Terminal low voltage treatment method and system without current sampling

Abstract

The embodiment of the invention provides a terminal low voltage treatment method and system without current sampling, and belongs to the technical field of terminal low voltage treatment. The method comprises the steps of collecting power grid data in real time by using a sensor network; carrying out load impedance inversion through the edge computing node, and generating a virtual impedance adjustment scheme according to a result of the load impedance inversion; impedance adjustment is carried out; and after impedance adjustment is completed, performing power grid state prediction by using a federated learning model, and executing power grid self-adaptive optimization regulation and control according to a power grid state prediction result to treat tail end low voltage. According to the invention, through cooperation of global sensing, edge intelligence and federated learning, terminal low-voltage precise treatment without current sampling is realized, a quantum optimization algorithm and adaptive closed-loop adjustment are combined, and the power grid stability, the electric energy quality and the intelligent level are significantly improved while data privacy is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of terminal low voltage management, and in particular to a method and system for terminal low voltage management without current sampling. Background Art

[0002] The low voltage control at the end of the power system is a key link to ensure the quality of power supply and the stability of the power grid. With the development of the economy and the increase in electricity demand, the problem of low voltage at the end is particularly prominent in rural and remote areas and areas with dense industrial loads. The problem of low voltage at the end is usually caused by factors such as line impedance mismatch, uneven load distribution and dynamic changes in the power grid, which seriously affects the user's power experience and the normal operation of equipment.

[0003] Traditional methods for managing low voltage at the end of the line mainly rely on current sampling technology. This method indirectly infers the state of the power grid by monitoring current changes, and uses a centralized control strategy to regulate voltage. In the early days when the power grid was small and the load did not change much, this method could meet basic voltage regulation needs. However, with the widespread access to distributed power sources and the development of smart grid technology, traditional current sampling methods have gradually exposed their limitations.

[0004] In recent years, with the rise of sensor networks and edge computing technologies, some studies have attempted to introduce distributed monitoring and real-time control mechanisms to improve governance efficiency. However, existing technologies still have significant limitations in terms of data real-time impedance regulation and grid status prediction capabilities. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a method and system for terminal low voltage management without current sampling, which can fully perceive the state of the power grid, accurately adjust the impedance and have efficient prediction capabilities, and can effectively manage the low voltage at the end of the power system.

[0006] In the first aspect, in order to achieve the above-mentioned purpose, an embodiment of the present invention provides a method for terminal low voltage control without current sampling, including establishing a global perception sensor network, and using the sensor network to collect power grid data in real time; based on the collected power grid data, load impedance inversion is performed through an edge computing node, and a virtual impedance adjustment scheme is generated based on the result of the load impedance inversion; according to the virtual impedance adjustment scheme, impedance matching verification is performed, the actual impedance value is verified through a four-wire measurement method, and impedance adjustment is performed; after the impedance adjustment is completed, a federated learning model is used to predict the state of the power grid, and based on the result of the power grid state prediction, a power grid adaptive optimization control is performed to control the terminal low voltage.

[0007] Optionally, the load impedance inversion through the edge computing node includes: preprocessing the electromagnetic field data collected by the sensor, and using the preprocessed data to construct an impedance inversion model based on the field distribution characteristics; defining a virtual impedance optimization objective function, using the power grid topology as a priori constraint, and establishing a block impedance matrix to reduce the complexity of multi-node calculations; using a quantum adaptive impedance matching algorithm to calculate the optimal virtual impedance value; using a quantum genetic algorithm to update the phase angle of the quantum bit according to the calculated optimal virtual impedance value, so that the optimal virtual impedance value gradually converges to the optimal solution.

[0008] Optionally, making the optimal virtual impedance value gradually converge to the optimal solution includes: using FPGA to perform parallel calculations of quantum chromosomes and introducing a random perturbation mechanism; adjusting part of the chromosomes after each iteration; and setting a maximum number of iterations.

[0009] Optionally, performing load impedance inversion through edge computing nodes also includes introducing a spatial interpolation algorithm to complete the field data of unmonitored points; introducing the spatial interpolation algorithm includes: calculating the interpolation weight for the area not covered by the sensor network based on known point data, and verifying whether the direction of the electric field is consistent with the direction of the conductor after the interpolation is completed; when there are new sensors or the grid topology changes, recalculating the interpolation; and inputting the calculated interpolation into the quantum genetic algorithm to optimize the virtual impedance.

[0010] Optionally, according to the virtual impedance adjustment scheme, impedance matching verification is performed, the actual impedance value is verified by a four-wire measurement method, and impedance adjustment is performed, including: configuring a GaN switch matrix for impedance adjustment; deploying a MEMS capacitor array for dynamic reactive power compensation; using a four-wire measurement method to verify the actual impedance value, and when the error between the actual impedance value and the target impedance value is greater than a set first threshold, triggering PID closed-loop compensation and automatically adjusting the impedance value; collecting and calculating the power grid data again, and if the error between the actual impedance value and the target impedance value is still greater than the first threshold, performing a second adjustment until the error between the actual impedance value and the target impedance value is less than or equal to the first threshold.

[0011] Optionally, the use of a federated learning model to predict grid status includes: designing an initial prediction model, and distributing the initial prediction model to grid edge devices; each regional device uses local historical data to train a local model, and uploads the local model training results in an encrypted manner; aggregating the encrypted uploaded local model training results and integrating them into a global model; and regularly pushing the global model to grid edge devices, which use the global model to predict grid status.

[0012] Optionally, the use of a federated learning model to predict grid status also includes: if the predicted grid status exceeds a set second error threshold three or more times in a row, a global model update will be triggered, and the second error threshold will be dynamically adjusted according to the error fluctuation.

[0013] Optionally, the execution of grid adaptive optimization and control to manage terminal low voltage includes: constructing a future grid state prediction model, and predicting the future voltage based on the future grid state prediction model; measuring the actual voltage at this time, calculating the difference between the future voltage and the actual voltage measurement value, and determining whether it is greater than a third voltage error threshold; if it is greater than the third voltage error threshold for three consecutive times, triggering an update; using an exponentially weighted moving average to update the third voltage error threshold; and changing each weight in the quantum genetic algorithm according to the updated third voltage error threshold.

[0014] Optionally, executing grid adaptive optimization and regulation to control terminal low voltage also includes: allocating corresponding weight parameters and trigger conditions to different fault types according to the characteristics and impact of the grid fault type.

[0015] On the other hand, the present invention provides a terminal low-voltage management system without current sampling, which is used to implement a terminal low-voltage management method without current sampling. The terminal low-voltage management system without current sampling includes: a control module, the control module includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executes the computer program to implement the terminal low-voltage management method without current sampling.

[0016] The above technical solution, by deploying high-precision sensors, collects electromagnetic field data, voltage, current and other multi-dimensional information of the power grid in real time, covering each node of the power grid; avoiding the data loss problem caused by insufficient sampling points in the traditional current sampling method. By constructing an impedance inversion model based on field distribution characteristics and combining it with a quantum adaptive impedance matching algorithm, the load impedance characteristics of the power grid can be accurately derived; the accuracy of impedance derivation is improved, and at the same time, the multi-node calculation complexity is reduced through block impedance matrix and prior constraints, ensuring the efficiency and real-time performance of the calculation. The actual impedance value is measured using a four-wire measurement method to ensure that the adjusted impedance value is consistent with the target value, and a PID closed-loop compensation mechanism is introduced. When the error between the actual impedance value and the target value exceeds the set threshold, the impedance value is automatically adjusted to ensure the reliability and stability of the adjustment; in addition, by integrating the local model training results of the equipment in each region, a global model is constructed to improve the generalization ability and real-time performance of the power grid state prediction. Based on the prediction results, adaptive optimization and control are performed to ensure that the power grid can maintain stable operation under different load and environmental conditions, and effectively control the low voltage problem at the end.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings: Figure 1 It is a flow chart of terminal low voltage management without current sampling.

[0019] Figure 2 It is the impedance adjustment flow chart.

[0020] Figure 3 It is a flowchart for determining whether to update the global model. DETAILED DESCRIPTION

[0021] The following is combined with Figure 1 - Attachment Figure 3 The specific implementation of the embodiment of the present invention is described in detail. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0022] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. In the embodiments of this application, some existing solutions in the industry such as certain software, components, and models may be mentioned, which should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.

[0023] Embodiment 1 Reference Figures 1 to 3 , which is the first embodiment of the present invention, and provides a method for managing low voltage at the terminal without current sampling, comprising: S100: Establishing a global sensing sensor network, and using the sensor network to collect power grid data in real time.

[0024] Specifically, magnetic field sensors, electric field sensors, phase voltage sensors, etc. are installed in transformers, line joints and load centers in key areas such as substations and various distributed power sources, as well as low-voltage distribution cabinets, smart meters and other terminal nodes to monitor electric fields, magnetic fields and parameters related to power transmission in real time. Sampling through the above sensors can accurately capture subtle changes in the operation of the power grid and provide detailed basic data for subsequent data processing and status analysis.

[0025] Preferably, synchronous sampling technology is used to collect power grid data to ensure that the power data at different detection points are compared under the same time reference system to ensure the consistency of the collected data. A monitoring network with high spatial coverage is formed through each monitoring point to capture subtle changes in the power transmission process and ensure that important status information can be recorded in a timely manner.

[0026] Preferably, the sensor types and locations are reasonably selected and arranged in consideration of the characteristics and applicable scope of different sensors. For example, a high-precision magnetic field sensor is installed near the transformer to accurately monitor the operating status of the transformer; an electric field sensor is installed at the key nodes of the transmission line to grasp the changes in the electric field strength in real time.

[0027] Preferably, by installing multiple sensors in key areas and terminal nodes of the power grid and reasonably arranging sensor types and locations according to the characteristics and applicable scope of different sensors, high accuracy of monitoring data can be ensured and comprehensive perception of the operating status of the power grid can be achieved; power grid data is collected in real time to ensure that subtle changes in power grid operation can be captured in a timely manner; synchronous sampling technology is used to ensure that data from different detection points are compared under the same time reference system, thereby improving data consistency.

[0028] S200: Perform load impedance inversion through an edge computing node according to the collected power grid data, and generate a virtual impedance adjustment scheme according to the result of the load impedance inversion.

[0029] Preferably, after data collection is completed, the collected raw data is preprocessed using a filter, and the preprocessed data is transmitted to the edge computing node in real time through a high-speed SPI interface, and the sampling frequency is set to 1MHz to ensure data collection at high resolution. The high-speed and high-precision data transmission method can effectively reduce the loss and distortion of data during transmission, and ensure the integrity and accuracy of the data.

[0030] Preferably, each edge computing node is not only responsible for data collection and preprocessing, but also deploys preliminary intelligent learning components, so that the edge computing node can run simple machine learning models and achieve preliminary analysis and learning of local data.

[0031] Furthermore, at the edge computing node, random noise is further removed through digital filtering algorithms (such as Kalman filtering). Time series data denoising and feature extraction algorithms (such as wavelet transform combined with principal component analysis dimensionality reduction method) are used to remove redundant information and noise, and high-noise, high-redundancy data is converted into understandable grid status signals. The spatial magnetic field and electric field distribution of each node is obtained to obtain a high-precision scenario data set, providing accurate and immediate grid operation status feedback for the follow-up.

[0032] Furthermore, the load impedance inversion without current sampling based on field data relies only on the spatiotemporal distribution of the electric field and the magnetic field, and estimates the equivalent load impedance by solving the inverse problem. The load impedance inversion formula without current sampling is as follows: in, is the sensor network coverage area, is the equivalent load impedance, expressed in The equivalent load impedance calculated based on the electromagnetic field data reflects the comprehensive grid impedance characteristics of the area. represents the electric field vector measured at spatial position r and time t, represents the magnetic field vector measured at the same position and time, Represents the dielectric constant, which characterizes the response of a medium to an electric field.

[0033] Preferably, a spatial interpolation algorithm is introduced to complete the field data of unmonitored points. The interpolation weights are calculated for the areas not covered by the sensor network based on the known point data. After the interpolation is completed, the electric field direction is checked to see if it is consistent with the conductor direction. When new sensors are added or the grid topology changes, the interpolation is recalculated.

[0034] Preferably, the calculated interpolation is input into a subsequent quantum genetic algorithm to optimize the virtual impedance. is the control target, the equivalent load impedance is the actual impedance of the grid calculated by inversion.

[0035] Furthermore, the optimal virtual impedance value is found through the quantum adaptive impedance matching algorithm. In the current-free sampling scheme, the virtual impedance adjustment needs to be optimized through field measurement feedback, and the virtual impedance optimization objective function is as follows: in, represents the virtual impedance, and represent the reference electric field and the reference magnetic field under ideal conditions respectively; and They respectively represent the actual measured electric field strength and the actual measured magnetic field strength under the current virtual impedance adjustment state; is the weight factor of the field ratio error term, reflecting the importance of electromagnetic field matching in optimization; is the weight factor of the harmonic distortion term, which is used to control the priority of power quality optimization; is the weight factor of the impedance matching item, which is used to ensure the consistency between the virtual impedance and the equivalent impedance of the power grid; THD is the total harmonic distortion rate, is the effective value of the fundamental voltage, is the effective value of the hth harmonic voltage, where h is the highest harmonic order considered, usually 50th.

[0036] Furthermore, the quantum genetic algorithm is used to update the quantum bits (corresponding to the impedance parameters) of each chromosome. The update formula is as follows: in, represents the phase angle of the quantum bit at the kth iteration, represents the phase angle at k+1 iterations, is the step size factor, indicating the magnitude of each update, and F represents the fitness function. Represents virtual impedance.

[0037] Preferably, by updating and fine-tuning the phase angle of the quantum bit, the corresponding virtual impedance gradually converges to the optimal solution, thus preventing the algorithm from falling into a local optimum.

[0038] Preferably, FPGA is used to realize parallel calculation of quantum chromosomes, which significantly improves the optimization speed. In order to avoid falling into local optimum, a random perturbation mechanism is introduced to fine-tune some chromosomes after each iteration, and the maximum number of iterations is set to ensure the convergence of the algorithm; the optimal virtual impedance value is output.

[0039] Preferably, data processing is performed through edge computing nodes to reduce loss and distortion during data transmission and improve data processing efficiency; filtering algorithms and feature extraction algorithms are used to remove noise and redundant information and improve the accuracy of data processing; virtual impedance is dynamically optimized through quantum genetic algorithms and adaptive filtering algorithms to generate virtual impedance solutions and avoid falling into local optimality. Current-free sampling inversion technology also simplifies the traditional current measurement process, reducing costs and complexity.

[0040] S300: According to the virtual impedance adjustment scheme, impedance matching verification is performed, the actual impedance value is verified by a four-wire measurement method, and impedance adjustment is performed.

[0041] Preferably, the measured electric field amplitude is used as the "voltage" information, and the right-hand side term (including displacement current) in Maxwell's equations reflects the "virtual current" effect that is not directly sampled. The ratio of the two is used as the estimated value of the load equivalent impedance, and the single-point noise effect is reduced by integration. When performing load impedance inversion without current sampling, prior knowledge (such as power grid topology) is introduced as a constraint condition. For multi-node systems, the block matrix method is used to reduce the computational complexity.

[0042] Furthermore, the virtual impedance is adjusted through the hardware GaN switch matrix and MEMS capacitor array. The GaN switch matrix improves the impedance adjustment accuracy by controlling the switch state; the MEMS capacitor array improves the voltage support capability by adjusting the reactive power compensation. The priority strategy here needs to ensure that voltage regulation takes precedence over harmonic compensation.

[0043] Furthermore, after completing the hardware adjustment, the actual impedance value is verified using the four-wire measurement method. When the measurement error is greater than the set threshold, the PID closed-loop compensation algorithm is triggered to automatically adjust the impedance value. The PID controller parameters are calibrated through experiments to ensure fast convergence and stability. The adjusted impedance value is recorded as historical data for subsequent federated learning.

[0044] Preferably, the grid status data is collected and calculated again to verify the regulation effect. If the error exceeds the set threshold, a second regulation (fine-tuning) is performed until the error converges to the set target. Such a feedback closed-loop mechanism ensures the adaptive ability of the grid during operation, and also provides valuable real-time data for subsequent optimization algorithms, making the entire regulation process more robust and reliable.

[0045] Preferably, the virtual impedance is adjusted through hardware (GaN switch matrix and MEMS capacitor array) to improve the adjustment accuracy and voltage support capability; the actual impedance value is verified by the four-wire measurement method, and combined with the PID closed-loop compensation algorithm to ensure that the adjusted impedance value converges quickly and remains stable; a priority strategy is formulated to ensure that voltage regulation takes precedence over harmonic compensation to improve the stability of grid operation; the grid status data is collected and calculated in real time to verify the adjustment effect and make fine adjustments to ensure the adaptive capability of the grid.

[0046] S400: After the impedance adjustment is completed, a federated learning model is used to predict the state of the power grid, and based on the result of the power grid state prediction, a power grid adaptive optimization control is performed to control the low voltage at the terminal.

[0047] Specifically, each edge computing node collects the grid operation data of the area it is responsible for, and trains the regional grid regulation model through the adaptive neural network (ANN). These regional grid regulation models can predict and respond to changes in the grid state, providing immediate reference for regulation decisions.

[0048] Furthermore, the technical framework of federated learning is used to collect the power grid operation experience of various regions. Each edge computing node transmits the learning gradient through encryption technology while maintaining the independence and privacy of local data. The cloud server summarizes and integrates the learning results of each node to build a more accurate and universal global regulation model, and regularly updates the edge devices. The model is pushed once every set time interval to enhance the real-time and adaptive capabilities of the model. The global regulation model not only improves the accuracy of power grid regulation, but also greatly enhances the adaptive capabilities of the entire system in the face of unknown situations. Over time, the global regulation model gradually forms the ability to identify and respond to various abnormal conditions in the process of continuously accumulating data and practical experience, and finally realizes true self-learning, adaptation and self-optimization.

[0049] Furthermore, after fully perceiving and adjusting the current state, a future grid state prediction model is constructed through a deep learning algorithm using historical data as a basis. The formula of the future grid state prediction model is as follows: in, Indicates the predicted The voltage value at the moment, Represents the nonlinear mapping function composed of LSTM network, which is obtained through training of historical data; represents the voltage history sequence of the past T time steps, Indicates the virtual impedance value; Indicates the power output of the photovoltaic system and reflects the power supply status of the distributed power source; Indicates the time step of the prediction, which is the future prediction duration.

[0050] Preferably, the future grid state prediction model integrates the historical operation data of the grid, the virtual impedance state and the distributed power source information, and predicts the future voltage changes through a deep learning model to provide a basis for dynamic compensation.

[0051] Preferably, each edge device predicts the future voltage value in real time and compares it with the actual measured value. If the prediction error exceeds the safety threshold three times in a row, it means that the current model can no longer accurately reflect the state of the power grid and must be updated immediately. For example, if a regional model predicts that the voltage in the next minute will be 10kV, but the actual measured voltage is 10.5kV, if the error exceeds the threshold and occurs three times in a row, an update is triggered.

[0052] Furthermore, in order to balance the stability and adaptability of the prediction model, the exponentially weighted moving average (EWMA) method is used to dynamically adjust the threshold.

[0053] Furthermore, a baseline threshold is set based on historical data, such as the maximum allowable error under normal operating conditions, and the latest prediction error is weighted. The weight of new data is 10% (configurable), and the weight of historical data is 90%. Each time new prediction error data is received, the long-term trend value and fluctuation range value are automatically updated, and the final threshold = long-term trend value + 2.5 times the fluctuation range value. If the threshold is triggered three times in a row, the system automatically increases the weight of new data to 20% to accelerate the capture of continuous anomalies. If it is not triggered for 24 consecutive hours, the weight is gradually reduced to 5% to improve stability.

[0054] Preferably, differentiated weight parameters and trigger conditions are preset for different fault types such as lightning strike, overload, harmonics, etc. (for example, an overload needs to exceed the threshold for five consecutive times to be triggered, while a lightning strike only needs to exceed the threshold once).

[0055] Preferably, a response strategy is formulated based on the predicted results of the future state, involving dynamic adjustment of virtual impedance, adjustment of the output of reactive compensation equipment, and reasonable distribution of loads. After virtual simulation and system simulation, the effects of different strategies are evaluated and compared to select the most practical and economically effective solution. This multi-dimensional strategy formulation method ensures that the system has sufficient flexibility and adaptability when facing complex power grid problems. Before the strategy is officially applied, the selected solution is fully tested using a simulation platform (MATLAB). The simulator is used to build an operating environment similar to the actual power grid, and the actual effect of the compensation strategy is fully evaluated to ensure the robustness and security of the overall solution.

[0056] Furthermore, after determining that the overall solution technology is mature, a representative area is selected as a pilot. The selection of the pilot area needs to take into account the complexity and particularity of the power grid operating environment in the area to ensure that the pilot results have broad reference value. In the pilot area, sensors, edge computing nodes, and control execution devices are fully deployed in accordance with the requirements of the solution, and a reliable communication network is established to achieve seamless collaboration between devices.

[0057] Furthermore, after the pilot deployment is completed, a longer period of operation observation will be entered. During this stage, the operation status will be comprehensively checked through real-time monitoring, regular evaluation and data feedback mechanism. During the operation, not only the improvement of voltage stability and harmonic indicators will be paid attention to, but also various emergencies and their response effects will be recorded to provide reliable data for subsequent optimization.

[0058] Preferably, a detailed operation log and evaluation report system should be established to regularly analyze and summarize the pilot operation data, promptly identify potential problems and make suggestions for improvement.

[0059] Preferably, after the pilot is completed, the entire program will be summarized and evaluated to compile a complete set of operating experience and technical specifications. Combined with the problems and improvement suggestions found during the pilot process, the details of the program will be further improved to form technical standards and promotion guidelines that meet industry requirements.

[0060] Preferably, a future grid state prediction model is constructed through deep learning algorithms, such as LSTM networks, which integrates historical data, virtual impedance state and distributed power supply information to improve the accuracy of prediction; the federated learning framework integrates multi-regional experience to build a global model while protecting data privacy, improving the adaptive ability of prediction and regulation. The threshold is dynamically adjusted through the exponentially weighted moving average method (EWMA) to balance the stability and adaptability of the prediction model; for different fault types, differentiated weight parameters and trigger conditions are preset to ensure that the system has sufficient flexibility and adaptability in the face of complex grid problems.

[0061] The present invention also provides a terminal low voltage management system without current sampling, which is used to implement a terminal low voltage management method without current sampling. The system includes a control module, and the control module includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the terminal low voltage management method without current sampling.

[0062] An embodiment of the present invention provides a storage medium on which a program is stored. When the program is executed by a processor, a method for terminal low voltage management without current sampling is implemented.

[0063] An embodiment of the present invention provides a processor, which is used to run a program, wherein the program executes a method for terminal low voltage management without current sampling when running.

[0064] The embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored in the memory and executable on the processor, and the above-mentioned related steps are implemented when the processor executes the program. The device in this article can be a server, a PC, a PAD, a mobile phone, etc.

[0065] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program initialized with the above-mentioned related steps.

[0066] Those skilled in the art will appreciate that the embodiments of the present application may provide methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0068] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for managing low voltage at the end without current sampling, characterized in that: include: Establishing a global sensing sensor network and using the sensor network to collect power grid data in real time; According to the collected power grid data, load impedance inversion is performed through an edge computing node, and a virtual impedance adjustment scheme is generated according to a result of the load impedance inversion; According to the virtual impedance adjustment scheme, impedance matching verification is performed, the actual impedance value is verified by a four-wire measurement method, and impedance adjustment is performed; After the impedance adjustment is completed, the federated learning model is used to predict the state of the power grid, and based on the results of the power grid state prediction, the power grid adaptive optimization control is performed to control the low voltage at the end.

2. The method for managing low voltage at the terminal without current sampling according to claim 1, characterized in that: The load impedance inversion by the edge computing node includes: Preprocess the electromagnetic field data collected by the sensor, and use the preprocessed data to build an impedance inversion model based on the field distribution characteristics; Define the virtual impedance optimization objective function, use the grid topology as a priori constraint, and establish a block impedance matrix to reduce the multi-node calculation complexity; The optimal virtual impedance value is calculated using a quantum adaptive impedance matching algorithm; A quantum genetic algorithm is used to update the phase angle of the quantum bit according to the calculated optimal virtual impedance value, so that the optimal virtual impedance value gradually converges to the optimal solution.

3. The method for managing low voltage at the terminal without current sampling according to claim 2, characterized in that: The step of making the optimal virtual impedance value gradually converge to the optimal solution includes: using FPGA to perform parallel calculation of quantum chromosomes and introducing a random perturbation mechanism; adjusting part of the chromosomes after each iteration; and setting a maximum number of iterations.

4. The method for managing low voltage at the terminal without current sampling according to claim 1, characterized in that: The load impedance inversion through the edge computing node also includes introducing a spatial interpolation algorithm to complete the field data of the unmonitored points; the introduction of the spatial interpolation algorithm includes: calculating the interpolation weight for the area not covered by the sensor network based on the known point data, and verifying whether the direction of the electric field is consistent with the direction of the conductor after the interpolation is completed; when there are new sensors or the grid topology changes, recalculating the interpolation; and inputting the calculated interpolation into the quantum genetic algorithm to optimize the virtual impedance.

5. The method for managing low voltage at the terminal without current sampling according to claim 1, characterized in that: According to the virtual impedance adjustment scheme, impedance matching verification is performed, the actual impedance value is verified by a four-wire measurement method, and impedance adjustment is performed, including: Configure the GaN switch matrix for impedance adjustment; Deploy MEMS capacitor arrays for dynamic reactive power compensation; Use the four-wire measurement method to verify the actual impedance value. When the error between the actual impedance value and the target impedance value is greater than the set first threshold, the PID closed-loop compensation is triggered to automatically adjust the impedance value. The grid data is collected and calculated again. If the error between the actual impedance value and the target impedance value is still greater than the first threshold, a second adjustment is performed until the error between the actual impedance value and the target impedance value is less than or equal to the first threshold.

6. The method for managing low voltage at the terminal without current sampling according to claim 1, characterized in that: The use of the federated learning model to predict the power grid state includes: Design an initial prediction model and distribute the initial prediction model to grid edge devices; Each regional device uses local historical data to train the local model and uploads the local model training results in encrypted form; Summarize the encrypted uploaded local model training results and integrate them into a global model; The global model is regularly pushed to the grid edge devices, which use the global model to predict the grid status.

7. The method for managing low voltage at the terminal without current sampling according to claim 6, characterized in that: The use of the federated learning model to predict the power grid state also includes: if the predicted power grid state exceeds the set second error threshold three or more times in a row, it will trigger a global model update, and dynamically adjust the second error threshold according to the error fluctuation.

8. The method for managing low voltage at the terminal without current sampling according to claim 1, characterized in that: The performing of grid adaptive optimization regulation to control the low voltage at the end includes: Construct a future grid state prediction model, and predict the future voltage based on the future grid state prediction model; Measure the actual voltage at this time, calculate the difference between the future voltage and the actual voltage measurement value, and determine whether it is greater than a third voltage error threshold; If it is greater than the third voltage error threshold for three consecutive times, an update is triggered; Adopting exponentially weighted moving average to update the third voltage error threshold; According to the updated third voltage error threshold, each weight in the quantum genetic algorithm is changed.

9. The method for managing low voltage at the terminal without current sampling according to claim 8, characterized in that: The performing of adaptive optimization and control of the power grid to control the low voltage at the terminal also includes: assigning corresponding weight parameters and trigger conditions to different fault types according to the characteristics and impact of the power grid fault types.

10. A terminal low voltage management system without current sampling, characterized in that: The system includes a control module, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the method for terminal low voltage management without current sampling according to any one of claims 1-9.

Images