A multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance

Abstract

The application discloses a kind of multi-source fusion intelligent power grid harmonic dynamic management systems based on harmonic impedance, and relate to the technical field of intelligent power grid, the system includes: multi-source data acquisition module: with the aid of intelligent electric meter, sensor equipment, collect harmonic impedance, power grid operation, weather and user power consumption behavior data, the application constructs multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance, realizes the comprehensive, accurate management to power grid harmonic problem, the system can automatically collect and process a large amount of data, including harmonic impedance, power grid operation, weather and user power consumption behavior data, provides strong support for the identification and analysis of harmonic problem, through the multi-source depth fusion of these data, system can generate more accurate, comprehensive comprehensive data set, to realize the accurate positioning tracking of harmonic source and the in-depth analysis of harmonic propagation path, improve the efficiency and accuracy of power grid harmonic management, provide strong guarantee for the safe and stable operation of power grid.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart grid, in particular to a multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance. BACKGROUND

[0002] With the rapid development of smart grid and the wide application of power electronic equipment, the large-scale construction of 750kV and above AC transmission network and the construction of large-scale power grid security and defense system have put forward higher requirements for power grid harmonic management. In the collaborative operation of the smart dispatching system, harmonic problems have become a key factor affecting the stability of power supply. Harmonics not only lead to a decrease in power quality, but also can cause equipment overheating, relay protection misoperation and other problems, which seriously threaten the safe operation of 750kV and above transmission lines. The traditional harmonic management technology mainly has the following shortcomings:

[0003] Single dimension of data acquisition

[0004] The existing system cannot comprehensively cover the harmonic impedance characteristic monitoring of 750kV and above transmission lines, and has insufficient collaborative influence analysis capability for distributed energy and nonlinear load in large-scale power grids, which cannot meet the demand of the smart dispatching system for multi-time and space scale data.

[0005] Limited analysis method accuracy

[0006] Traditional positioning algorithms rely on single-point measurement and cannot adapt to harmonic propagation path tracking under the complex topology structure of large-scale power grids, especially in 750kV and above power grids, where harmonic reflection and superposition effect is significant, and traditional methods are difficult to accurately identify harmonic source dynamics.

[0007] Lack of coordination in management decisions

[0008] The existing scheme is not deeply coupled with the smart dispatching system, and cannot respond in real time to the influence of changes in power grid operation mode (such as power flow transfer and equipment switching) on harmonic distribution, resulting in insufficient coordination optimization of treatment measures and large-scale power grid security and defense system. Therefore, developing a multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance will significantly improve the level of power grid harmonic management and ensure the safe and stable operation of smart grid. SUMMARY

[0009] The purpose of the present application is to make up for the deficiencies of the prior art, and provide a multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance, which integrates multiple functional modules such as multi-source data acquisition, data deep fusion, harmonic source positioning and tracking, harmonic propagation path analysis, and comprehensive evaluation and management decision-making, and realizes comprehensive monitoring, accurate analysis and scientific management of power grid harmonic problems through intelligent means, which can not only accurately identify and locate harmonic sources, but also effectively evaluate harmonic risk levels, and develop targeted harmonic control schemes according to the evaluation results.

[0010] To solve the above technical problems, the present application provides the following technical solutions: a multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance, which comprises:

[0011] The multi-source data acquisition module: with the aid of smart meters and sensor devices, the data of harmonic impedance, power grid operation, weather and user electricity behavior are collected, and through the embedded edge storage device, the data is cleaned and normalized locally, and then transmitted to the multi-source data deep fusion module;

[0012] The multi-source data deep fusion module: receiving the data transmitted from the multi-source data acquisition module, using principal component analysis to reduce the dimension of the data, and using multi-source data fusion algorithm to fuse different types of data to generate a comprehensive data set;

[0013] The harmonic source positioning and tracking module: receiving the comprehensive data set output by the data preprocessing and fusion module, based on the characteristics of harmonic impedance, using the positioning algorithm based on harmonic impedance gradient to calculate the harmonic source position coordinates, and simultaneously monitoring and continuously tracking the harmonic source dynamics;

[0014] The harmonic propagation path analysis module: obtaining the power grid topology structure data and the harmonic source information provided by the harmonic source positioning and tracking module, using the grid nodes and line abstraction to construct the graph structure model of the power grid, calculating the complex power flow based on the node voltage and line electrical parameters, determining the propagation direction of the harmonic according to the phase of the complex power flow, and determining the propagation intensity of the harmonic on different lines by calculating the propagation intensity coefficient;

[0015] The comprehensive evaluation and management decision-making module: according to the output of the multi-source data deep fusion module, the harmonic source positioning and tracking module and the harmonic propagation path analysis module, using the long short-term memory network to predict the probability of equipment failure caused by harmonic, and developing a harmonic management scheme according to the evaluation results.

[0016] Further, the data types and collection devices in the multi-source data acquisition module are:

[0017] Harmonic impedance data:

[0018] Data type: Amplitude, phase relationship of harmonic voltage and current in power grid, and harmonic impedance value at specific frequency;

[0019] Collecting equipment: Harmonic analyzer and harmonic impedance measuring device;

[0020] Power grid operation data:

[0021] Data type: Voltage, current, power and frequency;

[0022] Collecting equipment: Smart meter, voltage transformer, current transformer and synchronous phasor measurement device

[0023] Data type: Temperature, humidity, wind speed and air pressure;

[0024] Collecting equipment: Temperature sensor, humidity sensor, anemometer and barometer;

[0025] User electricity consumption behavior data:

[0026] Data type: Electricity consumption time, electricity consumption power, appliance usage pattern and abnormal electricity consumption behavior,

[0027] Collecting equipment: Smart meter, non-intrusive load monitoring device for identifying user electricity consumption equipment through current waveform analysis, and user side power quality monitoring sensor.

[0028] Further, the principal component analysis is used in the multi-source data deep fusion module to reduce the dimension of data, the data preprocessed by the multi-source data collection module is received, a covariance matrix C of the data is constructed, and a calculation formula is as follows: wherein is the i-th characteristic value of the k-th sample after standardization, n is the sample quantity, is the mean value of the i-th characteristic after standardization, is the j-th characteristic value of the k-th sample after standardization, indicates the mean value of the j-th characteristic after standardization, the eigenvalue λ i and the corresponding eigenvector e i of the covariance matrix C are obtained by eigenvalue decomposition of the covariance matrix C, i=1, 2, …, p, p is the original characteristic quantity, λ i is the i-th eigenvalue of the covariance matrix C, and indicates the variance of data in the direction, e i indicates the eigenvector of the covariance matrix C corresponding to the eigenvalue λ i , and indicates the direction of data change, the eigenvalues are arranged in descending order, the first m principal components with a cumulative contribution rate of 85%-95% are reserved according to the size of the eigenvalues, and a calculation formula is as follows: Where m is the number of principal components selected, the eigenvectors corresponding to the first m selected principal components form the transformation matrix X. * Multiplying by the transformation matrix E, X * The data matrix after standardization is used to obtain the dimensionality-reduced data matrix Y, which is expressed by the formula: Y = X * E is a transformation matrix composed of the eigenvectors corresponding to the first m principal components.

[0029] Furthermore, the multi-source data deep fusion module uses a multi-source data fusion algorithm to fuse different types of data. Let the collected i-th type of data be D. i Its adaptive weight is w i The merged data is D f The calculation formula is: Where c represents the number of data samples and o represents the number of data types.

[0030] Furthermore, the harmonic source location and tracking module uses a location algorithm based on harmonic impedance gradients to calculate the coordinates of the harmonic source. Let the harmonic impedance at a point x in the power grid be Z(x), and the harmonic source location be x0. The algorithm iteratively calculates x... k+1 To approximate the harmonic source location x0, the calculation formula is: in It is x k The gradient of harmonic impedance at x k It is the point that approximates the location of the harmonic source in the k-th iteration, x k+1 In the (k+1)th iteration, the point that approximates the location of the harmonic source is defined by α, which is a step size parameter ranging from 0.1 to 0.5. α is dynamically adjusted based on the fluctuations in the real-time harmonic impedance gradient, using the following formula: Where: α0 is the initial step size coefficient, and the calculation formula is: The harmonic impedance gradient at the initial iteration point, where η is a proportionality coefficient ranging from 0.2 to 0.8, and δ is a smoothing constant to prevent the denominator from failing when the gradient is zero. denoted as the harmonic impedance gradient at the k-th iteration, λ is the sensitivity adjustment factor, and a value of 0.5 to 1.5 is recommended. ∈ is a very small positive number to avoid a denominator of zero.

[0031] Furthermore, the harmonic propagation path analysis module utilizes the abstraction of power grid nodes and lines to construct a graph structure model of the power grid. The obtained power grid topology data is used to abstract nodes in the power grid as vertices of a graph, and lines as edges connecting those vertices, thus constructing a graph structure model of the power grid G ​​= (V, E), where α is the set of nodes, E is the set of edges, and each node v... i ∈V corresponds to an actual node in the power grid, and each edge eij ∈E connects to node v i and v j This represents the actual power transmission line and assigns electrical parameters to each side.

[0032] Furthermore, in the harmonic propagation path analysis module, the determination of the harmonic propagation direction involves calculating the complex power flow S between two nodes i and j connected by any line in the power grid. ij The calculation formula is: Where V i It is the voltage phasor of node i, obtained through a voltage measuring device installed at node i. isI ij The conjugate of the complex number, I ij I is the current phasor flowing from node i to node j. ij The voltage V through node i i The voltage V at node j j And the electrical parameters of the circuit are calculated to include resistance R. ij Inductor L ij and capacitor C ij The circuit, current phasor I ij The calculation formula is: Where ω is the angular frequency, and π = 2πf h f h It is the frequency of the h-th harmonic, and the calculation formula is: f h = h·f0, where h is the harmonic order, f0 is the fundamental frequency of the power grid (50Hz), and the direction of harmonic propagation is determined by the complex power flow S. ij The phase determines when Re(S) ij If Re > 0, Re represents taking the real part, indicating that power flows from node i to node j, that is, the harmonic propagates from node i to node j along this line; when Re(S) > 0, Re represents taking the real part, indicating that power flows from node i to node j, that is, the harmonic propagates from node i to node j along this line; ij If ) > 0, the harmonic propagation direction is opposite.

[0033] Furthermore, the harmonic propagation path analysis module determines the propagation intensity of harmonics on different paths by calculating the propagation intensity coefficient, let the propagation intensity coefficient be K. ij The calculation formula is: Where |S ij | is the complex power current S ij The modulus, N, represents the magnitude of the harmonic power on the line. i It is the set of adjacent nodes of node i. This represents the sum of the complex power flow modes between node i and all its adjacent nodes.

[0034] Further, the comprehensive evaluation and management decision module uses a long short-term memory network to predict the probability of harmonic-induced equipment failure, from the output of the multi-source data deep fusion module, the harmonic source positioning and tracking module, and the harmonic propagation path analysis module, extracts data related to harmonics and equipment status, cleans and standardizes the data, and arranges the data in chronological order to construct time series data, constructs an LSTM model, determines the number of layers and the number of neurons in each layer, and the LSTM unit contains an input gate f t , where x t is the current input, h t-I is the hidden state at the last time step, W ii and W hi are weight matrices, b i is a bias vector, and sigma is a sigmoid activation function, the forget gate f t : f t = sigma(W if x t + W hf h t-1 + b f ), the candidate memory cell where tanh is the hyperbolic tangent activation function, the memory cell C t : where denotes element-wise multiplication, the output gate o t : o t = sigma(W i ox t + W h oh t-1 + b o ), the hidden state h t : h t = o t ⊙ tanh(C t ), a fully connected layer is added, and the output is: y = W out h last + b out , where W out is the weight matrix of the fully connected layer, h last is the hidden state at the last time step of the last LSTM layer, and b out is the bias vector, then a fully connected layer is added, and the output is: y = W out h last + b out , where W out is the weight matrix of the fully connected layer, h last is the hidden state at the last time step of the last LSTM layer, and b outis a bias vector, the mean square error is selected as the loss function, the Adam optimizer is used for parameter updating, the preprocessed data is divided into a training set, a validation set and a test set, the total amount of data is N, the proportion of the training set is p1, the proportion of the validation set is p2, the proportion of the test set is p3, and p1+p2+p3=I, then the number of training set N1=p1N, the number of validation set N2=p2N, the number of test set N3=p3N, the training set data is input into the LSTM model for training, in each training step, the predicted value is calculated by forward propagation Then the loss L is calculated according to the loss function, the gradient is calculated by the back propagation algorithm, and finally the model parameters are updated using the optimizer. The real-time collected multi-source data is processed according to the data preprocessing steps and input into the trained LSTM model. The model outputs the predicted device failure probability P fault When P fault ≤T1, it is determined as a low risk level, indicating that the harmonic condition in the power grid is relatively good; when T1 fault ≤T2, it is determined as a medium risk level, and there is a certain degree of harmonic problem in the power grid; when , it is determined as a high risk level, indicating that the power grid harmonic pollution is serious, wherein T1 and T2 are threshold values set based on industry standards.

[0035] Further, the comprehensive evaluation and management decision module formulates a harmonic management scheme according to the evaluation results as follows:

[0036] Low risk level: adopt the strategy of regular monitoring, continuously track the harmonic condition of the power grid, and at the same time, carry out preventive maintenance on the equipment;

[0037] Medium risk level: strengthen the monitoring frequency, further analyze the harmonic source, determine the main harmonic generating equipment or area, and take measures such as adjusting the load distribution, optimizing the transformer tap, and installing small harmonic control devices;

[0038] High risk level: immediately start emergency treatment measures, shut down and rectify the seriously over-standard harmonic source equipment, replace large-capacity and high-performance active power filters and static var compensators, and take measures such as installing harmonic isolation devices on key equipment seriously affected by harmonics.

[0039] Compared with the prior art, a multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance has the following beneficial effects:

[0040] One, the application realizes comprehensive and accurate management of power grid harmonic problems by constructing a multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance, which can automatically collect and process a large amount of data including harmonic impedance, power grid operation, weather and user power consumption behavior data, providing strong support for harmonic problem identification and analysis, and through multi-source deep fusion of these data, the system can generate more accurate and comprehensive comprehensive data set, and then realize accurate positioning tracking of harmonic sources and in-depth analysis of harmonic propagation path, improving the efficiency and accuracy of power grid harmonic management, and providing strong guarantee for safe and stable operation of power grid.

[0041] Two, the positioning algorithm based on harmonic impedance gradient can quickly and accurately calculate the harmonic source position coordinates, realize real-time monitoring and continuous tracking of harmonic sources, and at the same time, the graph structure model of power grid is constructed by using power grid nodes and line abstraction, combined with complex power flow calculation and determination of harmonic propagation intensity coefficient, the system can comprehensively evaluate the propagation of harmonic in power grid and potential influence, not only improve the identification and analysis ability of harmonic problem, but also provide strong basis for formulating scientific and reasonable harmonic management scheme, significantly reduce the level of power grid harmonic pollution through the implementation of targeted management measures, and improve the efficiency and power quality of power grid operation.

[0042] Other advantages, objects and features of the present application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, without creative labor, other drawings can also be obtained from these drawings.

[0044] Figure 1 Flow chart of a multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance;

[0045] Figure 2 Framework diagram of a multi-source fusion intelligent power grid harmonic dynamic management system based on harmonic impedance. DETAILED DESCRIPTION

[0046] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purpose, the following will combine the drawings and preferred embodiments to explain the specific embodiments, structures, features and effects of the present application in detail.

[0047] Example 1:

[0048] Harmonic management of smart grids in urban commercial areas.

[0049] In urban commercial areas, numerous commercial electrical equipment, office equipment, and lighting systems pose complex harmonic problems to the power grid.

[0050] Data Acquisition: Smart meters, harmonic analyzers, voltage transformers, current transformers, and synchronous phasor measurement devices are installed at various distribution boxes, transformers, and major electrical equipment locations in the commercial area. Smart meters collect user electricity consumption behavior data, including electricity usage time, power, appliance usage patterns, voltage, current, power, and frequency data. Harmonic analyzers collect the amplitude and phase relationship of harmonic voltage and current, as well as harmonic impedance values ​​at specific frequencies. Temperature sensors, humidity sensors, anemometers, and barometers collect meteorological data. After being collected by the multi-source data acquisition module, this data is cleaned and normalized locally through embedded edge storage devices before being transmitted to the multi-source data deep fusion module.

[0051] Data fusion and dimensionality reduction: The multi-source data deep fusion module receives the preprocessed data and constructs the covariance matrix C of the data. The calculation formula is as follows: in It is the i-th feature value of the k-th sample after standardization, where n is the number of samples. It is the mean of the i-th feature after standardization. It is the j-th sample after standardization. Let λ represent the mean of the j-th feature after standardization. Then, perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues ​​λ. i and the corresponding eigenvector e i i = 1, 2, ..., p, where p is the number of original features, λ i The i-th eigenvalue of the covariance matrix C represents the magnitude of the variance of the data in that direction. The eigenvalues ​​are sorted from largest to smallest. Based on the magnitude of the eigenvalues, the top m principal components with a cumulative contribution rate of 85%-95% are retained. The formula for calculating the cumulative contribution rate is: Where m is the number of principal components selected, the eigenvectors corresponding to the first m selected principal components form the transformation matrix X. * Multiplying by the transformation matrix E, X * The data matrix after standardization is used to obtain the dimension-reduced data matrix Y, i.e., Y = X. * E, then, using a multi-source data fusion algorithm, let the collected i-th type of data be D. i Its adaptive weight is w i The merged data is D f The calculation formula is: Where c represents the number of data samples and o represents the number of data types, thus merging different types of data into a comprehensive dataset.

[0052] Harmonic source location and tracking: The harmonic source location and tracking module receives a comprehensive dataset and uses a location algorithm based on harmonic impedance gradients to calculate the coordinates of the harmonic source. Let the harmonic impedance at a point x in the power grid be Z(x), and the harmonic source location be x0. The algorithm iteratively calculates x... k+1 To approximate the harmonic source location x0, the calculation formula is: in It is x k The gradient of harmonic impedance at x k It is the point that approximates the location of the harmonic source in the k-th iteration, x k+1 In the (k+1)th iteration, the point that approximates the location of the harmonic source is defined by the step size parameter 'a', which ranges from 0.1 to 0.5. This parameter is dynamically adjusted based on the real-time fluctuations in the harmonic impedance gradient, using the following formula: Where: α0 is the initial step size coefficient, and the calculation formula is: The harmonic impedance gradient at the initial iteration point, where η is a proportionality coefficient ranging from 0.2 to 0.8, and δ is a smoothing constant to prevent the denominator from failing when the gradient is zero. denoted as the harmonic impedance gradient at the k-th iteration, λ is the sensitivity adjustment factor, and a value of 0.5 to 1.5 is recommended. ∈ is a very small positive number to avoid a denominator of zero.

[0053] Harmonic propagation path analysis: Obtain the topology data of the power grid in the commercial area, abstract the nodes in the power grid as vertices of a graph, and the lines as edges connecting the vertices, constructing a graph structure model G = (V, E), where V is the set of nodes, E is the set of edges, and each node v i ∈V corresponds to an actual node in the power grid, and each edge e ij ∈E connects to node v i and v j Let S represent the actual power transmission line. Assign electrical parameters to each edge. Taking the power transmission line connecting the shopping mall and the nearby office building as an example, calculate the complex power flow S between them. ij The formula for calculating complex power flow is: Where V i It is the voltage phasor of node i, obtained through a voltage measuring device installed at node i. isI ij The conjugate of the complex number, I ij I is the current phasor flowing from node i to node j. According to circuit theory, I ij The voltage V through node i i Voltage V at node j j And the electrical parameters of the circuit are calculated to include resistance R.ij , inductance L ij , and capacitance C ij , the line, the current phasor I ij , the calculation formula is: Where ω is the angular frequency, ω = 2πf h , f h is the frequency of the hth harmonic, the calculation formula is: f h = h·f0, h is the harmonic number, f0 is the fundamental frequency of the power grid, according to the phase of the complex power flow S ij , the direction of the harmonic propagation is determined, when Re(S ij ) > 0, Re represents the real part, which means that the power flows from node i to node j, that is, the harmonic propagates from node i to node j along the line; when Re(S ij ) > 0, the direction of the harmonic propagation is opposite, the propagation intensity coefficient is calculated to determine the propagation intensity of the harmonic on different lines, let the propagation intensity coefficient be K ij , the calculation formula is: Where |S ij | is the modulus of the complex power flow S ij , which represents the size of the harmonic power on the line, N i is the adjacent node set of node i, , which determines the proportion of the harmonic propagation intensity on the line in the entire power grid, where Ni is the adjacent node set of node i.

[0054] Comprehensive evaluation and management decision: the comprehensive evaluation and management decision module predicts the probability of device failure caused by harmonics based on the output results of the previous modules using the long short-term memory network, extracts data related to harmonics and device status from the output of the multi-source data deep fusion module, the harmonic source positioning and tracking module and the harmonic propagation path analysis module, cleans and standardizes the data, and arranges the data in chronological order to construct time series data, constructs an LSTM model, determines the number of layers and the number of neurons in each layer, and the LSTM unit contains an input gate f t , Where x t is the current input, h t-I is the hidden state at the previous time, W ii and W hi are weight matrices, b i is a bias vector, and σ is a sigmoid activation function. The forgetting gate f t : f t = σ(W if x t +W hf h t-1 +b f), candidate memory cell where tanh is the hyperbolic tangent activation function, memory cell C t : where represents element-wise multiplication, output gate o t : o t = sigma(W i oxt + W h oh t-1 + b o ), hidden state h t : h t = o t tanh(C t ), an additional fully connected layer is added, and its output is: y = W out h last + b out , where W out is the weight matrix of the fully connected layer, h last is the hidden state of the last time step of the last LSTM layer, b out is the bias vector, and then an additional fully connected layer is added, and its output is: y = W out h last + b out , where W out is the weight matrix of the fully connected layer, h last is the hidden state of the last time step of the last LSTM layer, b out is the bias vector, the mean square error is selected as the loss function, the Adam optimizer is used for parameter update, the preprocessed data is divided into training set, validation set and test set, the total amount of data is N, the training set ratio is p1, the validation set ratio is p2, and the test set ratio is p3, and p1+p2+p3=1, then the training set number N1=p1N, the validation set number N2=p2N, and the test set number N3=p3N, the training set data is input into the LSTM model for training, in each training step, the predicted value is calculated by forward propagation, then the loss L is calculated according to the loss function, the gradient is calculated by the back propagation algorithm, and finally the model parameters are updated using the optimizer, the real-time collected multi-source data is processed according to the data preprocessing steps, and then input into the trained LSTM model, and the model outputs the predicted device failure probability P fault , calculated T1 i≤T2 evaluation results for medium risk level, in addition to the frequency of enhanced monitoring, further analysis of harmonic source, found that part of the old elevator in the market frequency control system is the main harmonic generating equipment, take measures to adjust the elevator running time, optimize the transformer tap, in the elevator distribution box installation small harmonic control device measures, reduce the impact of harmonics on the power grid.

[0055] In summary, in the intelligent power grid harmonic management of urban commercial area, through the multi-source data acquisition module to collect all kinds of data, using multi-source data deep fusion module for dimension reduction and fusion processing, to provide strong support for subsequent analysis, harmonic source positioning tracking module accurately locates the harmonic source, harmonic propagation path analysis module determines the direction and intensity of harmonic propagation, comprehensive evaluation and management decision module evaluates the risk according to the results of each module, and takes effective measures for medium risk level. This series of operations constitutes a complete harmonic management system, which can timely find and solve the harmonic problem in commercial area, ensure the stable operation of power grid, improve the power quality, and ensure the normal work of various electrical equipment.

[0056] Example two:

[0057] Industrial park intelligent power grid harmonic management.

[0058] A large number of industrial production equipment such as motors, electric welders and rectifiers in industrial parks will generate a large amount of harmonics, which will threaten the stability of the power grid.

[0059] Data acquisition: install intelligent electric meters, harmonic analyzers, voltage transformers, current transformers and synchronous phasor measurement devices on the distribution room, production workshop distribution box and main production equipment in the industrial park to collect power consumption behavior, power grid operation and harmonic impedance data. At the same time, use the meteorological sensor in the park to collect meteorological data, and through the embedded edge storage device, clean and normalize these data locally, and then transmit them to the multi-source data deep fusion module.

[0060] Data fusion and dimension reduction: the multi-source data deep fusion module constructs the covariance matrix C, the calculation formula is: And the eigenvalue λ i and the corresponding eigenvector e i are obtained by eigenvalue decomposition, and the eigenvalues are sorted according to their size. According to the size of the eigenvalue, the first m principal components with cumulative contribution rate of 85%-95% are retained, and the calculation formula of the cumulative contribution rate is: The eigenvectors corresponding to the first m principal components selected form the conversion matrix, which is multiplied by the data matrix X * and the conversion matrix E, X * is the standardized data matrix, and the dimension-reduced data matrix Y is obtained, that is, Y=X * E, different types of data are fused by using multi-source data fusion algorithm, and the i-th type of data collected is Di , whose adaptive weight is w i , the fused data is D f , and the calculation formula is: Different types of data are fused into a comprehensive data set.

[0061] Harmonic source positioning and tracking: The harmonic source positioning and tracking module receives the comprehensive data set, calculates the harmonic source position coordinates using a positioning algorithm based on harmonic impedance gradient, sets the harmonic impedance at a point x in the power grid as Z(x), and the harmonic source position as x0, and iteratively calculates x k+1 to approximate the harmonic source position x0, and the calculation formula is: It is found through calculation that a large electric welding machine in a metal processing plant in the park is the main harmonic source, and its running state is tracked in real time.

[0062] Harmonic propagation path analysis: Obtain the topological structure data of the industrial park power grid, construct a graph structure model G=(V, E), and assign electrical parameters to the edges. Take the power transmission line connecting the metal processing plant and the neighboring electronics plant as an example, and calculate the complex power flow calculation formula: According to the phase of the complex power flow, the direction of harmonic propagation is determined, and the propagation intensity coefficient K ij is calculated to determine the propagation intensity of harmonics on different lines, and the calculation formula is:

[0063] Comprehensive evaluation and management decision: The comprehensive evaluation and management decision module predicts the probability of equipment failure caused by harmonics based on the output of each module using a long short-term memory network. From the output of the multi-source data deep fusion module, the harmonic source positioning and tracking module, and the harmonic propagation path analysis module, data related to harmonics and equipment status are extracted, cleaned and standardized, and arranged in chronological order to construct time series data. The LSTM model is constructed, the number of layers and the number of neurons in each layer are determined, and the LSTM unit contains an input gate f t , where x t is the current input, h t-1 is the hidden state at the previous time, W ii and W hi are weight matrices, b i is a bias vector, and σ is a sigmoid activation function, the forgetting gate f t : f t = σ(W if x t +W hf h t-1 +b f ), the candidate memory cell Where tanh is the hyperbolic tangent activation function. Memory unit C t : Where ⊙ represents element-wise multiplication, and the output gate ot ： o t =σ(W i oxt+W h oh t-1 +b o ), hidden state h t :h t =o t ⊙tanh(C t Add a fully connected layer, and its output is: y = W out h last +b out W old It is the weight matrix of the fully connected layer, h last It is the hidden state of the last time step of the last LSTM layer, b out It is a bias vector, then a fully connected layer is added, and its output is: y = W out h last +b out W out It is the weight matrix of the fully connected layer, h last It is the hidden state of the last time step of the last LSTM layer, b out The bias vector is used, and the mean squared error is chosen as the loss function. The Adam optimizer updates the parameters and divides the preprocessed data into training, validation, and test sets. Let the total amount of data be N, the proportion of the training set be p1, the proportion of the validation set be p2, and the proportion of the test set be p3, and p1 + p2 + p3 = 1. Then the number of training sets N1 = p1N, the number of validation sets N2 = p2N, and the number of test sets N3 = p3N. The training data is input into the LSTM model for training. In each training step, the predicted value is calculated through forward propagation. Then, the loss L is calculated based on the loss function, the gradient is calculated using the backpropagation algorithm, and finally, the optimizer is used to update the model parameters. The model is then optimized using the validation set. Real-time multi-source data is processed according to the data preprocessing steps and then input into the trained LSTM model. The model outputs the predicted equipment failure probability P. fault The assessment results indicated a high-risk level, prompting immediate implementation of emergency measures. This included shutting down and rectifying the welding machines at the metal processing plant, replacing them with high-capacity, high-performance active power filters, and installing harmonic isolation devices in front of critical production equipment at the electronics plant that was severely affected by harmonics. These measures aimed to ensure the stable operation of the power grid and the normal functioning of other equipment.

[0064] In summary, the industrial park smart grid harmonic management embodiment shows the application of the present application in a complex industrial power environment, the multi-source data acquisition device collects rich data in real time, the multi-source data deep fusion module optimizes the data processing, the harmonic source positioning and tracking module accurately finds out the harmonic source, the harmonic propagation path analysis module masters the harmonic propagation situation, and the comprehensive evaluation and management decision module decides to implement emergency control measures under high risk level according to the evaluation result. The system effectively deals with the problem of serious harmonic pollution in the industrial park, reduces the harm of harmonics to the power grid and equipment, ensures the continuity and stability of industrial production, and improves the overall power supply reliability of the industrial park.

[0065] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes, without departing from the technical solution of the present application. Any modification, equivalent change and modification of the above embodiments made according to the technical essence of the present application still belong to the scope of the technical solution of the present application.

Claims

1. A multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance, characterized in that, The system includes: Multi-source data acquisition module: With the help of smart meters and sensor devices, it collects data on harmonic impedance, power grid operation, weather and user electricity consumption behavior, and cleans and normalizes the data locally through embedded edge storage devices before transmitting it to the multi-source data deep fusion module. Multi-source data deep fusion module: Receives data transmitted from the multi-source data acquisition module, uses principal component analysis to reduce the dimensionality of the data, and uses a multi-source data fusion algorithm to fuse different types of data to generate a comprehensive dataset; Harmonic source location and tracking module: Receives the comprehensive dataset output by the data preprocessing and fusion module. Based on harmonic impedance characteristics, it uses a location algorithm based on harmonic impedance gradients to calculate the coordinates of the harmonic source. Let the harmonic impedance at a point x in the power grid be Z(x), and the harmonic source location be x0. The module iteratively calculates x... k+1 To approximate the harmonic source location x0, the calculation formula is: in It is x k The gradient of harmonic impedance at x k It is the point that approximates the location of the harmonic source in the k-th iteration, x k+1 In the (k+1)th iteration, the point that approximates the location of the harmonic source is defined by α, which is a step size parameter ranging from 0.1 to 0.

5. α is dynamically adjusted based on the fluctuations in the real-time harmonic impedance gradient, using the following formula: Where: α0 is the initial step size coefficient, and the calculation formula is: The harmonic impedance gradient at the initial iteration point, where η is a proportionality coefficient ranging from 0.2 to 0.8, and δ is a smoothing constant to prevent the denominator from failing when the gradient is zero. λ is the harmonic impedance gradient at the k-th iteration, λ is the sensitivity adjustment factor, and it is recommended to take a value of 0.5 to 1.

5. ∈ is to avoid extremely small positive numbers with zero denominator, and at the same time, the dynamics of the harmonic source are monitored and tracked in real time. The harmonic propagation path analysis module acquires power grid topology data and harmonic source information provided by the harmonic source location and tracking module. It constructs a graph model of the power grid using abstract data from power grid nodes and lines. Based on node voltages and line electrical parameters, it calculates complex power currents, determines the propagation direction of harmonics based on the phase of the complex power currents, and determines the propagation intensity of harmonics on different lines by calculating the propagation intensity coefficient. Let the propagation intensity coefficient be K. ij The calculation formula is: Where |S ij | is the complex power current S ij The modulus, N, represents the magnitude of the harmonic power on the line. i It is the set of adjacent nodes of node i. This represents the sum of the complex power flow modes between node i and all its adjacent nodes; Comprehensive assessment and management decision-making module: Based on the outputs of the multi-source data deep fusion module, harmonic source location and tracking module, and harmonic propagation path analysis module, the module uses long short-term memory network to predict the probability of equipment failure caused by harmonics, and formulates harmonic management plan based on the assessment results.

2. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, The data types and acquisition devices acquired in the multi-source data acquisition module are as follows: Harmonic impedance data: Data types: amplitude and phase relationship of harmonic voltage and harmonic current in the power grid, and harmonic impedance value at a specific frequency; Data acquisition equipment: harmonic analyzer and harmonic impedance measuring device; Power grid operation data: Data types: voltage, current, power, and frequency; Data collection equipment: smart meters, voltage transformers, current transformers, and synchronous phasor measurement devices. Meteorological data: Data types: temperature, humidity, wind speed, and air pressure; Data acquisition equipment: temperature sensor, humidity sensor, anemometer, and barometer; User electricity consumption behavior data: Data types: Electricity usage time, power consumption, appliance usage patterns, and abnormal electricity usage behavior. Data acquisition equipment: smart meters, non-intrusive load monitoring devices used to identify user electrical equipment through current waveform analysis, and user-side power quality monitoring sensors.

3. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, The multi-source data deep fusion module uses principal component analysis to reduce the dimensionality of the data, receives the preprocessed data from the multi-source data acquisition module, and constructs the covariance matrix C of the data. The calculation formula is as follows: in It is the i-th feature value of the k-th sample after standardization, where n is the number of samples. It is the mean of the i-th feature after standardization. It is the j-th digit of the k-th sample after standardization. Let λ represent the mean of the j-th feature after standardization. Then, perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues ​​λ. i and the corresponding eigenvector e i i = 1, 2, ..., p, where p is the number of original features, λ i The i-th eigenvalue of the covariance matrix C represents the magnitude of the variance of the data in that direction, e i This indicates that the covariance matrix C corresponds to the eigenvalue λ. i The eigenvectors represent the direction of data change. The eigenvalues ​​are arranged in descending order. Based on the magnitude of the eigenvalues, the top m principal components with a cumulative contribution rate of 85%-95% are retained. The calculation formula is as follows: Where m is the number of principal components selected, the eigenvectors corresponding to the first m selected principal components form the transformation matrix X. * Multiplying by the transformation matrix E, X * The data matrix after standardization is used to obtain the dimensionality-reduced data matrix Y, which is expressed by the formula: Y = X * E is a transformation matrix composed of the eigenvectors corresponding to the first m principal components.

4. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, The multi-source data deep fusion module uses a multi-source data fusion algorithm to fuse different types of data. Let the collected i-th type of data be D. i Its adaptive weight is w i The merged data is D f The calculation formula is: Where c represents the number of data samples and o represents the number of data types.

5. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, The harmonic propagation path analysis module utilizes the abstraction of power grid nodes and lines to construct a graph structure model of the power grid. It acquires power grid topology data, abstracts nodes in the power grid as vertices of a graph, and lines as edges connecting those vertices, constructing a graph structure model G = (V, E) for the power grid, where V is the set of nodes, E is the set of edges, and each node v... i ∈V corresponds to an actual node in the power grid, and each edge e ij ∈E connects to node v i and v j This represents the actual power transmission line and assigns electrical parameters to each side.

6. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, In the harmonic propagation path analysis module, the determination of the harmonic propagation direction involves calculating the complex power flow S between two nodes i and j connected by any line in the power grid. ij The calculation formula is: Where V i It is the voltage phasor of node i, obtained through a voltage measuring device installed at node i. isI ij The conjugate of the complex number, I ij I is the current phasor flowing from node i to node j. ij The voltage V through node i i Voltage V at node j j And the electrical parameters of the circuit are calculated to include resistance R. ij Inductor L ij and capacitor C ij The circuit, current phasor I ij The calculation formula is: Where ω is the angular frequency, ω = 2πf h f h It is the frequency of the h-th harmonic, and the calculation formula is: f h = h·f0, where h is the harmonic order, f0 is the fundamental frequency of the power grid (50Hz), and the direction of harmonic propagation is determined by the complex power flow S. ij The phase determines when Re(S) ij If Re > 0, Re represents taking the real part, indicating that power flows from node i to node j, that is, the harmonic propagates from node i to node j along this path; when Re(S) > 0, Re represents taking the real part, indicating that power flows from node i to node j, that is, the harmonic propagates from node i to node j along this path; ij If ) > 0, the harmonic propagation direction is opposite.

7. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, The comprehensive assessment and management decision-making module utilizes a long short-term memory network to predict the probability of equipment failures caused by harmonics. It extracts data related to harmonics and equipment status from the outputs of the multi-source data deep fusion module, the harmonic source localization and tracking module, and the harmonic propagation path analysis module. The data is cleaned and standardized, and then arranged chronologically to construct a time-series dataset. An LSTM model is then constructed, determining its number of layers and the number of neurons in each layer. Each LSTM unit contains an input gate f. t , Where x t This is the current input, h t-I It is the hidden state from the previous moment, W ii and W hi It is the weight matrix, b i It is the bias vector, and σ is the sigmoid activation function. Forgotten Gate f t :f t =σ(W if x t +W hf h t-1 +b f Candidate memory units Where tanh is the hyperbolic tangent activation function. Memory unit C t : Where ⊙ represents element-wise multiplication, and the output gate is 0. t :o t =σ(W i oxt+W h oh t-1 +b o ), hidden state h t :h t =o t ⊙tanh(C t Add a fully connected layer, and its output is: y = W out h last +b out W out It is the weight matrix of the fully connected layer, h last It is the hidden state of the last time step of the last LSTM layer, b out It is a bias vector, then a fully connected layer is added, and its output is: y = W out h last +b out W out It is the weight matrix of the fully connected layer, h last It is the hidden state of the last time step of the last LSTM layer, b out The bias vector is used, and the mean squared error is chosen as the loss function. The Adam optimizer updates the parameters and divides the preprocessed data into training, validation, and test sets. Let the total amount of data be N, and the proportion of the training set be p. I The validation set ratio is p2, the test set ratio is p3, and p1 + p2 + p3 = 1. Therefore, the number of training sets is N1 = p1N, the number of validation sets is N2 = p2N, and the number of test sets is N3 = p3N. The training set data is input into the LSTM model for training. In each training step, the predicted value is calculated through forward propagation. Then, the loss L is calculated based on the loss function, the gradient is calculated using the backpropagation algorithm, and finally, the optimizer is used to update the model parameters. The model is then optimized using the validation set. Real-time multi-source data is processed according to the data preprocessing steps and then input into the trained LSTM model. The model outputs the predicted equipment failure probability P. fault When P fault When T1 ≤ P, it is judged as a low-risk level, indicating that the harmonic situation in the power grid is relatively good; when T1 < P fault When T2 is ≤, it is judged to be of medium risk level, indicating that the power grid has a certain degree of harmonic problem; when When the time is determined to be high risk level, it indicates that the power grid harmonic pollution is relatively serious. Among them, T1 and T2 are thresholds set based on industry standards.

8. The multi-source fusion smart grid harmonic dynamic management system based on harmonic impedance according to claim 1, characterized in that, The comprehensive assessment and management decision-making module formulates a harmonic management plan based on the assessment results as follows: Low-risk level: Adopt a strategy of regular monitoring to continuously track the power grid harmonic status, and at the same time, carry out preventive maintenance on equipment; Medium risk level: In addition to strengthening the monitoring frequency, further analysis of harmonic sources should be conducted to identify the main harmonic generating equipment or areas, and measures such as adjusting load distribution, optimizing transformer taps, and installing small harmonic mitigation devices should be taken. High-risk level: Immediately initiate emergency remediation measures, shut down and rectify equipment with severely excessive harmonic sources, replace with high-capacity, high-performance active power filters and static var compensators, and install harmonic isolation devices on critical equipment severely affected by harmonics.

Images