A method and system for controlling voltage at a distribution network connection point

Abstract

The present invention belongs to the field of distributed photovoltaic technology, and specifically relates to a distribution network grid connection point voltage control method and system, which mainly includes the following steps: real-time monitoring of voltage, monitoring of grid load, power output, reactive power and data collection; pre-processing of the collected voltage data, current data, load data, power output data, reactive power data, and by analyzing the pre-processed data, using the LiBSVM model to predict the grid operation status in the future, and comprehensively considering voltage stability and economic factors, by optimizing control parameters, using an improved genetic algorithm to maximize the comprehensive benefits of voltage control. The present invention introduces an adaptive control strategy and a prediction algorithm, which can respond to changes in grid parameters in real time, and formulate an optimal control strategy based on the prediction results, improve the accuracy of voltage control, and maximize the comprehensive benefits of voltage control.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed photovoltaics, and in particular relates to a method and system for controlling voltage at a distribution network connection point. Background Art

[0002] With the grid-connected operation of large-scale photovoltaic power generation systems, the proportion of photovoltaic power generation capacity in the total system capacity has increased, and the impact on the power system has become increasingly greater. Photovoltaic power generation systems usually have to go through the power electronic interface-inverter, and connect to the grid through the low-voltage or medium-voltage distribution network. The traditional power system transmission and distribution network is designed as a one-way transmission and distribution system from the power generation unit to the load. The grid-connected operation of large-scale photovoltaic power generation systems may cause the problem of reverse flow, resulting in voltage increase or overvoltage at the grid connection point of the photovoltaic power generation system. The voltage increase not only affects the power supply quality of the local load, but also increases the loss of power transmission and distribution equipment such as lines and transformers, causing system overload, and limits the connection point to more photovoltaic power generation systems, affecting the penetration rate of photovoltaic power generation systems. Therefore, it is necessary to control the voltage at the grid connection point. However, relying solely on the voltage adjustment method of the traditional power system cannot completely and effectively and economically solve the problem of voltage increase at the grid connection point.

[0003] In terms of voltage control accuracy, the reactive current and voltage control strategy has good steady-state accuracy; from the perspective of economy, the reactive current and voltage adjustment strategy has better economic benefits than the active current and voltage adjustment strategy. Based on the above problems, the present invention proposes a distribution network connection point voltage control method and system. Summary of the invention

[0004] The present invention provides a distribution network connection point voltage control method and system, which are used to solve the technical problems of low data transmission efficiency, poor system stability and high programming complexity of the existing single-chip microcomputer system.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows: A method for controlling voltage at a distribution network connection point comprises the following steps:

[0006] (1) Real-time monitoring and data collection: Real-time monitoring and acquisition of real-time operating status data of the grid connection point, including voltage data, current data, load data, power output data, and reactive power data;

[0007] (2) Data analysis and state identification: The collected voltage data, current data, load data, power output data, and reactive power data are cleaned and preprocessed. The preprocessing includes filtering and denoising. The filtering adopts particle PF filtering with improved weights to improve data quality.

[0008] The particle PF filter with improved weights is used for the data x t, t=1,2,3...,N particle filter denoising process is as follows:

[0009] 1) Initialization

[0010] Given a state distribution p(x) and an initial value x0, randomly sample

[0011] The symbol xp indicates that x is an observed value of the distribution p(x), and the weight is initialized. Where t=1,2,3...,N;

[0012] 2) Loop

[0013] Set each time step t = 1, 2, 3..., N and do the following:

[0014] A. Importance distribution is defined as: It is assumed that the prior distribution is known sampling

[0015] B. Calculate importance weights: Assume the likelihood function It is known that the standard weight is calculated, that is, the importance weight is normalized, expressed as

[0016] C. Resampling, a set of N discrete random variables In time step t=1,2,3...,N, the values ​​are taken according to the following probabilities:

[0017] gather and

[0018] D. By iteratively updating the weights and positions of the particles, an accurate estimate of the running status data can be obtained.

[0019] (3) Prediction analysis: Based on the preprocessed data, the LiBSVM model is used to predict the grid operation status in the future;

[0020] (4) Multi-objective optimization analysis: taking into account voltage stability and economic factors, optimizing the real-time operating status data, and using an improved genetic algorithm to maximize the comprehensive benefits of voltage control;

[0021] The process of improving the genetic algorithm is as follows:

[0022] 1) Define control parameters: voltage set point, reactive power compensation equipment input, transformer tap position, capacitor / reactor switching status;

[0023] 2) Use of fitness function: The fitness function is defined using the weighted sum method, assuming four objective functions:

[0024] f1(x): represents voltage stability, i.e. the sum of the absolute values ​​of voltage deviations;

[0025] f2(x): represents economy, i.e. the operating cost of reactive power compensation equipment;

[0026] f3(x): represents the equipment life, which is an assessment based on the number of times the equipment is used;

[0027] f4(x): represents the maintenance cost, i.e. the estimated maintenance expense;

[0028] Each objective function is assigned a weight ω1, ω2, ω3, ω4

[0029] Satisfies ω1+ω2+ω3+ω4=1,

[0030] Then define the fitness function as:

[0031] f fitness (x) = ω1·f1(x)+ω2·f2(x)+ω3·f3(x)+ω4·f4(x), the specific function definition is given in combination with the actual model;

[0032] 3) Variable constraints:

[0033] Voltage set point: Within the allowable voltage range, V min ≤V set ≤V max ;

[0034] Reactive power compensation equipment investment: According to the equipment capacity limit, Q min ≤Q comp ≤Q max ;

[0035] Transformer tap position: According to the number of taps of the transformer, Tap min ≤tap≤Tap max ;

[0036] Capacitor / reactor switching status: binary variable, 0 means not switched on, 1 means switched on;

[0037] The specific process of improving genetic algorithm is as follows:

[0038] 1) Initialization: Randomly generate an initial population, i.e., a set of candidate solutions for the voltage control strategy, ensuring that each solution satisfies the variable constraints;

[0039] X i,j ~U(lb j,ub j )

[0040] Where X is the population, i∈{1,2,3...,N} and j∈{1,2,3...,sizepop}, N is the number of populations, sizepop is the population dimension, lb and ub are the lower and upper bounds of the exploration interval respectively; U satisfies random uniform distribution;

[0041] 2) Evaluation: Use the fitness function f fitness (x) Calculate the fitness value MM of each solution in the population i ;

[0042] 3) Selection: select individuals based on fitness values ​​for genetic operations, roulette wheel selection, and elite selection;

[0043] 4) Crossover: Randomly select two individuals for crossover operation to generate new offspring individuals; the crossover operation needs to ensure that the newly generated individuals still meet the variable constraints;

[0044] 5) Mutation: Perform mutation operations on individuals with a certain probability to increase the diversity of the population. The mutation operation also needs to meet variable constraints;

[0045] 6) Update the population: replace some individuals in the population with the newly generated offspring individuals, or merge them with individuals in the population and then select them;

[0046] 7) Termination condition: Determine whether the termination condition is met, that is, whether the maximum number of iterations has been reached and the quality of the solution is no longer improved. If so, stop the algorithm and output the optimal solution; otherwise, return to step 2) to continue iterating.

[0047] Furthermore, for the missing data of the real-time running status data, a linear interpolation method is used to repair the missing data with a small time interval:

[0048]

[0049] Where: x t and x t- j are the data parameters sampled at time t and time tj; x t-n is the missing data parameter at tn. i is the weight value at time i,

[0050] Furthermore, the mean smoothing method is used to process the distorted data. or In the formula, and are the thresholds of the errors of adjacent sampling data; x t is the current I t, voltage U t , Temperature T t , load data M t , power output data P t And reactive power data Q t The normalized average value at time t.

[0051] Furthermore, the specific process of using the LiBSVM model for prediction is as follows:

[0052] 1) Build the model:

[0053] Model 1: Establishing the LiBSVM prediction model between voltage monitoring data, grid status monitoring data and “grid load”;

[0054] Model 2: Establishing the LiBSVM prediction model between voltage monitoring data, grid status monitoring data and “power output”;

[0055] 2) Explanatory variables: Set X1-X6 to be the current I t , voltage U t , temperature T t , load data M t , power output data P t , reactive power data Q t ;

[0056] Model 1 Explained variables: Y1 is the grid load;

[0057] Model 2 explained variables: Y2 is the power output;

[0058] 3) Model prediction: Normalize the above variables, that is, map the feature values ​​from a large range to [-1,1];

[0059] f(x): x→y, y∈[-1, 1]

[0060]

[0061] The mean absolute percentage error MAPE and mean absolute error MAE are used for evaluation. The expressions of MAPE and MAE are as follows:

[0062]

[0063] A system comprising a method for controlling the voltage of a distribution network connection point comprises a data acquisition module, a data processing module, a data calculation module, an execution module and a control system, wherein the data acquisition module monitors the connection point voltage in real time, and simultaneously monitors the key parameters of the grid load, power output and reactive power, so as to obtain the real-time operation status data of the grid; the data processing module pre-processes the collected voltage data, current data, load data, power output data and reactive power data to improve the data quality, wherein the pre-processing comprises two processes, filtering and denoising, wherein the filtering adopts a particle PF filter with improved weights, and the data is cleaned before filtering; the data calculation module can use a LiBSVM model to predict the operation status of the grid in the future for the pre-processed data; the execution module optimizes the control parameters and uses an improved genetic algorithm to maximize the comprehensive benefits of voltage control.

[0064] Furthermore, the data acquisition module, data processing module, data calculation module and execution module are controlled by the control system.

[0065] Furthermore, the execution module can also adjust reactive power compensation equipment and transformers in the distribution network to achieve precise control of the voltage at the grid connection point.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] (1) The present invention can respond to changes in power grid parameters in real time by introducing adaptive control strategies and prediction algorithms, and formulate optimal control strategies based on prediction results to improve the accuracy of voltage control.

[0068] (2) The application of the multi-objective optimization algorithm in the present invention can ensure voltage stability while taking into account other factors such as economy, thereby maximizing the comprehensive benefits of voltage control. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of the voltage control method of the distribution network connection point of the present invention;

[0070] Figure 2 Improved genetic algorithm flow chart for the present invention. DETAILED DESCRIPTION

[0071] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0072] like Figure 1As shown, a method for controlling voltage at a distribution network connection point comprises the following steps:

[0073] (1) Real-time monitoring and data collection: Real-time monitoring of the grid connection point voltage, as well as key parameters of grid load, power output, and reactive power, to obtain real-time grid operation status data;

[0074] Voltage monitoring: Real-time monitoring of the grid connection point voltage, and obtaining the voltage value through a high-precision voltage sensor.

[0075] Grid status monitoring: monitor key parameters such as grid load, power output, reactive power, etc. to obtain the real-time operating status of the grid.

[0076] (2) Data analysis and state identification: The collected voltage data, current data, load data, power output data, and reactive power data are preprocessed to improve data quality. The preprocessing includes two processes: filtering and denoising. The filtering uses particle filtration (PF) filtering with improved weights. Data cleaning is performed before filtering.

[0077] Data preprocessing: preprocess the collected data, including filtering, denoising, etc., to improve data quality. The data includes: voltage data U, current data I, load data M, power output data P, reactive power data Q.

[0078] The filtering adopts the granular PF filtering with improved weights, and the data is cleaned before filtering.

[0079] The data is cleaned and repaired. For missing data, linear interpolation can be used to repair missing data with a small time interval:

[0080]

[0081] Where: x t and x t- j are the data parameters sampled at time t and time tj; x t-n is the missing data parameter at tn. i is the weight value at time i,

[0082] For distorted data, since the data is continuously sampled and transmitted back in sequence, the data sampled in adjacent periods will not change suddenly. If the range of sampled data at a certain moment is greater than or less than 10% of the monitored value before and after, the sampled data is considered distorted. The mean smoothing method is used:

[0083] or

[0084] In the formula, and are the thresholds of the errors of adjacent sampling data. t is the current I t , voltage U t , Temperature T t , load data M t , power output data P t And reactive power data Q t The normalized average value at time t.

[0085] Filtering: Use particle PF filtering with improved weights for filtering and denoising.

[0086] Particle filtering is a state estimation technology based on the Monte Carlo method, which is often used for state estimation and data fusion of nonlinear and non-Gaussian systems. In data denoising, particle filtering can effectively eliminate noise interference and extract the real state information of the system by randomly sampling and weighting the system state. By continuously iteratively updating the weights and positions of particles, accurate estimation and filtering of the operating state data can be obtained, thus improving the reliability and accuracy of the data.

[0087] The weighted particle (PF) filter is applied to the data x t , t=1,2,3...,N particle filter denoising process is as follows:

[0088] 1) Initialization

[0089] Given a state distribution p(x) and an initial value x0, randomly sample

[0090] The symbol xp indicates that x is an observed value of the distribution p(x), and the weight is initialized. Where t=1,2,3...,N;

[0091] 2) Loop

[0092] Set each time step t = 1, 2, 3..., N and do the following:

[0093] A. Importance distribution is defined as: It is assumed that the prior distribution is known sampling

[0094] B. Calculate importance weights: Assume the likelihood function It is known that the standard weight is calculated, that is, the importance weight is normalized, expressed as

[0095] C. Resampling, a set of N discrete random variables In time step t=1,2,3...,N, the values ​​are taken according to the following probabilities:

[0096] gather and

[0097] D. By iteratively updating the weights and positions of the particles, an accurate estimate of the running status data can be obtained.

[0098] (3) Predictive analysis: By analyzing the preprocessed data, the LiBSVM model is used to predict the grid operation status in the future; based on these prediction results, the system can formulate voltage control strategies in advance to achieve preventive control. This predictive control can adjust control parameters in advance before the grid load or power output changes significantly, thereby avoiding risks such as voltage over-limit or equipment damage.

[0099] The specific process of using the LiBSVM model for prediction is as follows:

[0100] 1) Build the model:

[0101] Model 1: Establishing the LiBSVM prediction model between voltage monitoring data, grid status monitoring data and “grid load”;

[0102] Model 2: Establishing the LiBSVM prediction model between voltage monitoring data, grid status monitoring data and “power output”;

[0103] 2) Explanatory variables: Set X1-X6 to be the current I t , voltage U t , temperature T t , load data M t , power output data P t , reactive power data Q t ;

[0104] Model 1 Explained variables: Y1 is the grid load;

[0105] Model 2 explained variables: Y2 is the power output;

[0106] 3) Model prediction: Normalize the above variables, that is, map the feature values ​​from a large range to [-1,1].

[0107] f(x): x→y, y∈[-1, 1]

[0108]

[0109] The mean absolute percentage error MAPE and mean absolute error MAE are used for evaluation. The expressions of MAPE and MAE are as follows:

[0110]

[0111] (4) Multi-objective optimization analysis: Taking voltage stability and economic factors into consideration, the comprehensive benefits of voltage control are maximized by optimizing control parameters and using an improved genetic algorithm.

[0112] The process of improving the genetic algorithm is as follows:

[0113] (1) Define control parameters: voltage set point, reactive power compensation equipment input, transformer tap position, capacitor / reactor switching status;

[0114] (2) Use of fitness function: The fitness function is defined using the weighted sum method, assuming four objective functions:

[0115] f1(x): represents voltage stability, i.e. the sum of the absolute values ​​of voltage deviations;

[0116] f2(x): represents economy, i.e. the operating cost of reactive power compensation equipment;

[0117] f3(x): represents the equipment life, which is an assessment based on the number of times the equipment is used;

[0118] f4(x): represents the maintenance cost, i.e. the estimated maintenance expense;

[0119] Each objective function is assigned a weight ω1, ω2, ω3, ω4

[0120] Satisfies ω1+ω2+ω3+ω4=1,

[0121] Then define the fitness function as:

[0122] f fitness (x)=ω1·f1(x)+ω2·f2(x)+ω3·f3(x)+ω4·f4(x), the specific function definition is given in combination with the actual model.

[0123] (3) Variable constraints:

[0124] Voltage set point: Within the allowable voltage range, V min ≤V set ≤V max

[0125] Reactive power compensation equipment investment: According to the equipment capacity limit, Q min ≤Q comp ≤Q max

[0126] Transformer tap position: According to the number of taps of the transformer, Tap min ≤tap≤Tap max

[0127] Capacitor / reactor switching status: binary variable, 0 means off, 1 means on.

[0128] like Figure 2 As shown, the specific calculation process is as follows:

[0129] 1) Initialization: Randomly generate an initial population (i.e., a set of candidate solutions for the voltage control strategy) to ensure that each solution satisfies the variable constraints;

[0130] X i,j ~U(lb j ,ub j )

[0131] Where X is the population, i∈{1,2,3...,N} and j∈{1,2,3...,sizepop}, N is the number of populations, sizepop is the population dimension, lb and ub are the lower and upper bounds of the exploration interval respectively; U satisfies random uniform distribution;

[0132] 2) Evaluation: Use the fitness function f fitness (x) Calculate the fitness value MM of each solution in the population i ;

[0133] 3) Selection: select individuals based on fitness values ​​for genetic operations, roulette wheel selection, and elite selection;

[0134] 4) Crossover: Randomly select two individuals for crossover operation to generate new offspring individuals; the crossover operation needs to ensure that the newly generated individuals still meet the variable constraints;

[0135] 5) Mutation: Perform mutation operations on individuals with a certain probability to increase the diversity of the population. The mutation operation also needs to meet variable constraints;

[0136] 6) Update the population: replace some individuals in the population with newly generated offspring individuals, or merge them with individuals in the population and then make a selection.

[0137] 7) Termination condition: Determine whether the termination condition is met (the maximum number of iterations is reached and the quality of the solution no longer improves). If so, stop the algorithm and output the optimal solution; otherwise, return to step (2) to continue iterating.

[0138] In practical applications, it may be necessary to adjust the fitness function and variable constraints according to the specific characteristics and needs of the power system. The selection of weights ω1, ω2, ω3, and ω4 will affect the optimization results. Appropriate weight values ​​can be determined through experiments or expert experience.

[0139] The performance of genetic algorithms is affected by many parameters, such as population size, crossover probability, mutation probability, etc. These parameters need to be adjusted according to actual problems.

[0140] A system including a method for controlling the voltage of a distribution network connection point includes a data acquisition module, a data processing module, a data calculation module, an execution module and a control system, wherein the data acquisition module monitors the connection point voltage in real time, and simultaneously monitors the key parameters of the grid load, power output and reactive power, so as to obtain the real-time operation status data of the grid; the data processing module pre-processes the collected voltage data, current data, load data, power output data and reactive power data to improve the data quality, and the pre-processing includes two processes, filtering and denoising, and the filtering adopts the particle PF filtering with improved weight, and the data is cleaned before filtering; the data calculation module can use the LiBSVM model to predict the operation status of the grid in the future period of time for the pre-processed data; the execution module optimizes the control parameters and uses the improved genetic algorithm to maximize the comprehensive benefits of voltage control. The data acquisition module, the data processing module, the data calculation module and the execution module are controlled by the control system. The execution module can also adjust the reactive compensation equipment and transformer in the distribution network to realize the precise control of the voltage of the connection point.

[0141] Through the control system and control method of the present invention, it is possible to achieve accurate control of the voltage at the distribution network connection point, improve voltage stability, reduce operating costs, and ensure safe and efficient operation of the power grid. At the same time, the system also has strong adaptive and predictive capabilities, can adapt to changes in power grid conditions, achieve preventive control, and further reduce the risk of power grid operation.

[0142] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for controlling voltage at a distribution network connection point, characterized in that: The following steps are involved: (1) Real-time monitoring and data collection: Real-time monitoring and acquisition of real-time operating status data of the grid connection point, including voltage data, current data, load data, power output data and reactive power data; (2) Data analysis and state identification: The collected voltage data, current data, load data, power output data, and reactive power data are cleaned and preprocessed. The preprocessing includes filtering and denoising. The filtering adopts particle PF filtering with improved weights to improve data quality. The particle PF filter with improved weights is used for the data x t , t=1,2,3...,N particle filter denoising process is as follows: 1) Initialization Given a state distribution p(x) and an initial value x0, randomly sample The symbol xp indicates that x is an observed value of the distribution p(x), and the weight is initialized. Where t=1,2,3...,N; 2) Loop Set each time step t = 1, 2, 3..., N and do the following: A. Importance distribution is defined as: It is assumed that the prior distribution is known sampling B. Calculate importance weights: Assume the likelihood function It is known that the standard weight is calculated, that is, the importance weight is normalized, expressed as C. Resampling, a set of N discrete random variables In time step t=1,2,3...,N, the values ​​are taken according to the following probabilities: gather and D. By iteratively updating the weights and positions of the particles, an accurate estimate of the running status data can be obtained. (3) Prediction analysis: Based on the preprocessed data, the LiBSVM model is used to predict the grid operation status in the future; (4) Multi-objective optimization analysis: taking into account voltage stability and economic factors, optimizing the real-time operating status data, and using an improved genetic algorithm to maximize the comprehensive benefits of voltage control; The process of improving the genetic algorithm is as follows: 1) Define control parameters: voltage set point, reactive power compensation equipment input, transformer tap position, capacitor / reactor switching status; 2) Use of fitness function: The fitness function is defined using the weighted sum method, assuming four objective functions: f1(x): represents voltage stability, i.e. the sum of the absolute values ​​of voltage deviations; f2(x): represents economy, i.e. the operating cost of reactive power compensation equipment; f3(x): represents the equipment life, which is an assessment based on the number of times the equipment is used; f4(x): represents the maintenance cost, i.e. the estimated maintenance expense; Each objective function is assigned a weight ω1, ω2, ω3, ω4 Satisfies ω1+ω2+ω3+ω4=1, Then define the fitness function as: f fitness (x) = ω1·f1(x)+ω2·f2(x)+ω3·f3(x)+ω4·f4(x), the specific function definition is given in combination with the actual model; 3) Variable constraints: Voltage setting point: Within the allowable voltage range, V min ≤V set ≤V max ; Reactive power compensation equipment investment: According to the equipment capacity limit, Q min ≤Q comp ≤Q max ; Transformer tap position: According to the number of taps of the transformer, Tap min ≤tap≤Tap max ; Capacitor / reactor switching status: binary variable, 0 means not switched on, 1 means switched on; The specific process of improving genetic algorithm is as follows: 1) Initialization: Randomly generate an initial population, i.e., a set of candidate solutions for the voltage control strategy, ensuring that each solution satisfies the variable constraints; X i,j ~U(lb j ,ub j ) Where X is the population, i∈{1,2,3...,N} and j∈{1,2,3...,sizepop}, N is the number of populations, sizepop is the population dimension, lb and ub are the lower and upper bounds of the exploration interval respectively; U satisfies random uniform distribution; 2) Evaluation: Use the fitness function f fitness (x) Calculate the fitness value MM of each solution in the population i ; 3) Selection: select individuals based on fitness values ​​for genetic operations, roulette wheel selection, and elite selection; 4) Crossover: Randomly select two individuals for crossover operation to generate new offspring individuals; the crossover operation needs to ensure that the newly generated individuals still meet the variable constraints; 5) Mutation: Perform mutation operations on individuals with a certain probability to increase the diversity of the population. The mutation operation also needs to meet variable constraints; 6) Update the population: replace some individuals in the population with the newly generated offspring individuals, or merge them with individuals in the population and then select them; 7) Termination condition: Determine whether the termination condition is met, that is, whether the maximum number of iterations has been reached and the quality of the solution is no longer improved. If so, stop the algorithm and output the optimal solution; otherwise, return to step 2) to continue iterating.

2. A method for controlling voltage at a distribution network connection point according to claim 1, characterized in that: For the missing data of the real-time running status data, linear interpolation is used to repair the missing data with a small time interval: Where: x t and x t-j are the data parameters sampled at time t and time tj respectively; x t-n is the missing data parameter at tn, ω i is the weight value at time i, 3. A method for controlling voltage at a distribution network connection point according to claim 1, characterized in that: The mean smoothing method is used to process the distorted data. or In the formula, and are the thresholds of the errors of adjacent sampling data; x t is the current I t , voltage U t , Temperature T t , load data M t , power output data P t And reactive power data Q t The normalized average value at time t.

4. A method for controlling voltage at a distribution network connection point according to claim 1, characterized in that: The specific process of using the LiBSVM model for prediction is as follows: 1) Build the model: Model 1: Establishing the LiBSVM prediction model between voltage monitoring data, grid status monitoring data and "grid load"; Model 2: Establishing the LiBSVM prediction model between voltage monitoring data, grid status monitoring data and "power output"; 2) Explanatory variables: Set X1-X6 to be the current I t , voltage U t , temperature T t , load data M t , power output data P t , reactive power data Q t ; Model 1 Explained variables: Y1 is the grid load; Model 2 explained variables: Y2 is the power output; 3) Model prediction: Normalize the above variables, that is, map the feature values ​​from a large range to [-1,1]; f(x): x→y, y∈[-1, 1] The mean absolute percentage error MAPE and mean absolute error MAE are used for evaluation. The expressions of MAPE and MAE are as follows:

5. A system for implementing a method for controlling voltage at a distribution network connection point according to any one of claims 1 to 4, characterized in that: It includes a data acquisition module, a data processing module, a data calculation module, an execution module and a control system, wherein the data acquisition module monitors the voltage of the grid connection point in real time, and simultaneously monitors the key parameters of the grid load, power output and reactive power to obtain the real-time operation status data of the grid; the data processing module pre-processes the collected voltage data, current data, load data, power output data and reactive power data to improve the data quality, and the pre-processing includes two processes: filtering and denoising. The filtering adopts the particle PF filtering with improved weights, and the data is cleaned before filtering; the data calculation module can use the LiBSVM model to predict the operation status of the grid in the future for the pre-processed data; the execution module optimizes the control parameters and uses the improved genetic algorithm to maximize the comprehensive benefits of voltage control.

6. The system of the method for controlling voltage at a distribution network connection point according to claim 5, characterized in that: The data acquisition module, data processing module, data calculation module and execution module are controlled by the control system.

7. The system of the method for controlling voltage at a distribution network connection point according to claim 6, characterized in that: The execution module can also adjust the reactive power compensation equipment and transformers in the distribution network to achieve precise control of the voltage at the grid connection point.

Images