, and analyzes the decision tree model based on the updated reactive current dataset until the training of the decision tree model is up to standard.

[0011] Furthermore, the simulation evaluation method includes: Step B1: Construct the digital twin and instruction evaluation model; Step B2: Input the control commands into the digital twin to obtain the twin's operating data, and input the twin's operating data into the command evaluation model to obtain the command evaluation score K; Step B3: Compare the instruction evaluation score K with the preset instruction evaluation score K0, determine the instruction's pass / fail status based on the comparison result, and output the simulation evaluation result based on the determination result, wherein: When K < K0, the simulation module determines the command's passability as unqualified, outputs the unqualified result as the simulation evaluation result, and performs a scoring adjustment on the control command conversion process. The scoring adjustment includes adjusting the preset calculated test accuracy Q0 value to 95%. When K≥K0, the simulation module determines the instruction to be qualified and outputs the qualified result as the simulation evaluation result.

[0012] Furthermore, the execution module inputs control commands into the PWM modulator, obtains the pulse signal output by the PWM modulator, and controls the networked SVG device according to the pulse signal.

[0013] Furthermore, the monitoring module acquires the post-execution running data, constructs a safety prediction model using a safety prediction model construction method, inputs the post-execution running data into the safety prediction model, and obtains the safety prediction results output by the safety prediction model. The safety prediction results include predictions of no fault and predictions of fault, wherein: When the safety prediction result indicates a predicted fault, the scoring adjustment process is optimized for safety. The preset instruction evaluation score K0 is optimized using the safety optimization coefficient α to obtain the optimized preset instruction evaluation score K0y. α is set to 1.25, K0y = K0 × α. The value of the preset instruction evaluation score K0 is replaced with the value of the optimized preset instruction evaluation score K0y, and the instruction evaluation score K is re-compared with the preset instruction evaluation score K0. When the safety prediction result is a prediction of no faults, no safety optimization is performed on the scoring adjustment process.

[0014] On the other hand, the present invention also provides a method for a low-voltage control system based on a photovoltaic power plant's in-grid SVG, comprising: Step S1: Collect the SVG data for the network structure; Step S2: Analyze the current device operating status based on the network SVG data to obtain the current device operating status. Also, construct a decision model using the decision model construction method, generate control decisions based on the decision model and the current device operating status, and convert control commands based on the control decisions to obtain control commands. Step S3: The control command is evaluated using a simulation evaluation method to obtain the simulation evaluation results, and the conversion process of the control command is scored and adjusted based on the simulation evaluation results. Step S4: Control the power converter according to the control command; Step S5 involves acquiring the post-execution running data, performing security predictions on the network SVG devices based on the post-execution running data, obtaining security prediction results, and optimizing the scoring adjustment process based on the security prediction results.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: the system collects network SVG data, analyzes equipment operating status, constructs a decision model to generate control commands, optimizes commands through simulation evaluation, controls the power converter, and monitors feedback to predict and optimize safety, achieving precise control and improving command reliability and equipment safety. The system accurately acquires core equipment data through the data acquisition module, providing a reliable foundation for subsequent analysis; the system also uses the decision module to combine data analysis of operating status and construct a dedicated decision model to generate control commands that better meet actual needs; the system also uses the simulation module to evaluate and adjust commands in advance, effectively avoiding the risks of unreasonable commands; the system also uses the execution module to precisely control the power converter, ensuring that the equipment operates as expected; the system also uses the monitoring feedback module to acquire operating data in real time and predict safety, and can also optimize the scoring process, significantly improving the overall reliability and safety of equipment operation and ensuring stable operation. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the low-voltage control system based on the SVG in a photovoltaic power plant in this embodiment; Figure 2 This is a flowchart illustrating the decision model construction method in this embodiment; Figure 3 This is a flowchart illustrating the simulation evaluation method in this embodiment; Figure 4 This is a flowchart illustrating the method of the low-voltage control system based on the SVG in a photovoltaic power plant in this embodiment. Detailed Implementation

[0017] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0018] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0019] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0020] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] Please see Figure 1 As shown, this is a schematic diagram of the low-voltage control system based on the SVG grid in a photovoltaic power plant in this embodiment. The system includes: The data acquisition module is used to acquire SVG network data; The decision module is used to analyze the current operating status of the device based on the network SVG data to obtain the current operating status of the device. It also constructs a decision model using a decision model construction method, generates control decisions based on the decision model and the current operating status of the device, and converts control commands based on the control decisions to obtain control commands. The decision module is connected to the data acquisition module. The simulation module is used to perform simulation evaluation on control commands using simulation evaluation methods, obtain simulation evaluation results, and score and adjust the conversion process of control commands based on the simulation evaluation results. The simulation module is connected to the decision module. An execution module is used to control the power converter according to control commands, and the execution module is connected to the simulation module; The monitoring and feedback module is used to acquire the running data after execution, and to make security predictions for the network SVG devices based on the running data after execution, thereby obtaining security prediction results. It also optimizes the scoring adjustment process based on the security prediction results. The monitoring and feedback module is connected to the execution module.

[0022] Specifically, the aforementioned low-voltage control system based on grid-connected SVG in photovoltaic power plants is applied to the control terminal of grid-connected SVG equipment in photovoltaic power plants. The system collects grid-connected SVG data, analyzes equipment operating status, constructs a decision model to generate control commands, optimizes commands through simulation evaluation, controls the power converter, and monitors feedback to predict and optimize safety. This enables precise control, improves command reliability and equipment safety. The system accurately acquires core equipment data through a data acquisition module, providing a reliable foundation for subsequent analysis. The system also uses a decision module to combine data analysis of operating status and construct a dedicated decision model to generate control commands that better meet actual needs. Furthermore, the system uses a simulation module to evaluate and adjust commands in advance, effectively avoiding the risks of unreasonable commands. The system also uses an execution module to precisely control the power converter, ensuring the equipment operates as expected. Finally, the system uses a monitoring and feedback module to acquire operating data in real time and predict safety, and optimizes the scoring process, significantly improving overall equipment operating reliability and safety, ensuring stable operation.

[0023] Specifically, the network SVG data includes operating voltage, decision dataset, actual value of DC-side capacitor voltage, and reactive current dataset. The operating voltage refers to the real-time operating voltage of the network SVG equipment, which is acquired by the data acquisition module using a voltmeter. The actual value of the DC-side capacitor voltage refers to the instantaneous measurement of the voltage across the supporting capacitor on the SVG DC bus, also acquired by the data acquisition module using a voltmeter. The decision dataset includes historically acquired current equipment states and corresponding control decisions, obtained by the data acquisition module from these historical current equipment states and corresponding control decisions. The reactive current dataset includes historically acquired voltage differences and corresponding reactive current output values, obtained by the data acquisition module from these historical voltage differences and corresponding reactive current output values.

[0024] Specifically, when the decision module analyzes the current device operating status based on the network SVG data, it compares the operating voltage Uy in the network SVG data with the preset operating voltage Uy0, judges the voltage situation based on the comparison result, and outputs the current device status based on the judgment result, wherein: When Uy≥Uy0, the decision module determines that the voltage is sufficient and outputs normal operation as the current device status. When Uy < Uy0, the decision module determines that the voltage is insufficient and outputs the fault operation as the current equipment status.

[0025] Specifically, the preset operating voltage refers to a preset value used to judge the voltage condition. This embodiment does not limit the specific value of the preset operating voltage. Those skilled in the art can set it according to the actual situation, as long as it meets the requirements for judging the voltage condition. For example, the preset operating voltage Uy0 can be set to 80% of the rated voltage. The rated voltage refers to the optimal voltage value specified for the network SVG equipment to operate normally and safely for a long time.

[0026] Please see Figure 2 As shown, this is a flowchart illustrating the decision model construction method in this embodiment. The decision model construction method includes: Step A1: Divide the decision dataset in the SVG network data into a 70% decision training set and a 30% decision validation set; Step A2: Select a recurrent neural network model as the neural network architecture for the decision model, select the Adam optimizer and cross-entropy loss function to train the recurrent neural network model, load the decision training set into the recurrent neural network model, perform forward propagation through the recurrent neural network model, and calculate the output value of the decision model. Step A3: Calculate the loss function value based on the output value and the true value of the recurrent neural network model, calculate the gradient through the backpropagation algorithm, and update the weights and biases of the recurrent neural network model. Repeat the forward propagation, loss calculation and backpropagation process until the preset training rounds are reached. Then, verify the accuracy of the recurrent neural network model through the decision validation set, and output the recurrent neural network model with an accuracy of 95%.

[0027] Specifically, the decision training set refers to the dataset used to train the recurrent neural network model, the decision validation set refers to the dataset used to validate the accuracy of the recurrent neural network model, and the decision dataset refers to the dataset used to construct the decision model. The decision dataset includes the historically acquired current device state and the control decisions corresponding to the historically acquired current device state.

[0028] Specifically, the decision module inputs the current device state into the decision model and obtains the control decision output by the decision model.

[0029] Specifically, the decision module calculates the voltage difference Uc based on the target capacitor voltage Um in the control decision and the actual value Us of the DC-side capacitor voltage in the network SVG data, setting Uc = Um - Udc. It then constructs a reactive current calculation model using a reactive current calculation model construction method, inputs the voltage difference Uc into the reactive current calculation model, obtains the reactive current output value from the model, and sends the reactive current output value as a control command to the execution module. The reactive current calculation model construction method includes: The reactive current dataset in the network SVG data is divided into a 70% training set, a 15% validation set, and a 15% test set. The training set is input into a decision tree model to train the model, resulting in a trained decision tree model. The validation set is then input into the trained decision tree model for hyperparameter iterative optimization, resulting in a hyperparameter-iterated decision tree model. An analysis test set is then input into the hyperparameter-iterated decision tree model to perform conflict analysis testing, yielding the test results. The total number of samples in the test set is set as q0, the number of correctly calculated test samples is set as q, and the test accuracy is set as Q, where Q = q / q0. The test accuracy Q is compared with a preset test accuracy Q0, with a setting of 85% ≤ Q0 ≤ 92%. Based on the comparison results, the training performance of the hyperparameter-iterated decision tree model is assessed, and the reactive current calculation model is output based on the assessment results. When Q≥Q0, the decision module determines that the training of the decision tree model after hyperparameter iteration has reached the target, and outputs the decision tree model after hyperparameter iteration as the reactive current calculation model. When Q < Q0, the decision module determines that the training of the decision tree model after hyperparameter iteration is not up to standard, updates the reactive current dataset to obtain an updated reactive current dataset, and retrains, iterates, optimizes, and analyzes the decision tree model based on the updated reactive current dataset until the training of the decision tree model is up to standard.

[0030] Specifically, the target capacitor voltage refers to the voltage value required for the supporting capacitor in the control decision; the reactive current calculation model refers to a decision tree model that takes voltage difference as input and reactive current output as output; the reactive current dataset includes historically acquired voltage differences and corresponding reactive current output values; the calculation training set refers to the dataset used to train the decision tree model within the reactive current dataset; the calculation validation set refers to the dataset used to validate the training results of the decision tree model within the reactive current dataset; the calculation test set refers to the dataset used to test the decision tree model within the reactive current dataset; the calculation test accuracy rate refers to the ratio of the number of correctly calculated test samples to the total number of samples in the calculation test set; the preset calculation test accuracy rate refers to a preset value used to judge the training compliance status of the iteratively optimized decision tree model; the training compliance status of the iteratively optimized decision tree model refers to the compliance status of the accuracy of the trained decision tree model; the training compliance status of the iteratively optimized decision tree model includes compliance and non-compliance.

[0031] Please see Figure 3 As shown, this is a flowchart illustrating the simulation evaluation method in this embodiment. The simulation evaluation method includes: Step B1: Construct the digital twin and instruction evaluation model; Step B2: Input the control commands into the digital twin to obtain the twin's operating data, and input the twin's operating data into the command evaluation model to obtain the command evaluation score K; Step B3: Compare the instruction evaluation score K with the preset instruction evaluation score K0, determine the instruction's pass / fail status based on the comparison result, and output the simulation evaluation result based on the determination result, wherein: When K < K0, the simulation module determines the command's passability as unqualified, outputs the unqualified result as the simulation evaluation result, and performs a scoring adjustment on the control command conversion process. The scoring adjustment includes adjusting the preset calculated test accuracy Q0 value to 95%. When K≥K0, the simulation module determines the instruction to be qualified and outputs the qualified result as the simulation evaluation result.

[0032] Specifically, the digital twin refers to a "digital twin" created in virtual space that completely corresponds to the networked SVG device. This simulation model, achieved through the use of control commands, enables the analysis and optimization of the networked SVG device. This embodiment does not limit the construction method of the digital twin; those skilled in the art can set it according to actual conditions. For example, the digital twin model can be trained using a predictive runtime dataset to obtain the digital twin. The predictive runtime dataset refers to the training dataset used to construct the digital twin, which includes historically acquired control commands and the corresponding twin runtime data. The command evaluation model refers to a machine learning model that takes the twin runtime data as input and outputs a command evaluation score. This embodiment does not specify the command evaluation model. The method of constructing the instruction evaluation model is limited, but those skilled in the art can set it according to the actual situation. For example, a machine learning model can be trained using a scoring dataset to obtain an instruction evaluation model. The scoring dataset refers to the training dataset used to construct the instruction evaluation model. The scoring dataset includes historically acquired twin operation data and instruction evaluation scores corresponding to the historically acquired twin operation data. The preset instruction evaluation score refers to a preset value used to judge the qualification status of an instruction. This embodiment does not limit the specific value of the preset instruction evaluation score, and those skilled in the art can set it according to the actual situation. For example, K0 can be set to 70 points. The qualification status of an instruction refers to whether the control instruction is qualified based on the instruction evaluation score and the preset instruction evaluation score. The qualification status of an instruction includes qualified and unqualified.

[0033] Specifically, the execution module inputs control commands into the PWM modulator, obtains the pulse signal output by the PWM modulator, and controls the networked SVG device according to the pulse signal.

[0034] Specifically, the PWM modulator refers to a modulator that converts a low-frequency execution signal into a series of high-frequency pulse signals with varying duty cycles, thereby controlling the on and off states of power switching devices. The full name of the PWM modulator is Pulse Width Modulation.

[0035] Specifically, the monitoring module acquires post-execution running data, constructs a safety prediction model using a safety prediction model construction method, inputs the post-execution running data into the safety prediction model, and obtains the safety prediction results output by the safety prediction model. The safety prediction results include predictions of no fault and predictions of fault, wherein: When the safety prediction result indicates a predicted fault, the scoring adjustment process is optimized for safety. The preset instruction evaluation score K0 is optimized using the safety optimization coefficient α to obtain the optimized preset instruction evaluation score K0y. α is set to 1.25, K0y = K0 × α. The value of the preset instruction evaluation score K0 is replaced with the value of the optimized preset instruction evaluation score K0y, and the instruction evaluation score K is re-compared with the preset instruction evaluation score K0. When the safety prediction result is a prediction of no faults, no safety optimization is performed on the scoring adjustment process.

[0036] Specifically, the post-execution operational data refers to all operational data generated during the operation of the power grid after the execution of the control command, such as three-phase voltage and three-phase current. This embodiment does not limit the specific implementation method for acquiring post-execution operational data. Those skilled in the art can set it according to the actual situation. For example, post-execution operational data can be acquired by a sensor cluster set at key nodes of the power grid. The key nodes of the power grid refer to the key operational nodes in the power grid system, such as substations, distribution rooms, and line switches. The sensor cluster refers to a set of sensors used to collect post-execution operational data, such as voltage transformers and current transformers. This embodiment does not limit the specific content of the safety prediction model construction method. Those skilled in the art can set it according to the actual situation. For example, a machine learning model can be trained using a safety prediction dataset to obtain a safety prediction model. The safety prediction dataset refers to the training dataset used to construct the safety prediction model. The safety prediction dataset includes historically acquired post-execution operational data and the safety prediction results corresponding to the historically acquired post-execution operational data.

[0037] Please see Figure 4 As shown, this is a flowchart illustrating the method for a low-voltage control system based on a photovoltaic power plant's in-grid SVG in this embodiment. The method includes: Step S1: Collect the SVG data for the network structure; Step S2: Analyze the current device operating status based on the network SVG data to obtain the current device operating status. Also, construct a decision model using the decision model construction method, generate control decisions based on the decision model and the current device operating status, and convert control commands based on the control decisions to obtain control commands. Step S3: The control command is evaluated using a simulation evaluation method to obtain the simulation evaluation results, and the conversion process of the control command is scored and adjusted based on the simulation evaluation results. Step S4: Control the power converter according to the control command; Step S5 involves acquiring the post-execution running data, performing security predictions on the network SVG devices based on the post-execution running data, obtaining security prediction results, and optimizing the scoring adjustment process based on the security prediction results.

[0038] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A low-voltage control system based on network construction SVG in a photovoltaic power station, characterized in that, The method comprises the following steps: a data acquisition module is configured to acquire network SVG data; a decision module is configured to analyze the current device operating state according to the network SVG data, obtain the current device operating state, build a decision model through a decision model building method, generate a control decision according to the decision model and the current device operating state, convert the control decision into a control instruction according to the control decision, and obtain the control instruction; a simulation simulation module is configured to simulate and evaluate the control instruction through a simulation evaluation method, obtain a simulation evaluation result, and adjust the conversion process of the control instruction according to the simulation evaluation result; an execution module is configured to control the power converter according to the control instruction; a monitoring feedback module is configured to obtain post-execution operating data, predict the safety of the network SVG device according to the post-execution operating data, obtain a safety prediction result, and optimize the process of score adjustment according to the safety prediction result. 2.The low voltage control system based on the network construction SVG in the photovoltaic power station of claim 1, characterized in that, When Uy≥Uy0, the decision module determines that the voltage condition is sufficient, and outputs normal operation as the current device state; When UyUy0, the decision module determines that the voltage condition is insufficient, and outputs fault operation as the current device state. The decision model building method comprises the following steps: 3.The low voltage control system based on the network construction SVG in the photovoltaic power station of claim 1, characterized in that, Step A1: divide the decision data set in the network SVG data into a 70% decision training set and a 30% decision verification set; Step A2: select a recurrent neural network model as the neural network architecture of the decision model, select an Adam optimizer and a cross-entropy loss function to train the recurrent neural network model, load the decision training set into the recurrent neural network model, perform forward propagation through the recurrent neural network model, and calculate the output value of the decision model; Step A3: calculate the loss function value according to the output value and the true value of the recurrent neural network model, calculate the gradient through the back propagation algorithm, update the weights and biases of the recurrent neural network model, repeat the processes of forward propagation, loss calculation, and back propagation until the preset training round is reached, verify the accuracy of the recurrent neural network model through the decision verification set, and output the recurrent neural network model with an accuracy of 95% as the decision model. The decision module inputs the current device state into the decision model to obtain the control decision output by the decision model. 4.The low voltage control system based on network construction SVG in a photovoltaic power station according to claim 1, characterized in that, ​ 5.The low voltage control system based on network construction SVG in a photovoltaic power station according to claim 1, characterized in that, The decision module calculates a voltage difference value Uc according to a target capacitor voltage Um in the control decision and an actual value Us of a direct current side capacitor voltage in the network configuration SVG data, sets Uc=Um-Udc, constructs a reactive current calculation model by a reactive current calculation model construction method, obtains the reactive current calculation model, inputs the voltage difference value Uc into the reactive current calculation model, acquires a reactive current output value output by the reactive current calculation model, and sends the reactive current output value as a control instruction to the execution module. 6.The low voltage control system based on network construction SVG in a photovoltaic power station according to claim 5, characterized in that, The reactive current calculation model construction method comprises: The reactive current data set in the network configuration SVG data is divided into a 70% calculation training set, a 15% calculation verification set and a 15% calculation test set, the calculation training set is input into the decision tree model to train the decision tree model, a trained decision tree model is obtained, the calculation verification set is input into the trained decision tree to perform hyperparameter iterative optimization on the trained decision tree, a hyperparameter-iterated decision tree model is obtained, and the analysis test set is input into the hyperparameter-iterated decision tree model to perform conflict analysis testing on the hyperparameter-iterated decision tree model, a calculation test result is obtained, the total number of calculation test samples is set as q0, the number of correctly calculated test samples is q, the calculation test accuracy is Q, Q=q / q0, the calculation test accuracy Q is compared with a pre-designed calculation test accuracy Q0, 85%≤Q0≤92% is set, the training of the hyperparameter-iterated decision tree model is judged according to the comparison result, and the reactive current calculation model is output according to the judgment result, wherein: When Q≥Q0, the decision module determines that the hyperparameter-iterated decision tree model training is up to standard, and outputs the hyperparameter-iterated decision tree model as the reactive current calculation model; When Q<Q0, the decision module determines that the hyperparameter-iterated decision tree model training is not up to standard, updates the reactive current data set to obtain an updated reactive current data set, and re-trains, hyperparameter-iteratively optimizes and analyzes the decision tree model according to the updated reactive current data set until the decision tree model training is up to standard.

7. The low voltage control system based on network construction SVG in a photovoltaic power station according to claim 6, characterized in that, The simulation evaluation method comprises: Step B1, constructing a digital twin and an instruction evaluation model; Step B2, inputting the control instruction into the digital twin to obtain twin running data, and inputting the twin running data into the instruction evaluation model to obtain an instruction evaluation score K; Step B3, comparing the instruction evaluation score K with a pre-set instruction evaluation score K0, judging the instruction qualification according to the comparison result, and outputting a simulation evaluation result according to the judgment result, wherein: When K<K0, the simulation module determines that the instruction qualification is unqualified, outputs the unqualified as the simulation evaluation result, and adjusts the conversion process of the control instruction, the score adjustment comprising: adjusting the value of the pre-designed calculation test accuracy Q0 to 95%; When K≥K0, the simulation module determines that the instruction qualification is qualified, and outputs the qualified as the simulation evaluation result. 8.The low voltage control system based on network construction SVG in a photovoltaic power station according to claim 6, characterized in that, The execution module inputs the control instruction into the PWM modulator, acquires the pulse signal output by the PWM modulator, and controls the network-constructed SVG device according to the pulse signal. 9.The low voltage control system based on network construction SVG in a photovoltaic power station according to claim 7, characterized in that, The monitoring module acquires the post-execution operation data, constructs the safety prediction model by the safety prediction model construction method, inputs the post-execution operation data into the safety prediction model, acquires the safety prediction result output by the safety prediction model, and the safety prediction result includes predicted failure-free and predicted failure, wherein: When the safety prediction result is predicted failure, the safety optimization is performed on the score adjustment process, the preset instruction evaluation score K0 is safety optimized by a safety optimization coefficient a, to obtain an optimized preset instruction evaluation score K0y, a is set to 1.25, K0y=K0×a, the value of the preset instruction evaluation score K0 is replaced by the value of the optimized preset instruction evaluation score K0y, and the instruction evaluation score K is compared with the preset instruction evaluation score K0 again; When the safety prediction result is predicted failure-free, the safety optimization is not performed on the score adjustment process.

10. A method applied to the low voltage control system of the network based SVG in the photovoltaic power station according to any one of claims 1-9, characterized in that, The method comprises the steps of: Step S1, acquiring network-constructed SVG data; Step S2, analyzing the current device operation state according to the network-constructed SVG data, obtaining the current device operation state, constructing a decision model by the decision model construction method, generating a control decision according to the decision model and the current device operation state, converting the control instruction according to the control decision, and obtaining the control instruction; Step S3, simulating and evaluating the control instruction by the simulation evaluation method, obtaining a simulation evaluation result, and adjusting the score according to the simulation evaluation result; Step S4, controlling the power converter according to the control instruction; Step S5, acquiring post-execution operation data, safety predicting the network-constructed SVG device according to the post-execution operation data, obtaining a safety prediction result, and safety optimizing the score adjustment process according to the safety prediction result.

Images