SVG compensation system and method based on reactive power closed loop and cloud cooperation

Abstract

The present application belongs to the technical field of electronic power, and particularly relates to an SVG compensation system and method based on reactive power closed loop and cloud cooperation, which comprises a high-voltage side acquisition unit, a low-voltage side static var generator and a cloud platform. The acquisition unit acquires real-time reactive power and harmonic data of the high-voltage side and uploads the data. The cloud platform calculates target reactive power based on the data and a preset power factor target value, takes the deviation between the real-time value and the target value as the core control quantity, combines voltage and harmonic suppression requirements, solves the multi-objective optimization problem through a model predictive control algorithm based on a discrete state space model, generates a cooperative compensation instruction meeting the equipment and safety constraints, and issues the instruction to the SVG for execution. The SVG feeds back new data to the cloud after execution, which is used for adaptive updating of model parameters. The system ensures fast and accurate linkage between high-voltage side sampling and low-voltage side compensation, and avoids the frequent misadjustment and voltage flicker problems in the traditional high-sampling and low-compensation scheme.

Description

Technical Field

[0001] This invention belongs to the field of electronic power technology, specifically relating to an SVG compensation system and method based on reactive power closed-loop and cloud collaboration. Background Technology

[0002] With the large-scale grid connection of new energy power generation and the continuous increase in the proportion of power electronic loads, the reactive power in the distribution network exhibits complex characteristics of rapid fluctuation and bidirectional flow, leading to increased line losses and deterioration of voltage quality. Traditional static reactive power compensation devices are no longer sufficient to meet the needs of dynamic management. The current mainstream dynamic reactive power compensation technologies mainly follow two architectures: one is "high-voltage side sampling and high-voltage side compensation" (high sampling and high compensation), which involves directly installing a static var generator (SVG) on the high-voltage bus side of the transformer. Although this scheme can directly adjust the power quality on the high-voltage side, it has poor economic efficiency and flexibility due to the high insulation level of the equipment, large investment costs, and the need for power outages for maintenance; the other is "low-voltage side sampling and low-voltage side compensation" (low sampling and low compensation), which involves installing a compensation device on the low-voltage side of the transformer. Although it has lower costs and is easy to install, its sampling and control are limited to the low-voltage side, and it cannot sense and compensate for the reactive power of the transformer body excitation and the reactive power of the high-voltage side line-to-ground distributed capacitance, resulting in a management blind spot. To balance economic efficiency and governance scope, a hybrid approach of "high-voltage side sampling and low-voltage side compensation" (high sampling and low compensation) has emerged in the existing technology. This approach collects the high-voltage side current and low-voltage side voltage in the low-voltage side controller and performs closed-loop PID regulation with the high-voltage side average power factor as the target, attempting to incorporate the reactive power state of the high-voltage side into the control loop.

[0003] However, this type of high-voltage sampling and low-voltage compensation scheme still has significant limitations: First, its control objective is singular, focusing solely on the power factor. The power factor is affected by both active and reactive power components. When active power fluctuates drastically due to renewable energy output or impulsive loads, even if reactive power remains unchanged, the power factor will change significantly, leading to controller malfunctions, unnecessary adjustments, or even oscillations, resulting in poor dynamic response performance. Second, its system architecture is rigid, relying on a controller with fixed local parameters. It lacks multi-objective coordination and global optimization capabilities, failing to effectively suppress voltage fluctuations and background harmonics while compensating for reactive power. In complex scenarios, frequent switching may even exacerbate power quality problems. Furthermore, existing schemes suffer from communication delays and time asynchrony between high-voltage side sampling and low-voltage side compensation command execution, affecting compensation accuracy. Therefore, how to achieve more accurate, faster, and multi-objective coordinated dynamic reactive power compensation while maintaining the economical high-voltage sampling and low-voltage compensation architecture has become a pressing technical problem to be solved in this field. Summary of the Invention

[0004] The purpose of this invention is to provide an SVG compensation system and method based on reactive power closed-loop and cloud collaboration, so as to solve the problems of existing high-collection and low-compensation reactive power compensation schemes with single control objectives and insufficient dynamic response, which make it difficult to achieve coordinated optimization and management of reactive power, voltage fluctuation and harmonic content in complex power grid scenarios.

[0005] The present invention achieves the above objectives through the following technical solutions:

[0006] Firstly, this invention proposes an SVG compensation system based on reactive power closed-loop and cloud collaboration, applied to a topology architecture of high-voltage side sampling and low-voltage side compensation of transformers. The system includes a data acquisition unit, a static var generator, and a cloud platform that wirelessly interacts with both.

[0007] The acquisition unit is connected to the high-voltage side to acquire and process the electrical quantities of the power grid on the high-voltage side, so as to obtain the first state data containing the real-time reactive power value and harmonic analysis data of the high-voltage side.

[0008] The cloud platform is used to receive real-time updated first status data, and based on the real-time reactive power value of the high-voltage side and the preset power factor target value in the first status data, to determine the target reactive power value of the high-voltage side; using the deviation between the real-time reactive power value and the target reactive power value as the core control quantity, and generating reactive power compensation instructions through a preset multi-objective optimization algorithm;

[0009] The static var generator is connected to the low-voltage side to receive and respond to the reactive power compensation command, determine and output the corresponding compensation current to the low-voltage side of the transformer to compensate the reactive power on the high-voltage side.

[0010] The acquisition unit is also used to acquire the electrical quantities of the power grid on the high-voltage side after the static var generator performs compensation, and send the updated first state data to the cloud platform.

[0011] Furthermore, the cloud platform includes:

[0012] The data receiving and synchronization unit is used to receive and align first state data from the acquisition unit and second state data from the low-voltage side static var generator based on a unified time reference. The second state data includes the IGBT module temperature, DC bus voltage, cooling fan speed and actual output compensation current data of the static var generator.

[0013] The data processing unit is used to generate reactive power compensation instructions through a preset multi-objective optimization algorithm, and to maintain and optimize the multi-objective optimization algorithm online based on the measured data fed back after the instructions are executed; wherein, the multi-objective optimization algorithm includes: in each control cycle, generating the reactive power compensation instructions on a rolling basis based on a preset discrete state space model and applying preset constraints.

[0014] The health management unit is used to predict lifespan and provide early warning of failures based on the second status data.

[0015] Furthermore, the data processing unit generates the reactive power compensation instruction by including the following steps:

[0016] Calculate the target reactive power value on the high-voltage side based on the current real-time reactive power value on the high-voltage side and the preset power factor target value;

[0017] Using the deviation between the real-time reactive power value and the target reactive power value as the reactive power control benchmark, and combining the harmonic content suppression requirements and voltage deviation suppression requirements based on the harmonic analysis data in the first state data, a multi-objective optimization problem in the finite time domain is constructed.

[0018] In each control cycle, the optimization problem is solved in a rolling manner according to a multi-objective optimization algorithm to generate a reactive power compensation command that meets the preset constraints.

[0019] Furthermore, in the discrete state-space model, the output variables include the reactive power of the high-voltage side bus of the transformer, the voltage at the common coupling point connecting the static var generator, and the amplitude or effective value of the characteristic harmonic current of a specific order; the model parameters are updated through online system identification, and the input data for online system identification includes the following time-aligned historical sequences:

[0020] The reactive power compensation command sequence issued to the low-voltage side static var generator, the high-voltage side reactive power sequence obtained by the high-voltage side acquisition unit, and the common coupling point voltage and characteristic harmonic current sequence obtained by the low-voltage side static var generator.

[0021] Furthermore, the discrete state-space model includes the following equation:

[0022] ;

[0023] Where k is the current time, The system state vector includes the high-voltage side reactive power deviation, the common coupling point voltage deviation, and the characteristic harmonic current deviation. The control input vector represents the adjustment amount of the reactive power output of the static var generator; This represents the deviation of the load disturbance. , The measured value of the load active power at time k. For the k-th control cycle, the power factor angle of the total load on the low-voltage side of the transformer is... Let k be the 5th harmonic current generated by the load at time k. The 7th harmonic current generated by the load at time k; These are model output variables; , , , , This is the model matrix, whose parameters are updated through online system identification.

[0024] Furthermore, the optimization objective of the multi-objective optimization problem includes a composite objective function J, specifically:

[0025] ;

[0026] in, For the future control sequence to be optimized, , To control the time domain, ;

[0027] For reactive power tracking, ,in, The weighting coefficients for the reactive power tracking term. To base on discrete state-space model The predicted value of reactive power on the high-voltage side at any given time. Yes Time based on active power prediction value Calculated future target value;

[0028] This is a voltage deviation penalty term. ,in, The weighting coefficient for the voltage deviation penalty term. This is a predicted value for the point of common coupling voltage connected to the static var generator. This is the voltage reference value;

[0029] This is a harmonic suppression term. , These are the weighting coefficients for each harmonic in the harmonic suppression term. This is a predicted estimate of the effective value of the mth harmonic current on the high-voltage side at time k+i in the future.

[0030] To control the incremental smoothing term, ,make This is the actual execution value from the previous cycle. The amount of reactive power adjustment applied to the SVG output at time k+i in the future. Let be the reactive power regulation amount of the SVG applied at time k+i preceding time, where To control the weighting coefficients of the incremental smoothing term.

[0031] Furthermore, the preset constraints include system dynamic constraints, control input constraints, control increment constraints, and output safety constraints; wherein:

[0032] The predicted future state variables in the system's dynamic constraints satisfy the dynamic relationships defined by the discrete state-space model;

[0033] Control input constraints are ,in , The baseline value for SVG output at the current working point. , These represent the maximum inductive reactive power and the maximum capacitive reactive power that the SVG device can continuously output, respectively. The amount of reactive power to be adjusted for the SVG output in the i-th step in the future;

[0034] Control incremental constraints are ,in , The minimum allowable change in SVG reactive power regulation between two consecutive control cycles. The maximum allowable change in SVG reactive power regulation between two consecutive control cycles;

[0035] Output safety constraints are , ,in , As the lower limit of voltage safety, To ensure voltage safety, an upper limit is set. The predicted value of the effective value of the voltage at point PCC at time k+i in the future. Upper limit of total harmonic distortion, The predicted effective value of the m-th harmonic current at time k+i in the future. The predicted value of the fundamental current at time k+i in the future.

[0036] Furthermore, the rolling solution process includes:

[0037] The objective function and constraints of the multi-objective optimization problem are mathematically reconstructed in the form of quadratic programming to generate the corresponding coefficient matrix and vector.

[0038] The preset quadratic programming solver is invoked to numerically solve the coefficient matrix and vector to obtain the optimal control increment sequence in the future finite time domain;

[0039] The control increment corresponding to the current moment is extracted from the optimal control increment sequence, and combined with the control output of the previous moment, the current reactive power compensation command is calculated.

[0040] Furthermore, the rolling solution process also includes:

[0041] In the current control cycle k, the reactive power compensation instruction and its corresponding execution timestamp are sent to the static var generator, and the received confirmation is received from it.

[0042] At a preset sampling time corresponding to the execution of the instruction, the data receiving and synchronization unit collects the compensated electrical quantity measurement data from the high-voltage side acquisition unit, as well as the actual output current data from the static var generator.

[0043] The error between the compensated electrical quantity measurement data and the corresponding predicted value of the previous cycle is calculated, and the error is input into the state estimator to update the estimate of the internal state of the system.

[0044] If the prediction error exceeds a preset threshold, the online system identification process of the discrete state space model is triggered.

[0045] Using the corrected system state as the new initial state, the prediction time domain is rolled forward by one control cycle, and the optimization calculation for the next cycle k+1 is started.

[0046] Secondly, this invention proposes an SVG compensation method based on reactive power closed-loop and cloud-based collaboration, executed based on the aforementioned SVG compensation system. The method includes the following steps:

[0047] S1: By connecting to the acquisition unit on the high-voltage side of the transformer, the electrical quantities of the high-voltage side power grid are obtained, the real-time reactive power value and harmonic analysis data on the high-voltage side are calculated, and packaged into first-state data;

[0048] S2: Upload the first status data to the cloud platform via wireless communication;

[0049] S3: In the cloud platform, a target reactive power value is determined based on the real-time reactive power value of the high-voltage side and the preset power factor target value in the first state data; the deviation between the real-time reactive power value and the target reactive power value is used as the core control quantity, and a reactive power compensation command is generated through a preset multi-objective optimization algorithm.

[0050] S4: Send the reactive power compensation command to the static var generator connected to the low-voltage side of the transformer;

[0051] S5: Control the static var generator to output the corresponding compensation current according to the reactive power compensation command;

[0052] S6: The power grid electrical quantities on the high-voltage side after compensation are collected again by the acquisition unit, new first state data is generated and fed back to the cloud platform;

[0053] S7: In the cloud platform, the parameters or optimization strategy of the discrete state space model are corrected and updated online by using the deviation between the first state data fed back in S6 and the predicted value in S3.

[0054] The beneficial effects of this invention are as follows:

[0055] 1. By changing the control target from power factor to reactive power on the high-voltage side, this invention directly eliminates the interference of active power changes caused by photovoltaic output fluctuations or motor start-stop to the control system, thereby avoiding the frequent misregulation and voltage flicker problems in the traditional high-voltage sampling and low-voltage compensation scheme.

[0056] 2. Through the model predictive control algorithm deployed on the cloud platform, the system can predict changes in the power grid state in advance based on the state space model, and calculate the optimal compensation command that enables reactive power to quickly track the target while suppressing bus voltage fluctuations and specific subharmonics, thus achieving multi-control with one machine.

[0057] 3. This invention utilizes measured data after each control cycle to update model parameters and correct state online, enabling it to adapt to the slow changes in power grid structure and load characteristics, maintaining long-term control accuracy. A high-precision time synchronization mechanism solves the communication delay asynchrony problem between high-voltage side sampling and low-voltage side compensation commands, ensuring the spatiotemporal consistency of control actions. Attached Figure Description

[0058] Figure 1 This is a system block diagram of an SVG compensation system based on reactive power closed-loop and cloud collaboration in this invention;

[0059] Figure 2 This is a system block diagram of a cloud platform generating reactive power compensation instructions in this invention;

[0060] Figure 3 This is a flowchart of an SVG compensation method in this invention. Detailed Implementation

[0061] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0062] This disclosure proposes a specific embodiment of an SVG compensation system based on reactive power closed-loop and cloud-based collaboration, applied to a transformer high-voltage side sampling and low-voltage side compensation topology. The system includes a high-voltage side acquisition unit, a cloud platform, a wireless communication unit, and a low-voltage side static var generator (SVG). The wireless communication unit ensures wireless communication interaction between the acquisition unit, the SVG, and the cloud platform. The core solution of this disclosure is to directly target the high-voltage side reactive power as the control object, generating precise control commands by calculating the reactive power deviation in real time and combining it with the intelligent control algorithm proposed in this disclosure. This system not only achieves complete identification and full compensation of load reactive power, transformer excitation reactive power, and high-voltage side-to-ground capacitance reactive power, but also addresses multiple power quality issues such as reactive power, voltage fluctuations, and harmonic content. Through the edge-cloud collaborative architecture and high-precision time synchronization mechanism, the system ensures rapid and accurate linkage between high-voltage side sampling and low-voltage side compensation.

[0063] Please see Figure 1 and Figure 2 Specifically, the acquisition unit is connected to the high-voltage side to acquire and process the electrical quantities of the power grid on the high-voltage side, including the voltage and current signals of the high-voltage bus. It performs real-time calculations to obtain the real-time reactive power value and harmonic analysis data of the high-voltage side. The harmonic analysis data includes at least the amplitude or effective value of a preset major harmonic (e.g., the 5th and 7th harmonics). This data is packaged into first-state data. The cloud platform receives the real-time updated first-state data and, based on the real-time reactive power value of the high-voltage side and the preset power factor target value in the first-state data, calculates the reactive power value using the formula... The target reactive power value on the high-voltage side is calculated, where P(k) is the active power on the high-voltage side measured in real time. The preset power factor target value is used as the core control variable, and the deviation between the real-time reactive power value and the target reactive power value is used as the core control variable. Combined with the harmonic analysis data in the first state data, a preset multi-objective optimization algorithm is used to generate reactive power compensation instructions. The static var generator is connected to the low-voltage side to receive and respond to the reactive power compensation instructions, determine and output the corresponding compensation current to the low-voltage side of the transformer to compensate the reactive power on the high-voltage side.

[0064] The acquisition unit is also used to acquire the electrical quantities of the power grid on the high-voltage side after the static var generator performs compensation, and send the updated first state data to the cloud platform.

[0065] As a preferred embodiment, the cloud platform includes a data receiving and synchronization unit, a data processing unit, and a device health management unit.

[0066] The data receiving and synchronization unit is used to receive and align first state data from the acquisition unit and second state data from the SVG based on a unified time reference. The second state data includes the IGBT module temperature, DC bus voltage, cooling fan speed, and actual output compensation current data of the static var generator. The data is aligned based on a unified time reference to eliminate errors caused by communication delays.

[0067] The data processing unit is used to generate reactive power compensation instructions through a preset multi-objective optimization algorithm, and to maintain and optimize the multi-objective optimization algorithm online based on the measured data fed back after the instructions are executed. The multi-objective optimization algorithm includes: in each control cycle, the unit predicts the high-voltage side reactive power, point of common coupling (PCC) voltage and main characteristic harmonic current in the future period based on the harmonic analysis data in the first state data and other real-time data, combined with a pre-configured discrete state-space prediction model, and solves an optimization problem that simultaneously satisfies multiple objectives such as reactive power tracking, voltage fluctuation suppression, harmonic content reduction and control increment smoothing, and is constrained by SVG equipment capacity, output change rate and grid voltage safety range, and finally generates a coordinated compensation instruction.

[0068] The equipment health management unit analyzes the second status data reported by the SVG to predict the lifespan of key components such as IGBTs and DC capacitors and provide early warning of failures.

[0069] As an example, the Coffin-Manson model based on temperature cycling is used to predict the fatigue life of IGBT solder layers, or a model based on the growth of the equivalent series resistance (ESR) of electrolytic capacitors is used to predict its life. When the predicted remaining life is lower than the threshold or abnormal features (such as a sudden change in the rate of temperature growth) are detected, the platform generates an early warning message and notifies the operation and maintenance personnel through the human-machine interface or SMS, while also providing preventive maintenance suggestions.

[0070] As a preferred embodiment, the data processing unit generates reactive power compensation instructions by the following steps:

[0071] S11: Calculate the target reactive power value on the high-voltage side based on the current real-time reactive power value on the high-voltage side and the preset power factor target value;

[0072] S12: Using the deviation between the real-time reactive power value and the target reactive power value as the reactive power control benchmark, combined with the harmonic content suppression requirements and voltage deviation suppression requirements derived from the harmonic analysis data, and considering the control increment smoothing requirements, a finite time domain multi-objective optimization problem is constructed.

[0073] S13: In each control cycle, the optimization problem is solved in a rolling manner according to the multi-objective optimization algorithm to generate a reactive power compensation command that meets the preset constraints; the preset constraints include system dynamic constraints, control input constraints, control increment constraints and output safety constraints.

[0074] As a preferred embodiment, in the discrete state-space model, the output variables include the reactive power of the high-voltage side bus of the transformer, the voltage of the point of common coupling (PCC) connecting the static var generator, and the amplitude or effective value of the characteristic harmonic current of a specific order (such as the 5th or 7th order) preset according to the harmonic characteristics of the power grid background. The model parameters are updated through online system identification, which preferably adopts the recursive least squares method with a forgetting factor. The input data for online system identification includes the following time-aligned historical sequence:

[0075] The reactive power compensation command sequence sent to the low-voltage side static var generator, the high-voltage side reactive power sequence obtained by the high-voltage side acquisition unit, and the common coupling point voltage and characteristic harmonic current sequence obtained by the low-voltage side static var generator.

[0076] As a preferred option, the core mathematical expression of the discrete state-space model includes the following formula:

[0077] ;

[0078] in, Let be the system state vector. Let k be the state vector of the system at the next time step (k+1). It includes high-voltage side reactive power deviation, point of common coupling voltage deviation, and characteristic harmonic current deviation. State variables represent the deviations of each physical quantity relative to a certain steady-state operating point. Indicates reactive power on the high-voltage side. Indicates the voltage amplitude at critical nodes (such as the PCC point). , Represents the effective value of the main subharmonic current (such as the 5th and 7th harmonics); To control the input vector, , indicating the amount of reactive power adjustment in the SVG output; This represents the deviation of the load disturbance. , The measured value of the load active power at time k. For the k-th control cycle, the power factor angle of the total load on the low-voltage side of the transformer is... Let k be the 5th harmonic current generated by the load at time k. The 7th harmonic current generated by the load at time k; These are the model output variables, which can be the same as or a subset of the state variables; , , , , This is the model matrix, whose parameters are updated through online system identification.

[0079] Furthermore, the specific optimization objective of the multi-objective optimization problem is embodied in a composite objective function J, which is specifically:

[0080] ;

[0081] in, For the future control sequence to be optimized, , To control the time domain, ;

[0082] For reactive power tracking, ,in, The weighting coefficients for the reactive power tracking term. To base on discrete state-space model The predicted value of reactive power on the high-voltage side at any given time. Yes Time based on active power prediction value Calculated future target value;

[0083] This is a voltage deviation penalty term. ,in, The weighting coefficient for the voltage deviation penalty term. This is a predicted value for the point of common coupling voltage connected to the static var generator. This is the voltage reference value;

[0084] This is a harmonic suppression term. Minimizing the predicted value of future major subharmonic currents directly corresponds to reducing the total harmonic distortion (THD). These are the weighting coefficients for each harmonic in the harmonic suppression term. This is a predicted estimate of the effective value of the mth harmonic current on the high-voltage side at time k+i in the future.

[0085] To control the incremental smoothing term, This is used to constrain the rate of change of reactive power output by the SVG, preventing impact on the power grid and equipment; This is the actual execution value from the previous cycle. The amount of reactive power adjustment applied to the SVG output at time k+i in the future. Let be the reactive power regulation amount of the SVG applied at time k+i preceding time, where To control the weighting coefficients of the incremental smoothing term.

[0086] As a preferred option, the preset constraints include system dynamic constraints, control input constraints, control increment constraints, and output safety constraints; among which:

[0087] The system's dynamic constraints are manifested in the following way: during the rolling solution process, the predicted sequence of future states must follow the recursive relationship defined by the discrete state-space model.

[0088] Control input constraints are ,in , The baseline value for SVG output at the current working point. , These represent the maximum inductive reactive power and the maximum capacitive reactive power that the SVG device can continuously output, respectively. The amount of reactive power to be adjusted for the SVG output in the i-th step in the future;

[0089] Control incremental constraints are ,in , The minimum allowable change in SVG reactive power regulation between two consecutive control cycles. The maximum allowable change in SVG reactive power regulation between two consecutive control cycles;

[0090] Output safety constraints are , ,in , As the lower limit of voltage safety, To ensure voltage safety, an upper limit is set. The predicted value of the effective value of the voltage at point PCC at time k+i in the future. Upper limit of total harmonic distortion, The predicted effective value of the m-th harmonic current at time k+i in the future. The predicted value of the fundamental current at time k+i in the future.

[0091] As a preferred embodiment, the data processing unit includes the following steps in performing the rolling solution process:

[0092] S131: The objective function and constraints of the multi-objective optimization problem are mathematically reconstructed in the form of quadratic programming (QP) to generate the corresponding coefficient matrix and vector, as shown in the following equation:

[0093] ;

[0094] Among them, matrix , , , , , It is systematically constructed from the coefficients in the objective function and constraints. The cloud-based decision platform calls a high-efficiency QP solver for online solution. H and f are completely determined by the weight coefficients in the objective function and the dynamic prediction model of the system; G and h are constructed from all physical limits, safety boundaries and corresponding dynamic relationships. and This is determined by specific equations. Through rigorous mathematical derivation, this construction process transforms the original, physically-based optimization and control problem into a standard convex quadratic programming problem that can be efficiently computed by a standard optimization solver, thus achieving online real-time solution.

[0095] S132: Invoke a preset quadratic programming solver (such as an interior-point method solver or an effective set method solver) to numerically solve the coefficient matrix and vectors, and obtain the optimal control increment sequence in the future finite time domain. ;

[0096] S133: Extract the control increment corresponding to the current moment from the optimal control increment sequence, and combine it with the control output of the previous moment to calculate the current reactive power compensation command, as shown in the following formula:

[0097] ;

[0098] As a preferred embodiment, the rolling solution process of the data processing unit, combined with a closed-loop adaptive mechanism, constitutes a complete control cycle, including the following steps:

[0099] S134: In the current control cycle k, send the reactive power compensation command and its corresponding execution timestamp to the static var generator and receive the reception confirmation returned by it.

[0100] S135: At the preset sampling time corresponding to the command execution (usually a fixed time point after the command is issued, after delay compensation), the data receiving and synchronization unit collects the compensated electrical quantity measurement data (mainly including high-voltage side reactive power and PCC point voltage) from the high-voltage side acquisition unit, as well as the actual output current data from the static var generator.

[0101] S136: Calculate the error between the compensated electrical quantity measurement data and the corresponding predicted value of the previous cycle, and input the error into the state estimator (e.g., Kalman filter) to update the estimate of the internal state of the system and obtain a more accurate initial state value for the next cycle prediction.

[0102] S137: If the prediction error continues to exceed the preset threshold, it indicates that the model and the actual system are dynamically mismatched. Then, the online system identification process of the discrete state space model is triggered, and the model matrix parameters are updated using recent historical data. The corrected system state is used as the new initial state, and the prediction time domain is rolled forward by one control cycle to start the optimization calculation of the next cycle k+1.

[0103] It should be noted that the SVG compensation system proposed in this disclosure is particularly suitable for complex scenarios with high requirements for power quality and operational economy, such as smart distribution networks, industrial microgrids, and rail transit power supply systems containing distributed renewable energy sources and impulsive loads.

[0104] As an example, in smart distribution networks with a high proportion of distributed photovoltaic (PV) grids, the intermittency and randomness of PV output can cause rapid fluctuations in reactive power and voltage at the point of common coupling (PCC). This system, through direct closed-loop control of reactive power on the high-voltage side and rapid rolling optimization using a multi-objective optimization algorithm on a cloud platform, can track and smooth these fluctuations in real time. While maintaining the power factor, it effectively suppresses voltage exceedances and avoids erroneous adjustments caused by sudden changes in active power, as is common in traditional solutions.

[0105] In the above scheme, the wireless communication unit supports adaptive switching between 4G / 5G cellular networks and LoRa wide area networks. Specifically, the unit continuously monitors the signal quality of each link and dynamically decides on switching based on the real-time priority of data services (with reactive power compensation commands issued by the cloud having the highest priority).

[0106] As a preferred embodiment, this disclosure also includes a high-precision time synchronization system to provide a unified time reference for data acquisition by the high-voltage side acquisition unit, data processing by the cloud platform, and command execution by the static var generator. Specifically, the system uses a Network Time Protocol (NTP) server deployed on the cloud platform or a Global Navigation Satellite System (GNSS) module integrated into the high-voltage side acquisition unit as the master clock source to distribute synchronization signals to all network nodes. Both data acquisition and command execution are based on this unified time standard, and strict timing consistency between high-voltage side sampling, cloud-based decision-making, and low-voltage side compensation actions is ensured through timestamp alignment and network transmission delay compensation techniques.

[0107] Understandably, in this disclosure, the cloud platform, serving as the system's decision-making hub, is deployed on a cloud server. Based on real-time synchronized data uploaded by the high-voltage side acquisition unit and the low-voltage side static var generator, it executes a multi-objective optimization algorithm centered on model predictive control (MPC). Its core function is to use the high-voltage side reactive power deviation as a direct control variable, and, under the premise of satisfying equipment and grid safety constraints, to continuously solve for the optimal compensation command that coordinates and optimizes reactive power, voltage fluctuation, and harmonic content, and then send it out for execution through a highly reliable link.

[0108] Please see Figure 3 Another specific embodiment of this disclosure proposes an SVG compensation method based on reactive power closed-loop and cloud collaboration, which is executed based on the SVG compensation system in the above embodiment. The method includes the following steps:

[0109] S1: By connecting to the acquisition unit on the high-voltage side of the transformer, the voltage and current signals on the high-voltage side are acquired synchronously, the real-time reactive power value and harmonic analysis data on the high-voltage side are calculated, and packaged into first-state data.

[0110] S2: Upload the first state data to the cloud platform via wireless communication;

[0111] S3: In the cloud platform, the target reactive power value is determined based on the real-time reactive power value on the high-voltage side and the preset power factor target value. The deviation between the two is used as the core control quantity. Combined with the harmonic content reflected by the harmonic analysis data in the first state data and the voltage suppression requirements, a reactive power compensation command that meets the constraints is generated through a multi-objective optimization algorithm based on the discrete state space model.

[0112] S4: Send the reactive power compensation command to the static var generator connected to the low-voltage side of the transformer;

[0113] S5: Controls the static var generator to output the corresponding compensation current according to the command;

[0114] S6: The power grid electrical quantities on the high-voltage side after compensation are collected again by the acquisition unit, new first state data is generated and fed back to the cloud platform;

[0115] S7: In the cloud platform, the deviation between the measured data fed back from S6 and the predicted values ​​in S3 is used to perform online correction and update of the parameters or optimization strategies of the discrete state-space model; the predicted values ​​include the predicted value sequence of high-voltage side reactive power, common coupling point voltage and characteristic harmonic current.

[0116] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0117] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0118] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An SVG compensation system based on reactive power closed-loop and cloud-based collaboration, characterized in that, The topology architecture applied to high-voltage side sampling and low-voltage side compensation of transformers includes a data acquisition unit, a static var generator, and a cloud platform that wirelessly interacts with both. The acquisition unit is connected to the high-voltage side to acquire and process the electrical quantities of the power grid on the high-voltage side, so as to obtain the first state data containing the real-time reactive power value and harmonic analysis data of the high-voltage side. The cloud platform is used to receive real-time updated first status data, and based on the real-time reactive power value of the high-voltage side and the preset power factor target value in the first status data, to determine the target reactive power value of the high-voltage side; using the deviation between the real-time reactive power value and the target reactive power value as the core control quantity, and generating reactive power compensation instructions through a preset multi-objective optimization algorithm; The static var generator is connected to the low-voltage side to receive and respond to the reactive power compensation command, determine and output the corresponding compensation current to the low-voltage side of the transformer to compensate the reactive power on the high-voltage side. The acquisition unit is also used to acquire the electrical quantities of the power grid on the high-voltage side after the static var generator performs compensation, and send the updated first state data to the cloud platform. The cloud platform includes: The data receiving and synchronization unit is used to receive and align first state data from the acquisition unit and second state data from the low-voltage side static var generator based on a unified time reference. The second state data includes the IGBT module temperature, DC bus voltage, cooling fan speed and actual output compensation current data of the static var generator. The data processing unit is used to generate reactive power compensation instructions through a preset multi-objective optimization algorithm, and to maintain and optimize the multi-objective optimization algorithm online based on the measured data fed back after the instructions are executed; wherein, the multi-objective optimization algorithm includes: in each control cycle, generating the reactive power compensation instructions on a rolling basis based on a preset discrete state space model and applying preset constraints. The health management unit is used to predict lifespan and provide early warning of failures based on the second status data.

2. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 1, characterized in that, The data processing unit generates the reactive power compensation instruction by the following steps: Calculate the target reactive power value on the high-voltage side based on the current real-time reactive power value on the high-voltage side and the preset power factor target value; Using the deviation between the real-time reactive power value and the target reactive power value as the reactive power control benchmark, and combining the harmonic content suppression requirements and voltage deviation suppression requirements based on the harmonic analysis data in the first state data, a multi-objective optimization problem in the finite time domain is constructed. In each control cycle, the multi-objective optimization problem is solved in a rolling manner according to the multi-objective optimization algorithm to generate reactive power compensation commands that meet the preset constraints.

3. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 2, characterized in that, In the discrete state-space model, the output variables include the reactive power of the high-voltage side bus of the transformer, the voltage at the common coupling point connecting the static var generator, and the amplitude or effective value of the characteristic harmonic current of a specific order. The model parameters are updated through online system identification, and the input data for online system identification includes the following time-aligned historical sequences: The reactive power compensation command sequence issued to the low-voltage side static var generator, the high-voltage side reactive power sequence obtained by the high-voltage side acquisition unit, and the common coupling point voltage and characteristic harmonic current sequence obtained by the low-voltage side static var generator.

4. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 2, characterized in that, The discrete state-space model includes the following equation: ； Where k is the current time, The system state vector includes the high-voltage side reactive power deviation, the common coupling point voltage deviation, and the characteristic harmonic current deviation. The control input vector represents the adjustment amount of the reactive power output of the static var generator; This represents the deviation of the load disturbance. , The measured value of the load active power at time k. For the k-th control cycle, the power factor angle of the total load on the low-voltage side of the transformer is... Let k be the 5th harmonic current generated by the load at time k. The 7th harmonic current generated by the load at time k; These are model output variables; This is the model matrix, whose parameters are updated through online system identification.

5. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 4, characterized in that, The optimization objective of the multi-objective optimization problem includes a composite objective function J, which is specifically: ； in, For the future control sequence to be optimized, , To control the time domain, ; For reactive power tracking, ,in, The weighting coefficients for the reactive power tracking term. To base on discrete state-space model The predicted value of reactive power on the high-voltage side at any given time. Yes Time based on active power prediction value Calculated future target value; This is a voltage deviation penalty term. ,in, The weighting coefficient for the voltage deviation penalty term. This is a predicted value for the point of common coupling voltage connected to the static var generator. This is the voltage reference value; This is a harmonic suppression term. , These are the weighting coefficients for each harmonic in the harmonic suppression term. For the future High-voltage side at the moment m Predicted estimates of the effective value of the subharmonic current; To control the incremental smoothing term, ,make This is the actual execution value from the previous cycle. For the future The amount of reactive power regulation applied to the SVG output at any given time. for The reactive power regulation of the SVG applied at the time preceding time, where, To control the weighting coefficients of the incremental smoothing term.

6. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 5, characterized in that, The preset constraints include system dynamic constraints, control input constraints, control increment constraints, and output safety constraints; wherein: The predicted future state variables in the system's dynamic constraints satisfy the dynamic relationships defined by the discrete state-space model; Control input constraints are ,in , The baseline value for SVG output at the current working point. These represent the maximum inductive reactive power and the maximum capacitive reactive power that the SVG device can continuously output, respectively. For the future The amount of reactive power adjustment of the SVG output; Control incremental constraints are ,in , The minimum allowable change in SVG reactive power regulation between two consecutive control cycles. The maximum allowable change in SVG reactive power regulation between two consecutive control cycles; Output safety constraints are , ,in , As the lower limit of voltage safety, As the upper limit of voltage safety, For the future Predicted value of the effective voltage at point PCC at time 10:

00. Upper limit of total harmonic distortion, For the future Time of the first m Predicted value of the effective value of the subharmonic current For the future The predicted value of the effective value of the fundamental current at time t.

7. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 2, characterized in that, The rolling solution process includes: The objective function and constraints of the multi-objective optimization problem are mathematically reconstructed in the form of quadratic programming to generate the corresponding coefficient matrix and vector. The preset quadratic programming solver is invoked to numerically solve the coefficient matrix and vector to obtain the optimal control increment sequence in the future finite time domain; The control increment corresponding to the current moment is extracted from the optimal control increment sequence, and combined with the control output of the previous moment, the current reactive power compensation command is calculated.

8. The SVG compensation system based on reactive power closed-loop and cloud collaboration according to claim 7, characterized in that, The rolling solution process also includes: In the current control cycle k, the reactive power compensation instruction and its corresponding execution timestamp are sent to the static var generator, and the received confirmation is received from it. At a preset sampling time corresponding to the execution of the instruction, the data receiving and synchronization unit collects the compensated electrical quantity measurement data from the high-voltage side acquisition unit, as well as the actual output current data from the static var generator. The error between the compensated electrical quantity measurement data and the corresponding predicted value of the previous cycle is calculated, and the error is input into the state estimator to update the estimate of the internal state of the system. If the error exceeds a preset threshold, the online system identification process of the discrete state space model is triggered. Using the corrected system state as the new initial state, the prediction time domain is rolled forward by one control cycle, and the optimization calculation for the next cycle k+1 is started.

9. An SVG compensation method based on reactive power closed-loop and cloud-based collaboration, executed based on the SVG compensation system according to any one of claims 1-8, characterized in that, The method includes the following steps: S1: By acquiring the electrical quantities of the high-voltage power grid through the acquisition unit connected to the high-voltage side of the transformer, the real-time reactive power value and harmonic analysis data of the high-voltage side are calculated and packaged into first-state data; S2: Upload the first status data to the cloud platform via wireless communication; S3: In the cloud platform, a target reactive power value is determined based on the real-time reactive power value of the high-voltage side and the preset power factor target value in the first state data; the deviation between the real-time reactive power value and the target reactive power value is used as the core control quantity, and a reactive power compensation command is generated through a preset multi-objective optimization algorithm. S4: Send the reactive power compensation command to the static var generator connected to the low-voltage side of the transformer; S5: Control the static var generator to output the corresponding compensation current according to the reactive power compensation command; S6: The power grid electrical quantities on the high-voltage side after compensation are collected again by the acquisition unit, new first state data is generated and fed back to the cloud platform; S7: In the cloud platform, the parameters or optimization strategy of the discrete state space model are corrected and updated online by using the deviation between the first state data fed back in S6 and the predicted value in S3.

Images