Power grid impedance prediction and equalization adjustment method based on deep learning

Abstract

The invention belongs to the technical field of communication, and discloses a power grid impedance prediction and balance adjustment method based on deep learning. Comprising the following steps: dividing a power grid region according to a power grid topological structure chart, and identifying access points in the power grid region; collecting power grid impedance data and power grid operation state data of each access point within a preset time; preprocessing the power grid impedance data and the power grid operation state data of each access point to obtain a comprehensive feature data set; training according to the comprehensive feature data set to obtain a power grid impedance prediction model, and predicting based on the power grid impedance prediction model to obtain a power grid impedance change trend and a power grid impedance value in future n periods of time; according to the power grid impedance change trend and the power grid impedance value in the future n periods of time, the predicted power grid impedance change rate in the future n periods of time is obtained; and measures can be taken in advance to avoid unnecessary interference, and the stability and the operation efficiency of a power grid system are ensured.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and more specifically, to a power grid impedance prediction and balancing adjustment method based on deep learning. Background Art

[0002] Patent announcement number CN113625054A discloses a grid impedance detection device, an inverter and a grid impedance detection method, which are used to detect the grid impedance at the common coupling point where the inverter is connected to the grid. The detection device includes the following modules: a sampling module, which is used to sample the three-phase voltage signal and the three-phase current signal at the common coupling point; an intelligent chip, which is used to receive the three-phase voltage signal and the three-phase current signal sampled by the sampling module, and input them into the internal grid impedance detection neural network model. The grid impedance detection neural network model is pre-established and trained to be able to output the predicted value of the grid impedance according to the input three-phase voltage signal and the three-phase current signal. The grid impedance detection device of the present invention does not need to inject harmonic current into the common coupling point, and can accurately predict the grid impedance in real time through the deep neural network model to achieve adaptive control of the inverter.

[0003] The existing grid impedance prediction and balancing adjustment methods have the following main problems:

[0004] The LMS filter was not used to remove high-frequency noise. The noise in the grid impedance and operating parameter data may affect the quality of the signal, causing the data to fluctuate too much, resulting in inaccurate or unstable subsequent analysis and modeling results, reducing the predictive ability and robustness of the model, and affecting the reliability of grid management and BB-PLC signal transmission; the extraction of low-frequency components and filtering out short-term fluctuations was not considered, which may cause the processed signal to retain many unnecessary fluctuations and noise, making the long-term trend of the data unclear, thus affecting the trend prediction and modeling results; the step factor was not dynamically adjusted through the step factor limit formula, and the learning process of the LMS filter may become unstable, the convergence speed slows down or stagnates near the local optimal solution, resulting in the filter being unable to effectively learn the characteristics of the data;

[0005] If the multi-dimensional data of the grid impedance, such as frequency, amplitude, phase angle and its rate of change, are not considered, it may not be possible to fully evaluate the impact of grid impedance changes on the bit error rate of BB-PLC signal transmission, which may lead to inaccurate evaluation of the interference between the grid and the BB-PLC signal, and thus fail to effectively identify and optimize the interference source, affecting the stability and efficiency of the communication system; if the introduction of the bit error rate adjustment coefficient and the adjustment range limit formula is not considered, changes such as grid load fluctuations may not be reflected in real time in the evaluation of the BB-PLC signal bit error rate. Due to the lack of a reasonable adjustment mechanism, the bit error rate may be too high or too low, resulting in the system being unable to flexibly adapt to changes in the power grid environment, affecting the quality and performance of BB-PLC communication; in the case of power grid load fluctuations or other dynamic changes, it is impossible to evaluate the quality of BB-PLC signals in real time and optimize transmission strategies, which may lead to increased interference and even signal loss or unstable transmission, affecting the communication between the power network and BB-PLC equipment, reducing the reliability and efficiency of the entire system; without an evaluation and adjustment mechanism to suppress the impact of changes in the power grid environment, the robustness of the BB-PLC signal may not be effectively improved; the system may not be able to cope with interference in a complex power grid environment, resulting in a decrease in the stability and performance of BB-PLC communication when the power grid disturbance is large, affecting the management and operation efficiency of power equipment.

[0006] In view of this, the present invention proposes a grid impedance prediction and balancing adjustment method based on deep learning to solve the above problems. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for predicting and balancing power grid impedance based on deep learning, comprising:

[0008] S1. Divide the power grid area according to the power grid topology diagram and identify the access points in the power grid area;

[0009] S2, collecting grid impedance data and grid operation status data of each access point within a preset time;

[0010] S3, preprocessing the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set;

[0011] S4. A grid impedance prediction model is obtained by training the comprehensive feature data set, and a grid impedance change trend and a grid impedance value in the next n periods of time are predicted based on the grid impedance prediction model; a predicted grid impedance change rate in the next n periods of time is obtained based on the grid impedance change trend and the grid impedance value in the next n periods of time;

[0012] S5. According to the predicted grid impedance change rate in the next n periods of time, determine whether the grid impedance change has an interference effect on the BB-PLC signal transmission; if an interference effect is produced, evaluate the interference effect of the grid impedance change on the BB-PLC signal transmission and obtain the interference effect degree data; the grid impedance prediction terminal automatically generates corresponding interference warning information according to the interference effect degree data.

[0013] Furthermore, the method of dividing the power grid area according to the power grid topology diagram and identifying the access points in the power grid area includes:

[0014] A grid topology diagram is obtained, which includes components of the grid, including substations, distribution equipment, transmission lines, access points and load points; the grid is divided into regions using a connected component algorithm in graph theory according to the electrical connection and geographical distribution of the grid, forming v grid regions, and the devices in each grid region are interconnected; nodes in each grid region that are connected to external devices or other regions are identified as access points, and the access points include substations and distribution boxes.

[0015] Furthermore, the grid impedance data includes the impedance value, amplitude impedance, phase angle of impedance and frequency of impedance at the grid access point; the grid operation status data includes voltage data, current data, grid power data, load current and load voltage at the grid access point.

[0016] Furthermore, the method of preprocessing the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set includes:

[0017] Based on the box plot method, outliers in the grid impedance data and grid operation status data of each access point are identified and eliminated, and missing values ​​in the grid impedance data and grid operation status data of each access point are filled by interpolation method;

[0018] Performing timestamp alignment and standard deviation normalization on the grid impedance data and grid operation status data of each access point after missing value filling, to obtain a grid impedance characteristic data set and a grid operation status characteristic data set;

[0019] The LMS filter is used to perform denoising on the grid impedance parameter data and the grid operation parameter data to obtain the denoised grid impedance parameter data and the grid operation parameter data, and the steps include:

[0020] S41, extracting low-frequency components from the grid impedance parameter data and the grid operation parameter data by a moving average method, and constructing an expected signal vector d(m);

[0021] S42, initializing LMS filter parameters, initializing LMS filter parameters includes presetting filter order M, initializing a weight vector w(m) and setting an initial step size factor μ 0 ;

[0022] S43. Use the grid impedance parameter data and the grid operation parameter data as input signals and define the input signal vector: x(m) = [x(m), x(m-1), ..., x(m-M+1)] T ; Where x(m) is the input signal vector at the mth iteration; m is the number of iterations; T is the transposition operator symbol;

[0023] S44. Based on the weight vector w(m) and the input signal vector x(m), calculate the output signal vector of the LMS filter: y(m)=w T (m)·x(m); where y(m) is the output signal vector at the mth iteration; w T (m) is the transpose of the weight vector at the mth iteration;

[0024] By comparing the output signal vector y(m) and the expected signal vector d(m), the error signal vector is calculated: e(m)=d(m)-y(m); wherein e(m) is the error signal vector at the mth iteration, indicating the deviation between the output signal vector y(m) and the expected signal vector d(m);

[0025] S45. Based on the obtained error signal vector e(m) and the input signal vector x(m), the weight vector is updated using the gradient descent method; the updated weight vector is: w(m+1)=w(m)+μ m ·e(m)x(m); where μ m is the step factor at the mth iteration; e(m)x(m) is the gradient term of the weight; the step factor is limited by the step factor limitation formula, and the size of the step factor is dynamically adjusted; the mean square value MSE of the error signal vector is calculated by the mean square error calculation formula;

[0026] S46. Repeat steps S43-S45 until the mean square value MSE of the error signal vector is less than the preset mean square value threshold θ, then stop the iteration, output the output signal vector of the current LMS filter, and obtain a power grid impedance characteristic data set and a power grid operation status characteristic data set; perform weighted fusion on the power grid impedance characteristic data set and the power grid operation status characteristic data set to obtain a comprehensive characteristic data set.

[0027] Furthermore, the method of limiting the step factor by a step factor limiting formula and dynamically adjusting the size of the step factor includes:

[0028] The step factor limiting formula is: Among them, μ′ m is the step size factor after limitation; k is the attenuation factor that controls the rate at which the step size factor decreases with the number of iterations; b is the step size adaptive adjustment factor.

[0029] Furthermore, the method of performing weighted fusion on the grid impedance characteristic data set and the grid operation status characteristic data set to obtain a comprehensive characteristic data set includes:

[0030] The grid impedance characteristic data set and the grid operation status characteristic data set are fused through a weighted formula to obtain a comprehensive characteristic data set; the grid impedance characteristic data set is denoted as E 1 , the power grid operation status characteristic data set is recorded as E 2 ;

[0031] The weighted formula is: Rb = E 1 ·η 1 +E 2 ·η 2 ; Wherein, Rb is a comprehensive feature data set; η 1 is the weight coefficient of the grid impedance characteristic data set; η 2 is the weight coefficient of the power grid operation status characteristic data set.

[0032] Furthermore, the training method of the power grid impedance prediction model includes:

[0033] The data set is divided into training set, validation set and test set to train the model and evaluate the model performance; the sample set is a subset of the data set, each sample set includes the historical comprehensive feature data set and the corresponding grid impedance change trend and grid impedance value in the next n periods of time; the deep learning library TensorFlow is used to build a grid impedance prediction model; the grid impedance prediction model includes input layer, LSTM layer and output layer;

[0034] The input layer of the model is used to input the historical comprehensive feature data set; the output layer of the model is used to output the corresponding grid impedance change trend and grid impedance value in the future n periods of time; the grid impedance prediction model is an LSTM model;

[0035] Define the loss function of the model, and use the L2 regularized mean square error loss function to measure the grid impedance change trend predicted by the grid impedance prediction model in the next n periods of time and the difference between the grid impedance value and the actual value; use the training set to train the grid impedance prediction model, and update the model parameters through the back propagation algorithm to minimize the loss function; use the validation set to evaluate the performance of the grid impedance prediction model by calculating the accuracy index;

[0036] The Adam optimization algorithm is selected as the optimizer, and the power grid impedance prediction model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves or the preset number of iterations is reached. The test set is used to evaluate the performance of the model in the prediction task, and the trained power grid impedance prediction model is used to predict the current comprehensive feature data set to obtain the power grid impedance change trend and power grid impedance value in the next n period of time.

[0037] Furthermore, the method for obtaining the predicted grid impedance change rate within the next n periods of time according to the grid impedance change trend and the grid impedance value within the next n periods of time includes:

[0038] According to the predicted grid impedance value in the next n periods of time, the impedance change at adjacent time points is calculated by the impedance change rate calculation formula to obtain the grid impedance change rate in the next n periods of time; the impedance change rate calculation formula is: Where R(t) is the rate of change of grid impedance between time point t and time point t+1; Z pd (t) is the predicted grid impedance value at time point t; Z pd (t+1) is the predicted grid impedance value at time point t+1; t is the current time point; t+1 is the next adjacent time point; T′ is the time difference between the two adjacent time points of time point t and time point t+1.

[0039] Furthermore, the method for judging whether the grid impedance change has an interference effect on the BB-PLC signal transmission according to the predicted grid impedance change rate in the next n periods of time includes:

[0040] Compare the predicted grid impedance change rate within the next n periods of time with a preset grid impedance change rate threshold within the next n periods of time;

[0041] If the predicted grid impedance change rate within the next n periods of time is greater than or equal to the preset grid impedance change rate threshold within the next n periods of time, it is determined that the grid impedance change has an impact on the BB-PLC signal transmission;

[0042] If the predicted grid impedance change rate within the next n periods of time is less than the preset grid impedance change rate threshold within the next n periods of time, it is determined that the grid impedance change has no impact on the BB-PLC signal transmission.

[0043] Furthermore, the method for evaluating the interference impact degree of the grid impedance change on the BB-PLC signal transmission and obtaining the interference impact degree data includes:

[0044] The impact degree evaluation formula is defined to evaluate the interference degree of the grid impedance change on the BB-PLC signal transmission; the impact degree evaluation formula is: Where, BER(t) is the bit error rate of BB-PLC signal transmission at time point t; h z,t is the frequency of the impedance at time point t; g t is the impedance amplitude at time point t; θ z,t is the phase angle of impedance at time point t; |sin(θ z,t )| is the absolute value of the sine value of the phase angle of the impedance; ω 1 is the influence factor of the grid impedance change rate on the BB-PLC signal transmission bit error rate; ω 2 is the influence factor of impedance frequency on the bit error rate of BB-PLC signal transmission; ω 3 is the influence factor of impedance amplitude on the bit error rate of BB-PLC signal transmission; ω 4 is the influence factor of the impedance phase angle on the BB-PLC signal transmission bit error rate; σ is the BB-PLC signal transmission bit error rate adjustment coefficient;

[0045] The adjustment range of the BB-PLC signal transmission bit error rate adjustment coefficient is limited by the adjustment coefficient limit formula, and the adjustment coefficient limit formula is: Among them, σ′ is the bit error rate adjustment coefficient of BB-PLC signal transmission after limitation; a is the constant that controls the initial value of the adjustment coefficient; b is the constant that controls the decay speed of the adjustment coefficient over time; σ min is the minimum value of the bit error rate adjustment coefficient of BB-PLC signal transmission; Indicates taking and σ min The maximum of two values;

[0046] The interference degree of the grid impedance change on the BB-PLC signal transmission is evaluated by the BB-PLC signal transmission bit error rate at time point t; the first threshold BER of the BB-PLC signal transmission bit error rate at time point t is preset 1 and the second threshold BER of the BB-PLC signal transmission bit error rate at time point t 2 ;

[0047] The BB-PLC signal transmission bit error rate BER(t) at time point t is compared with BER 1 and BER 2 For comparison; if BER(t) <BER 1 , then it is determined that the interference degree of the grid impedance change on the BB-PLC signal transmission is at a low level; if BER 1 ≤BER(t), then the interference degree of grid impedance change on BB-PLC signal transmission is judged to be medium level; if BER 1 ≤BER(t) <BER 2, then it is determined that the interference effect of the grid impedance change on the BB-PLC signal transmission is at a high level;

[0048] The BB-PLC signal transmission bit error rate BER(t) and different levels of interference degree data at time point t are stored in the MySQL database to obtain the interference impact degree data.

[0049] The technical effects and advantages of the personalized image content recommendation method based on big data analysis of the present invention are as follows:

[0050] The present invention can effectively remove high-frequency noise in grid impedance and operating parameter data through LMS filter, improve signal stability, and provide a more reliable basis for subsequent data analysis and modeling; extract low-frequency components, effectively filter out short-term fluctuations, highlight long-term trends, and help avoid noise interference in the processing process; dynamically adjust the size of the step factor through the restriction formula, optimize the learning process of the LMS filter, and improve the algorithm convergence speed and accuracy; use MSE as the stopping criterion to ensure that the filter stops when the error is small, avoid overfitting and improve the robustness of the model;

[0051] Through the impact evaluation formula, combined with the frequency, amplitude, phase angle and change rate of the grid impedance and other multi-dimensional data, the impact of the grid impedance change on the BB-PLC signal transmission bit error rate can be accurately evaluated. This quantitative analysis can provide a scientific basis for the interference between the power network and the BB-PLC equipment, and help identify and optimize potential interference sources; by introducing the bit error rate adjustment coefficient and the adjustment range limit formula, the bit error rate evaluation model of the BB-PLC signal can be dynamically adjusted so that the bit error rate can be reasonably adjusted under different grid conditions; it can help to evaluate the quality of the BB-PLC signal in real time and optimize the transmission strategy when the grid load fluctuates, thereby reducing interference; based on the evaluation and adjustment mechanism, the system can effectively suppress the impact of changes in the grid environment, thereby improving the robustness of the BB-PLC signal and ensuring its stable operation in a complex grid environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of a grid impedance prediction and balanced adjustment method based on deep learning;

[0053] Figure 2 Schematic diagram of the grid impedance prediction and balancing adjustment module based on deep learning. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Example 1, please refer to Figure 1 As shown, the grid impedance prediction and balancing adjustment method based on deep learning described in this embodiment includes:

[0056] S1. Divide the power grid area according to the power grid topology diagram and identify the access points in the power grid area;

[0057] S2, collecting grid impedance data and grid operation status data of each access point within a preset time;

[0058] S3, preprocessing the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set;

[0059] S4. A grid impedance prediction model is obtained by training the comprehensive feature data set, and a grid impedance change trend and a grid impedance value in the next n periods of time are predicted based on the grid impedance prediction model; a predicted grid impedance change rate in the next n periods of time is obtained based on the grid impedance change trend and the grid impedance value in the next n periods of time;

[0060] S5. According to the predicted grid impedance change rate in the next n periods of time, determine whether the grid impedance change has an interference effect on the BB-PLC signal transmission; if an interference effect is produced, evaluate the interference effect of the grid impedance change on the BB-PLC signal transmission and obtain the interference effect degree data; the grid impedance prediction terminal automatically generates corresponding interference warning information according to the interference effect degree data.

[0061] The method of dividing the power grid area according to the power grid topology diagram and identifying the access points in the power grid area includes:

[0062] A grid topology diagram is obtained, which includes components of the grid, including substations, distribution equipment, transmission lines, access points and load points; the grid is divided into regions using a connected component algorithm in graph theory according to the electrical connection and geographical distribution of the grid, forming v grid regions, and the devices in each grid region are interconnected; nodes in each grid region that are connected to external devices or other regions are identified as access points, and the access points include substations and distribution boxes.

[0063] The purpose of regional division is to divide the power grid into multiple relatively independent areas. The equipment in each area is interconnected, which can simplify the management of the power grid and optimize the load distribution.

[0064] Connected component algorithm: In graph theory, a connected component refers to the set of all interconnected nodes in an undirected graph. By applying a connected component algorithm (such as depth-first search DFS or breadth-first search BFS) to the power grid topology graph, it is possible to identify the various connected areas in the power grid, where the devices in each area are directly connected and there are no connections between devices in other areas.

[0065] Algorithm steps:

[0066] Input topology diagram, each node represents a device in the power grid (such as substation, distribution box, transmission line, etc.), and each edge represents the electrical connection between devices.

[0067] Use the DFS / BFS algorithm to traverse the nodes and edges in the graph, identify all interconnected devices, and form multiple connected areas.

[0068] Regional division: Based on the division of connected components, the power grid area is divided into v areas, and the equipment in each area is directly connected through electrical connections.

[0069] Access point definition: Access point refers to the connection point between each grid area and external equipment or other areas. These access points are usually substations and distribution boxes, which are responsible for connecting with external load points, grids or other grid areas.

[0070] Access point identification: Based on the division of power grid areas, nodes connected to the outside in each area are identified as access points. Specifically, access points usually have the following characteristics:

[0071] Connection to external grids or other areas: The access point connects the transmission lines between different grid areas or external load points.

[0072] Substation: As the main interface between the power grid and external grids or different regions, the substation is usually located at the access point of the power grid.

[0073] Distribution box: A distribution box may be a point of connection between a small or medium-sized power grid area and other areas, or used to distribute electricity to different load points.

[0074] By dividing the power grid area and identifying the access points, the power grid can be monitored, load dispatched and fault isolated more efficiently. Regional division helps to quickly locate and isolate the fault area when a fault occurs, reducing the scope of the accident.

[0075] The grid impedance data includes the impedance value, impedance amplitude, phase angle of impedance and frequency of impedance at the grid access point; the grid operation status data includes the voltage data, current data, grid power data, load current and load voltage at the grid access point.

[0076] The method of preprocessing the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set includes:

[0077] Based on the box plot method, outliers in the grid impedance data and grid operation status data of each access point are identified and eliminated, and missing values ​​in the grid impedance data and grid operation status data of each access point are filled by interpolation method;

[0078] Performing timestamp alignment and standard deviation normalization on the grid impedance data and grid operation status data of each access point after missing value filling, to obtain a grid impedance characteristic data set and a grid operation status characteristic data set;

[0079] The LMS filter is used to perform denoising on the grid impedance parameter data and the grid operation parameter data to obtain the denoised grid impedance parameter data and the grid operation parameter data, and the steps include:

[0080] S41, extracting low-frequency components from the grid impedance parameter data and the grid operation parameter data by a moving average method, and constructing an expected signal vector d(m);

[0081] S42, initializing LMS filter parameters, initializing LMS filter parameters includes presetting the filter order M (i.e., the dimension of the weight vector), initializing a weight vector w(m), and setting an initial step size factor μ 0 (used to control the speed of weight update);

[0082] S43. Use the grid impedance parameter data and the grid operation parameter data as input signals and define the input signal vector: x(m) = [x(m), x(m-1), ..., x(m-M+1)] T ; Where x(m) is the input signal vector at the mth iteration; m is the number of iterations; T is the transposition operator symbol;

[0083] S44. Based on the weight vector w(m) and the input signal vector x(m), calculate the output signal vector of the LMS filter: y(m)=w T (m)·x(m); where y(m) is the output signal vector at the mth iteration; w T (m) is the transpose of the weight vector at the mth iteration;

[0084] By comparing the output signal vector y(m) and the expected signal vector d(m), the error signal vector is calculated: e(m)=d(m)-y(m); wherein e(m) is the error signal vector at the mth iteration, indicating the deviation between the output signal vector y(m) and the expected signal vector d(m);

[0085] S45. Based on the obtained error signal vector e(m) and input signal vector x(m), the weight vector is updated using the gradient descent method so that the output signal vector gradually approaches the expected signal vector to achieve denoising; the updated weight vector is: w(m+1)=w(m)+μ m ·e(m)x(m); where μ m is the step factor at the mth iteration; e(m)x(m) is the gradient term of the weight; the step factor is limited by the step factor limitation formula, and the size of the step factor is dynamically adjusted to ensure that the weight update in the gradient descent process can converge quickly and maintain convergence stability, preventing oscillation caused by too large a step factor or slow convergence caused by too small a step factor, thereby improving the optimization efficiency and reliability of the algorithm in the denoising task; the mean square error MSE of the error signal vector is calculated by the mean square error calculation formula;

[0086] S46. Repeat steps S43-S45 until the mean square value MSE of the error signal vector is less than the preset mean square value threshold θ, then stop the iteration, output the output signal vector of the current LMS filter, and obtain a power grid impedance characteristic data set and a power grid operation status characteristic data set; perform weighted fusion on the power grid impedance characteristic data set and the power grid operation status characteristic data set to obtain a comprehensive characteristic data set.

[0087] The step factor is limited by the step factor limiting formula. The methods for dynamically adjusting the size of the step factor include:

[0088] The step factor limiting formula is: Among them, μ′ m is the restricted step size factor; k is the attenuation factor that controls the rate at which the step size factor decreases with the number of iterations; b is the step size adaptive adjustment factor, which indicates the degree of influence of the step size factor itself on the adjustment. It is used to dynamically adjust the step size factor so that it depends not only on the number of iterations but also on the historical step size factor, thereby enhancing the flexibility and stability of the optimization process. For example, the initial step size factor μ 0 is 0.1, the attenuation factor k that controls the rate at which the step size factor decreases with the number of iterations is 0.01, and the step size adaptive adjustment factor b is 0.5. Assuming that the current number of iterations m is 50, the step size factor μ at the mth iteration m is 0.02, the step size factor after restriction

[0089] The method of performing weighted fusion on the grid impedance characteristic data set and the grid operation state characteristic data set to obtain a comprehensive characteristic data set includes:

[0090] The grid impedance characteristic data set and the grid operation status characteristic data set are fused through a weighted formula to obtain a comprehensive characteristic data set; the grid impedance characteristic data set is denoted as E 1 , the power grid operation status characteristic data set is recorded as E 2 ;

[0091] The weighted formula is: Rb = E 1 ·η 1 +E 2 ·η 2 ; Wherein, Rb is a comprehensive feature data set; η 1 is the weight coefficient of the grid impedance characteristic data set; η 2 is the weight coefficient of the power grid operation status characteristic data set.

[0092] The training method of the grid impedance prediction model includes:

[0093] The data set is divided into training set, validation set and test set to train the model and evaluate the model performance; the sample set is a subset of the data set, each sample set includes the historical comprehensive feature data set and the corresponding grid impedance change trend and grid impedance value in the next n periods of time; the deep learning library TensorFlow is used to build a grid impedance prediction model; the grid impedance prediction model includes input layer, LSTM layer and output layer;

[0094] The input layer of the model is used to input the historical comprehensive feature data set; the output layer of the model is used to output the corresponding grid impedance change trend and grid impedance value in the future n periods of time; the grid impedance prediction model is an LSTM model;

[0095] Define the loss function of the model, and use the L2 regularized mean square error loss function to measure the grid impedance change trend predicted by the grid impedance prediction model in the next n periods of time and the difference between the grid impedance value and the actual value; use the training set to train the grid impedance prediction model, and update the model parameters through the back propagation algorithm to minimize the loss function; use the validation set to evaluate the performance of the grid impedance prediction model by calculating the accuracy index;

[0096] The Adam optimization algorithm is selected as the optimizer, and the power grid impedance prediction model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves or the preset number of iterations is reached. The test set is used to evaluate the performance of the model in the prediction task, and the trained power grid impedance prediction model is used to predict the current comprehensive feature data set to obtain the power grid impedance change trend and power grid impedance value in the next n period of time.

[0097] According to the grid impedance change trend and grid impedance value in the next n periods of time, the method for obtaining the predicted grid impedance change rate in the next n periods of time includes:

[0098] According to the predicted grid impedance value in the next n periods of time, the impedance change at adjacent time points is calculated by the impedance change rate calculation formula to obtain the grid impedance change rate in the next n periods of time; the impedance change rate calculation formula is: Where R(t) is the rate of change of grid impedance between time point t and time point t+1; Z pd (t) is the predicted grid impedance value at time point t; Z pd (t+1) is the predicted grid impedance value at time point t+1; t is the current time point; t+1 is the next adjacent time point; T is the time difference between the two adjacent time points of time point t and time point t+1.

[0099] According to the predicted grid impedance change rate in the next n period of time, the method for judging whether the grid impedance change has an interference effect on the BB-PLC signal transmission includes:

[0100] Compare the predicted grid impedance change rate within the next n periods of time with a preset grid impedance change rate threshold within the next n periods of time;

[0101] If the predicted grid impedance change rate within the next n periods of time is greater than or equal to the preset grid impedance change rate threshold within the next n periods of time, it is determined that the grid impedance change has an impact on the BB-PLC signal transmission;

[0102] If the predicted grid impedance change rate within the next n periods of time is less than the preset grid impedance change rate threshold within the next n periods of time, it is determined that the grid impedance change has no impact on the BB-PLC signal transmission.

[0103] Methods for evaluating the interference effect of grid impedance change on BB-PLC signal transmission and obtaining interference effect data include:

[0104] The impact degree evaluation formula is defined to evaluate the interference degree of the grid impedance change on the BB-PLC signal transmission; the impact degree evaluation formula is: Where, BER(t) is the bit error rate of BB-PLC signal transmission at time point t; h z,t is the frequency of the impedance at time point t; g t is the impedance amplitude at time point t; θ z,t is the phase angle of impedance at time point t; |sin(θ z,t )| is the absolute value of the sine value of the phase angle of the impedance; ω 1is the influence factor of the grid impedance change rate on the BB-PLC signal transmission bit error rate; ω 2 is the influence factor of impedance frequency on the bit error rate of BB-PLC signal transmission; ω 3 is the influence factor of impedance amplitude on the bit error rate of BB-PLC signal transmission; ω 4 is the influence factor of the impedance phase angle on the BB-PLC signal transmission bit error rate; σ is the BB-PLC signal transmission bit error rate adjustment coefficient;

[0105] For example, the grid impedance change rate R(t) between time point t and time point t+1 is 0.05, and the impedance frequency h at time point t is z,t is 50Hz, the impedance amplitude g at time point t t The impedance phase angle θ at time point t is 100Ω. z,t is 45°, the influence factor of the grid impedance change rate on the BB-PLC signal transmission bit error rate ω 1 is 0.25, the influence factor of impedance frequency on the bit error rate of BB-PLC signal transmission ω 2 is 0.25, the influence factor of impedance amplitude on the bit error rate of BB-PLC signal transmission ω 3 =0.25, the influence factor of the impedance phase angle on the bit error rate of BB-PLC signal transmission ω 4 is 0.25, and the BB-PLC signal transmission bit error rate adjustment coefficient σ is 100. Then the BB-PLC signal transmission bit error rate at time point t is The adjustment range of the BB-PLC signal transmission bit error rate adjustment coefficient is limited by the adjustment coefficient limit formula, and the adjustment coefficient limit formula is: Among them, σ′ is the bit error rate adjustment coefficient of BB-PLC signal transmission after limitation; a is the constant that controls the initial value of the adjustment coefficient; b is the constant that controls the decay speed of the adjustment coefficient over time; σ min is the minimum value of the bit error rate adjustment coefficient of BB-PLC signal transmission; Indicates taking and σ min The maximum of two values;

[0106] For example, the BB-PLC signal transmission bit error rate adjustment coefficient σ is 100, the constant a that controls the initial value of the adjustment coefficient is 1, the constant b that controls the decay speed of the adjustment coefficient over time is 0.1, and the minimum value σ of the BB-PLC signal transmission bit error rate adjustment coefficient is min is 50;

[0107] When t = 0, When t = 1, When t = 2, When t=3, The interference degree of the grid impedance change on the BB-PLC signal transmission is evaluated by the BB-PLC signal transmission bit error rate at time point t; the first threshold BER of the BB-PLC signal transmission bit error rate at time point t is preset 1 and the second threshold BER of the BB-PLC signal transmission bit error rate at time point t 2 ;

[0108] The BB-PLC signal transmission bit error rate BER(t) at time point t is compared with BER 1 and BER 2 For comparison; if BER(t) <BER 1 , then it is determined that the interference degree of the grid impedance change on the BB-PLC signal transmission is at a low level; if BER 1 ≤BER(t), then the interference degree of grid impedance change on BB-PLC signal transmission is judged to be medium level; if BER 1 ≤BER(t) <BER 2 , then it is determined that the interference effect of the grid impedance change on the BB-PLC signal transmission is at a high level;

[0109] The BB-PLC signal transmission bit error rate BER(t) and different levels of interference degree data at time point t are stored in the MySQL database to obtain the interference impact degree data.

[0110] The preset grid impedance change rate threshold within the next n periods of time is set by the staff, and the grid impedance change rates in different future n periods of time are collected through the grid impedance prediction terminal, and the average value of the grid impedance change rates in multiple future n periods of time is taken as the preset grid impedance change rate threshold within the future n periods of time; similarly, the preset mean square value threshold and the first threshold BER of the BB-PLC signal transmission bit error rate at time point t are set 1 and the second threshold BER of the BB-PLC signal transmission bit error rate at time point t 2 .

[0111] This embodiment can effectively remove high-frequency noise in grid impedance and operating parameter data through the LMS filter, improve the stability of the signal, and provide a more reliable basis for subsequent data analysis and modeling; extract low-frequency components, effectively filter out short-term fluctuations, highlight long-term trends, and help avoid noise interference during the processing process; dynamically adjust the size of the step factor through the restriction formula, optimize the learning process of the LMS filter, and improve the algorithm convergence speed and accuracy; use MSE as the stopping criterion to ensure that the filter stops when the error is small, avoid overfitting and improve the robustness of the model;

[0112] Through the impact evaluation formula, combined with the frequency, amplitude, phase angle and change rate of the grid impedance and other multi-dimensional data, the impact of the grid impedance change on the BB-PLC signal transmission bit error rate can be accurately evaluated. This quantitative analysis can provide a scientific basis for the interference between the power network and the BB-PLC equipment, and help identify and optimize potential interference sources; by introducing the bit error rate adjustment coefficient and the adjustment range limit formula, the bit error rate evaluation model of the BB-PLC signal can be dynamically adjusted so that the bit error rate can be reasonably adjusted under different grid conditions; it can help to evaluate the quality of the BB-PLC signal in real time and optimize the transmission strategy when the grid load fluctuates, thereby reducing interference; based on the evaluation and adjustment mechanism, the system can effectively suppress the impact of changes in the grid environment, thereby improving the robustness of the BB-PLC signal and ensuring its stable operation in a complex grid environment.

[0113] Example 2

[0114] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a grid impedance prediction and balancing adjustment system based on deep learning is provided, including:

[0115] A power grid node division module is used to divide the power grid area according to the power grid topology diagram and identify the access points in the power grid area;

[0116] A data acquisition module, used to collect grid impedance data and grid operation status data of each access point within a preset time;

[0117] A data processing module is used to pre-process the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set;

[0118] The power grid impedance prediction module is used to obtain a power grid impedance prediction model based on the comprehensive feature data set training, and to predict the power grid impedance change trend and power grid impedance value in the next n periods of time based on the power grid impedance prediction model; and to obtain the predicted power grid impedance change rate in the next n periods of time based on the power grid impedance change trend and power grid impedance value in the next n periods of time;

[0119] The grid impedance interference warning module is used to determine whether the grid impedance change has an interference effect on the BB-PLC signal transmission based on the predicted grid impedance change rate in the next n periods of time; if an interference effect occurs, the interference effect of the grid impedance change on the BB-PLC signal transmission is evaluated to obtain the interference effect degree data; the grid impedance prediction terminal automatically generates interference warning information based on the interference effect degree data.

[0120] Since the electronic device introduced in this embodiment is an electronic device used to implement the grid impedance prediction and balanced adjustment method based on deep learning in the embodiment of the present application, based on the grid impedance prediction and balanced adjustment method based on deep learning introduced in the embodiment of the present application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as the technical personnel of this field implement the electronic device used by the grid impedance prediction and balanced adjustment method based on deep learning in the embodiment of the present application, it belongs to the scope of protection of this application.

[0121] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0122] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A grid impedance prediction and balancing adjustment method based on deep learning, characterized in that: include: S1. Divide the power grid area according to the power grid topology diagram and identify the access points in the power grid area; S2, collecting grid impedance data and grid operation status data of each access point within a preset time; S3, preprocessing the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set; S4. A grid impedance prediction model is obtained by training the comprehensive feature data set, and a grid impedance change trend and a grid impedance value in the next n periods of time are predicted based on the grid impedance prediction model; a predicted grid impedance change rate in the next n periods of time is obtained based on the grid impedance change trend and the grid impedance value in the next n periods of time; S5. According to the predicted grid impedance change rate in the next n periods of time, determine whether the grid impedance change has an interference effect on the BB-PLC signal transmission; if an interference effect is produced, evaluate the interference effect of the grid impedance change on the BB-PLC signal transmission and obtain the interference effect degree data; the grid impedance prediction terminal automatically generates corresponding interference warning information according to the interference effect degree data.

2. The method for predicting and balancing power grid impedance based on deep learning according to claim 1, characterized in that: The method of dividing the power grid area according to the power grid topology diagram and identifying the access points in the power grid area includes: A grid topology diagram is obtained, which includes components of the grid, including substations, distribution equipment, transmission lines, access points and load points; the grid is divided into regions using a connected component algorithm in graph theory according to the electrical connection and geographical distribution of the grid, forming v grid regions, and the devices in each grid region are interconnected; nodes in each grid region that are connected to external devices or other regions are identified as access points, and the access points include substations and distribution boxes.

3. The grid impedance prediction and balanced adjustment method based on deep learning according to claim 2 is characterized in that: The grid impedance data includes the impedance value, impedance amplitude, phase angle of impedance and frequency of impedance at the grid access point; the grid operation status data includes voltage data, current data, grid power data, load current and load voltage at the grid access point.

4. The method for predicting and balancing power grid impedance based on deep learning according to claim 3, characterized in that: The method of preprocessing the grid impedance data and grid operation status data of each access point to obtain a comprehensive feature data set includes: Based on the box plot method, outliers in the grid impedance data and grid operation status data of each access point are identified and eliminated, and missing values ​​in the grid impedance data and grid operation status data of each access point are filled by interpolation method; Performing timestamp alignment and standard deviation normalization on the grid impedance data and grid operation status data of each access point after missing value filling, to obtain a grid impedance characteristic data set and a grid operation status characteristic data set; The LMS filter is used to perform denoising on the grid impedance parameter data and the grid operation parameter data to obtain the denoised grid impedance parameter data and the grid operation parameter data, and the steps include: S41, extracting low-frequency components from the grid impedance parameter data and the grid operation parameter data by a moving average method, and constructing an expected signal vector d(m); S42, initializing LMS filter parameters, initializing LMS filter parameters includes presetting filter order M, initializing a weight vector w(m) and setting an initial step size factor μ0; S43. Use the grid impedance parameter data and the grid operation parameter data as input signals and define the input signal vector: x(m) = [x(m), x(m-1), ..., x(m-M+1)] T ; Where x(m) is the input signal vector at the mth iteration; m is the number of iterations; T is the transposition operator symbol; S44. Based on the weight vector w(m) and the input signal vector x(m), calculate the output signal vector of the LMS filter: y(m)=w T (m)·x(m); where y(m) is the output signal vector at the mth iteration; w T (m) is the transpose of the weight vector at the mth iteration; By comparing the output signal vector y(m) and the desired signal vector d(m), the error signal vector is calculated: e(m)=d(m)-y(m); where e(m) is the error signal vector at the mth iteration, representing the deviation between the output signal vector y(m) and the expected signal vector d(m); S45. Based on the obtained error signal vector e(m) and the input signal vector x(m), the weight vector is updated using the gradient descent method; the updated weight vector is: w(m+1)=w(m)+μ m ·e(m)x(m); where μ m is the step factor at the mth iteration; e(m)x(m) is the gradient term of the weight; the step factor is limited by the step factor limitation formula, and the size of the step factor is dynamically adjusted; the mean square value MSE of the error signal vector is calculated by the mean square error calculation formula; S46. Repeat steps S43-S45 until the mean square value MSE of the error signal vector is less than the preset mean square value threshold θ, then stop the iteration, output the output signal vector of the current LMS filter, and obtain a power grid impedance characteristic data set and a power grid operation status characteristic data set; perform weighted fusion on the power grid impedance characteristic data set and the power grid operation status characteristic data set to obtain a comprehensive characteristic data set.

5. The method for predicting and balancing power grid impedance based on deep learning according to claim 4, characterized in that: The method of limiting the step factor by a step factor limiting formula and dynamically adjusting the size of the step factor includes: The step factor limiting formula is: Among them, μ′ m is the step size factor after limitation; k is the attenuation factor that controls the rate at which the step size factor decreases with the number of iterations; b is the step size adaptive adjustment factor.

6. The method for predicting and balancing power grid impedance based on deep learning according to claim 5, characterized in that: The method of performing weighted fusion on the grid impedance characteristic data set and the grid operation state characteristic data set to obtain a comprehensive characteristic data set includes: The grid impedance characteristic data set and the grid operation state characteristic data set are fused by a weighted formula to obtain a comprehensive characteristic data set; the grid impedance characteristic data set is recorded as E1, and the grid operation state characteristic data set is recorded as E2; The weighted formula is: Rb=E1·η1+E2·η2; wherein Rb is a comprehensive characteristic data set; η1 is a weight coefficient of a power grid impedance characteristic data set; and η2 is a weight coefficient of a power grid operation status characteristic data set.

7. The method for predicting and balancing power grid impedance based on deep learning according to claim 6, characterized in that: The training method of the power grid impedance prediction model includes: The data set is divided into training set, validation set and test set to train the model and evaluate the model performance; the sample set is a subset of the data set, each sample set includes the historical comprehensive feature data set and the corresponding grid impedance change trend and grid impedance value in the next n periods of time; the deep learning library TensorFlow is used to build a grid impedance prediction model; the grid impedance prediction model includes input layer, LSTM layer and output layer; The input layer of the model is used to input the historical comprehensive feature data set; the output layer of the model is used to output the corresponding grid impedance change trend and grid impedance value in the future n periods of time; the grid impedance prediction model is an LSTM model; Define the loss function of the model, and use the L2 regularized mean square error loss function to measure the grid impedance change trend predicted by the grid impedance prediction model in the next n periods of time and the difference between the grid impedance value and the actual value; use the training set to train the grid impedance prediction model, and update the model parameters through the back propagation algorithm to minimize the loss function; use the validation set to evaluate the performance of the grid impedance prediction model by calculating the accuracy index; The Adam optimization algorithm is selected as the optimizer, and the power grid impedance prediction model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves or the preset number of iterations is reached. The test set is used to evaluate the performance of the model in the prediction task, and the trained power grid impedance prediction model is used to predict the current comprehensive feature data set to obtain the power grid impedance change trend and power grid impedance value in the next n period of time.

8. The method for predicting and balancing power grid impedance based on deep learning according to claim 7, characterized in that: The method for obtaining the predicted grid impedance change rate within the next n time periods based on the grid impedance change trend and grid impedance value within the next n time periods includes: According to the predicted grid impedance value in the next n periods of time, the impedance change at adjacent time points is calculated by the impedance change rate calculation formula to obtain the grid impedance change rate in the next n periods of time; the impedance change rate calculation formula is: Where R(t) is the rate of change of grid impedance between time point t and time point t+1; Z pd (t) is the predicted grid impedance value at time point t; Z pd (t+1) is the predicted grid impedance value at time point t+1; t is the current time point; t+1 is the next adjacent time point; T′ is the time difference between the two adjacent time points of time point t and time point t+1.

9. The method for predicting and balancing power grid impedance based on deep learning according to claim 8, characterized in that: The method for determining whether the grid impedance change has an interference effect on the BB-PLC signal transmission based on the predicted grid impedance change rate within the next n time periods includes: Comparing the predicted grid impedance change rate within the next n time periods with the preset grid impedance change rate threshold within the next n time periods; If the predicted grid impedance change rate within the next n time periods is greater than or equal to the preset grid impedance change rate threshold within the next n time periods, it is determined that the grid impedance change has an impact on the BB-PLC signal transmission; If the predicted grid impedance change rate within the next n time periods is less than the preset grid impedance change rate threshold within the next n time periods, it is determined that the grid impedance change has no impact on the BB-PLC signal transmission.

10. The method for predicting and balancing power grid impedance based on deep learning according to claim 9, characterized in that: The method for evaluating the interference effect degree of the grid impedance change on the BB-PLC signal transmission and obtaining the interference effect degree data includes: The impact degree evaluation formula is defined to evaluate the interference degree of the grid impedance change on the BB-PLC signal transmission; the impact degree evaluation formula is: Where, BER(t) is the bit error rate of BB-PLC signal transmission at time point t; h z,t is the frequency of the impedance at time point t; g t is the impedance amplitude at time point t; θ z,t is the phase angle of impedance at time point t; |sin(θ z,t )| is the absolute value of the sine value of the phase angle of the impedance; ω1 is the influence factor of the grid impedance change rate on the BB-PLC signal transmission bit error rate; ω2 is the influence factor of the impedance frequency on the BB-PLC signal transmission bit error rate; ω3 is the influence factor of the impedance amplitude on the BB-PLC signal transmission bit error rate; ω4 is the influence factor of the impedance phase angle on the BB-PLC signal transmission bit error rate; σ is the BB-PLC signal transmission bit error rate adjustment coefficient; The adjustment range of the BB-PLC signal transmission bit error rate adjustment coefficient is limited by the adjustment coefficient limit formula, and the adjustment coefficient limit formula is: Among them, σ′ is the bit error rate adjustment coefficient of BB-PLC signal transmission after limitation; a is the constant that controls the initial value of the adjustment coefficient; b is the constant that controls the decay speed of the adjustment coefficient over time; σ min is the minimum value of the bit error rate adjustment coefficient of BB-PLC signal transmission; Indicates taking and σ min The maximum of two values; Evaluating the interference effect degree of the grid impedance change on the BB-PLC signal transmission through the bit error rate of the BB-PLC signal transmission at time point t; presetting the first threshold BER1 of the BB-PLC signal transmission bit error rate and the second threshold BER2 of the BB-PLC signal transmission bit error rate at time point t; Comparing the bit error rate BER(t) of the BB-PLC signal transmission at time point t with BER1 and BER2 respectively; if BER(t) < BER1, it is determined that the interference effect degree of the grid impedance change on the BB-PLC signal transmission is at a low level; if BER1 ≤ BER(t), it is determined that the interference effect degree of the grid impedance change on the BB-PLC signal transmission is at a medium level; if BER1 ≤ BER(t) < BER2, it is determined that the interference effect degree of the grid impedance change on the BB-PLC signal transmission is at a high level; Storing the bit error rate BER(t) of the BB-PLC signal transmission at time point t and the interference degree data of different levels into the MySQL database to obtain the interference effect degree data.

Images