Multi-source random load reactive flow correlation feature analysis method based on zebra optimization

Abstract

The invention discloses a multi-source random load reactive flow correlation feature analysis method based on zebra optimization, and relates to the technical field of power system analysis. According to the method, active power, reactive power, voltage amplitude and phase angle of each node in a multi-source random load are collected and preprocessed, and a reactive flow correlation characteristic matrix is constructed; and utilizing a zebra optimization algorithm to optimize correlation feature analysis model parameters, training a deep neural network model containing alternately arranged full connection layers and convolutional layers in combination with the optimization parameters, and finally analyzing and outputting a result based on the model. According to the method, the accuracy and the reliability of multi-source random load reactive flow correlation characteristic analysis are improved, a visual result and a quantitative index are provided for operation of a power system, optimization of a scheduling decision is facilitated, and the stability and the economical efficiency of the system are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system analysis, in particular to a multi-source random load reactive power flow correlation feature analysis method based on zebra optimization. BACKGROUND

[0002] In modern power systems, the existence of multi-source random load makes the reactive power flow present complex and variable characteristics. With the large number of distributed power sources and the diversification of power load types, the traditional reactive power flow analysis method has been difficult to meet the requirements of accuracy and timeliness. In the prior art, some methods rely on simplified mathematical models and cannot fully reflect the influence of load randomness and multi-source; some methods introduce intelligent optimization algorithms, but there are deficiencies in the pertinence of parameter optimization and the feature extraction ability of the model, which limits the reliability and practicality of the analysis results. Therefore, there is an urgent need for a method that can effectively fuse multi-source load data, accurately optimize model parameters and deeply mine reactive power flow correlation features to improve the operation stability and economy of the power system. SUMMARY

[0003] The main purpose of the present application is to provide a multi-source random load reactive power flow correlation feature analysis method based on zebra optimization, which aims to solve the problems of poor adaptability of existing reactive power control strategies, difficulty in balancing voltage stability and economy.

[0004] To achieve the above-mentioned purpose, the present application provides the following solutions:

[0005] The present application provides a multi-source random load reactive power flow correlation feature analysis method based on zebra optimization, comprising the following steps:

[0006] Collecting real-time active power, reactive power, voltage amplitude and voltage phase angle of each load node in the multi-source random load and preprocessing;

[0007] Based on the preprocessed active power, reactive power, voltage amplitude and voltage phase angle, a reactive power flow correlation feature matrix is constructed;

[0008] The parameters of the correlation feature analysis model are optimized using the zebra optimization algorithm;

[0009] The reactive power flow correlation feature analysis model is constructed and trained combined with the optimized parameters;

[0010] Based on the trained model, the reactive power flow correlation features of the multi-source random load are analyzed and the results are output.

[0011] Optionally, the collecting real-time active power, reactive power, voltage amplitude and voltage phase angle of each load node in the multi-source random load and preprocessing comprises:

[0012] The active power, the reactive power, the voltage amplitude, and the voltage phase angle of each load node are collected in real time by the power monitoring device;

[0013] The collected active power, reactive power, voltage amplitude, and voltage phase angle are standardized respectively;

[0014] The standardized data are denoised by wavelet transform, and the data with missing values are interpolated and filled by K-Nearest Neighbor algorithm.

[0015] Optionally, the reactive flow correlation feature matrix is constructed based on the preprocessed active power, reactive power, voltage amplitude, and voltage phase angle, comprising:

[0016] A n*n order reactive flow correlation feature matrix M is constructed based on the preprocessed active power, reactive power, voltage amplitude, and voltage phase angle of each load node;

[0017] The element M ij in the ith row and jth column of the matrix M is used to represent the reactive flow correlation between the ith load node and the jth load node, and the calculation formula is:

[0018]

[0019] Wherein, P i and P j are the preprocessed active power of the ith and jth load nodes respectively;

[0020] Q i and Q j are the preprocessed reactive power of the ith and jth load nodes respectively;

[0021] V i and V j are the preprocessed voltage amplitude of the ith and jth load nodes respectively;

[0022] θ i and θ j are the preprocessed voltage phase angle of the ith and jth load nodes respectively.

[0023] Optionally, the parameters of the correlation feature analysis model are optimized by using Zebra Optimization algorithm, comprising:

[0024] The optimization range of the model parameters is determined, and an initial Zebra population is generated;

[0025] The fitness function is constructed according to the target of correlation feature analysis, and the optimization effect of individuals in the population is evaluated;

[0026] An optimal parameter is found through a population iterative updating mechanism until a preset optimization termination condition is met.

[0027] Optionally, the optimization range of the model parameter is determined, and an initial zebra population is generated, specifically:

[0028] According to the physical meaning and value constraint of the parameters in the correlation characteristic analysis model, the optimization interval of each parameter is determined.

[0029] N zebra individuals are randomly generated in the interval, and the position vector of each individual corresponds to a set of model parameters. The parameter dimension is related to the dimension of the correlation characteristic matrix.

[0030] Optionally, the fitness function is constructed according to the correlation characteristic analysis target, and the optimization effect of the individual in the population is evaluated, specifically:

[0031] The fitness function is constructed by minimizing the deviation between the actual value of the correlation characteristic matrix and the predicted value of the model based on the current parameters.

[0032] The optimization effect of the parameter combination represented by each zebra individual is evaluated by calculating the fitness value corresponding to each zebra individual. The smaller the fitness value, the better the optimization effect.

[0033] Optionally, the optimal parameter is found through a population iterative updating mechanism until a preset optimization termination condition is met, specifically:

[0034] During the population iteration process, the individual updates the position based on the characteristic information of the collected data through exploration, utilization and information interaction strategy;

[0035] The preset optimization termination condition includes the maximum number of iterations and the convergence threshold of the fitness value. When any condition is met, the iteration is stopped and the parameters corresponding to the current optimal individual are output.

[0036] Optionally, the reactive power flow correlation characteristic analysis model is constructed and trained combined with the optimized parameters, including:

[0037] A network model suitable for multi-source random load reactive power flow correlation characteristic analysis is constructed, which includes multiple hidden layers for feature extraction and processing;

[0038] The optimal parameters obtained by the zebra optimization algorithm are configured into the model;

[0039] The model is trained using preprocessed correlation characteristic data to improve the analysis accuracy of the model.

[0040] Optionally, the hidden layer adopts an alternating structure of fully connected layers and convolutional layers.

[0041] The full connection layer is used for global feature mapping of the input associated feature matrix, and the convolution layer is used for extracting local key features in the associated feature matrix, and deep mining of complex associated features of reactive power flow is realized through multi-layer alternating processing.

[0042] Optionally, the trained model is used to analyze the multi-source random load reactive power flow associated features and output results, including:

[0043] The trained model is used to analyze the flow relationship of the reactive power between nodes, and the associated features of the voltage parameters and the reactive power flow;

[0044] Output the reactive power flow chart, the voltage-reactive power relationship curve and the quantitative index, and the quantitative index includes the reactive power transmission coefficient between nodes and the voltage stability index.

[0045] Through the above technical solutions, the beneficial effects of the present application are: through the fine pretreatment of the multi-source random load data, the feature matrix reflecting the reactive power flow association between nodes is constructed, the model parameters are accurately optimized by combining the zebra optimization algorithm, and the feature rules are efficiently mined by using the deep neural network, not only the accuracy and timeliness of the reactive power flow analysis are improved, but also the intuitive visual results and quantitative indexes are output, which provides scientific decision support for power system dispatching, and effectively enhances the stability and economy of system operation. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description only some embodiments of the present application, and the embodiments in the drawings do not constitute any limitation to the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0047] Figure 1 The flowchart of the multi-source random load reactive power flow associated feature analysis method based on zebra optimization provided by an embodiment of the present application is shown in the figure.

[0048] Figure 2 The structure diagram of the reactive power flow associated feature analysis model provided by an embodiment of the present application is shown in the figure.

[0049] Figure 3 The schematic diagram of the reactive power flow direction and intensity distribution between nodes of the multi-source random load provided by an embodiment of the present application is shown in the figure.

[0050] Figure 4 The dynamic relationship curve diagram of the voltage per unit value and the reactive power associated feature value of each load node provided by an embodiment of the present application is shown in the figure.

[0051] Figure 5 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application;

[0052] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0053] It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the application. Rather, these embodiments are provided to make the disclosure more thorough and complete, and to fully convey the scope of the disclosure to those skilled in the art.

[0054] The foregoing and other technical contents, features and effects of the present invention are described in conjunction with the appendix below. Figures 1-5 The detailed description of the embodiments will make this clear. All structural details mentioned in the following embodiments are based on the accompanying drawings.

[0055] Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings.

[0056] In one exemplary embodiment, such as Figure 1 As shown, a method for analyzing the correlation characteristics of multi-source stochastic load reactive power flow based on zebra optimization is provided. This method is executed by a control terminal with computing capabilities and specifically includes the following steps:

[0057] Step S1: Collect the real-time active power, reactive power, voltage amplitude, and voltage phase angle of each load node in the multi-source random load, and perform preprocessing.

[0058] In one specific embodiment, firstly, the control terminal establishes a communication connection with the monitoring equipment of each load node in the power system, and collects the active power P of each load node in real time. i Reactive power Q i Voltage amplitude V i Voltage phase angle θ i (where 9i = 1, 2, ..., n, and n is the number of load nodes). The data collection frequency is set to 3 minutes / time to ensure data timeliness while avoiding data redundancy. The collected data is classified and stored according to node number and collection time to form the original data sequence.

[0059] After data collection, the data standardization process begins. The core of this step is to eliminate the interference caused by differences in the magnitude of different parameters in subsequent analysis, ensuring that all parameters are within the same comparable order of magnitude. For active power P... i Calculate its overall mean μ during the collection period. P and standard deviation σ P :

[0060]

[0061] Wherein, T is the total number of collection.

[0062] The standardization formula is By dividing by 3 times the standard deviation, most of the data can be distributed in the [-1, 1] interval.

[0063] Similarly, for the reactive power Q i , the mean μ Q and the standard deviation σ Q are calculated, and the standardization formula is

[0064] For the voltage amplitude V i , the mean μ V and the standard deviation σ V are calculated, and the standardization formula is

[0065] For the voltage phase angle θ i , the mean μ θ and the standard deviation σ θ are calculated, and the standardization formula is

[0066] The data after standardization needs to be denoised, and the improved wavelet packet transform method is used to remove the interference noise. Specifically, the sym5 wavelet packet is selected to decompose the normalized active power sequence P i ′ for 4 layers, and the wavelet packet coefficients of each frequency band are obtained. For the coefficients obtained by decomposition, a threshold processing method based on signal-to-noise ratio is used, and the threshold is Wherein, SNR is a preset signal-to-noise ratio, and c k is the wavelet packet coefficient.

[0067] The coefficients less than λ in absolute value are set to zero, and the coefficients greater than or equal to λ are retained, and then the denoised active power sequence P is reconstructed by wavelet packet inverse transform. The same method is used to denoise Q i ′ , V i ′ , and θ i ′ , respectively, to obtain to ensure the purity of the data.

[0068] Finally, the missing data is filled, and the K nearest neighbor algorithm based on weighted Euclidean distance is used for the missing data in the data collection process. Specifically, for the identified missing data point, such as P i (t0), 6 adjacent complete data points P i are selected in its historical data sequence.(t1), P i (t2), …, P i (t6), calculating the weighted Euclidean distance between the missing point and each adjacent point wherein, a is a weight coefficient, is the local mean value of the active power of the node.

[0069] Based on the weighted Euclidean distance d k Calculate the weight w k The formula is:

[0070]

[0071] The filling formula is:

[0072]

[0073] For other types of missing data Q i (t0), V i (t0), θ i (t0), the same method is used to fill in the missing values, and finally the complete and pure preprocessed data is obtained, which provides reliable data support for subsequent construction of correlation feature matrix and model training.

[0074] Step S2, based on the preprocessed active power, reactive power, voltage amplitude and voltage phase angle, a reactive flow correlation feature matrix is constructed.

[0075] In a specific embodiment, first, the control terminal extracts each load node data after standardization, denoising and missing value filling from the data storage module, including and an index table is established according to the node number to ensure the accurate correspondence of each node parameter when calling data.

[0076] Next, the control terminal initializes an n×n order reactive flow correlation feature matrix M as a carrier for storing correlation feature values. For each element M ij , used to represent the reactive flow correlation between the i th load node and the j th load node, the calculation formula is:

[0077]

[0078] Wherein, P i and P j are the preprocessed active power of the i th and j th load nodes, respectively;

[0079] Q i and Q j are the preprocessed reactive power of the i th and j th load nodes, respectively;

[0080] Vi and V j These are the preprocessed voltage amplitudes of the i-th and j-th load nodes, respectively;

[0081] θ i and θ j These are the voltage phase angles of the i-th and j-th load nodes after preprocessing, respectively.

[0082] In this formula, the first term, by dividing the sum of the product of the active power and the product of the reactive power of the two nodes by the product of the voltage amplitudes of the two nodes, intuitively reflects the basic relationship between nodes under the interaction of power and voltage. The second term uses the cosine function to calculate the trigonometric function value of the voltage phase angle difference between the two nodes, accurately capturing the influence of the phase difference on the direction and intensity of reactive flow. The combination of the two terms comprehensively describes the reactive flow relationship between nodes.

[0083] During the calculation process, the control terminal uses a parallel computing method to process all elements M. ij In synchronous calculations, when i = j, the results reflect the inherent correlation between the reactive power, voltage, and phase of that node; when i ≠ j, the results reflect the interaction strength between different nodes during the reactive power flow process. Since M... ij With M ji Because it is physically symmetric, after all elements are calculated, the control terminal performs a symmetry check on matrix M. If the check passes, the matrix is ​​stored in the feature database, providing structured input data for subsequent Zebra Optimization Algorithm to optimize model parameters and for deep neural network model training.

[0084] By constructing the reactive power flow correlation feature matrix as described above, the complex and abstract reactive power flow correlations among multi-source random load nodes are transformed into a quantifiable and structured numerical matrix. This transformation not only makes the dynamic characteristics of the power system, which were originally difficult to analyze directly, measurable and calculable, but its structured form can also accurately adapt to the input requirements of subsequent models, helping deep neural networks to efficiently extract global and local correlation features. At the same time, the correlation strength information between nodes contained in the matrix provides a clear optimization guide for the Zebra optimization algorithm, ensuring that the algorithm parameters are optimized in accordance with the actual laws of reactive power flow, thereby improving the accuracy and reliability of the analysis method as a whole.

[0085] Step S3: Optimize the parameters of the correlation feature analysis model using the Zebra Optimization Algorithm.

[0086] In one specific embodiment, firstly, the control terminal determines the optimization range of each parameter based on the structure of the correlation feature analysis model and the physical constraints of reactive power flow in the power system. For example, for the weight parameters of the neural network, their value range is set to [-w max ,w max ], where wmax is the maximum weight value set empirically. And a d-dimensional parameter search space is constructed according to the total number of parameters d, which is related to the dimension of the associated feature matrix, d = n 2 .

[0087] Subsequently, the control terminal randomly generates N zebra individuals in the space, and the position vector Z k = [z k1 ,z k2 ,…,z kd ](k = 1, 2, …, N) of each individual corresponds to a complete set of model parameters, thereby forming an initial zebra population.

[0088] Next, the control terminal constructs a fitness function according to the target of the associated feature analysis to quantitatively evaluate the optimization effect of each individual in the population. The fitness function takes the deviation between the actual value M of the associated feature matrix and the predicted value of the model based on the current parameters as the core, and is defined as:

[0089]

[0090] where ω ij is the weight coefficient. Higher weight is given to the associated features between key nodes, and satisfies By calculating the fitness value of each zebra individual, the control terminal can select the optimal individual Z best in the current population, i.e., the individual with the smallest fitness value.

[0091] In the population iterative update phase, the control terminal starts the iteration process according to the preset maximum iteration number T max . In each iteration, the individual updates its position based on the feature information of the collected data through exploration, utilization and information interaction strategies. The specific steps are as follows:

[0092] The exploration strategy aims to expand the search range and avoid the algorithm falling into local optimum. The control terminal first generates a random probability value p for each individual. When p ≤ p exp , the individual executes the exploration strategy, where p exp is the preset exploration probability, usually taking 0.3. At this time, the control terminal calculates the comprehensive feature mean of the collected data, and the calculation formula is:

[0093]

[0094] The individual position update formula is:

[0095] where δ is the exploration step size, and rand() is a random number in the interval [0, 1]. By introducing The search direction is associated with the load data characteristics, and the search is more targeted.

[0096] The utilization strategy focuses on approaching the current optimal solution, improving the local search accuracy. The control terminal first determines the optimal individual in the current population and calculates the mean of the normalized voltage amplitude For non-optimal individuals, the position update formula is Where β is the utilization coefficient, The introduction of can adjust the amplitude of approaching the optimal solution according to the voltage stability characteristics. When the voltage fluctuation is small, it is more stable, and the convergence speed of the utilization strategy is accelerated.

[0097] The information exchange strategy improves the group optimization efficiency through cooperation between individuals. The control terminal defines a neighbor set N k for each individual k, usually selecting the top 20% of individuals in terms of fitness value as neighbors, and then calculating the mean of the normalized voltage phase angle The individual position update formula is Where γ is the interaction coefficient, and |N k | is the number of neighbors. The introduction of makes the information exchange intensity match the phase angle correlation characteristics. When the phase difference between nodes is small, it is more stable, and the cooperation between individuals is closer, accelerating the propagation of high-quality parameter information.

[0098] In each iteration, the control terminal executes the three strategies in the above order, and each individual in the same iteration only executes one strategy, which is allocated by probability. Then, the fitness values of all individuals are recalculated, and the optimal individual The iteration process is repeated until the maximum iteration number T max or the optimal individual fitness value does not change significantly for T conv consecutive iterations, ensuring that the final obtained parameters have both global optimality and adaptability to actual load characteristics.

[0099] Step S4, build a reactive power flow correlation feature analysis model and train it with the optimized parameters. The goal is to enable the model to accurately capture the complex correlation characteristics of multi-source random load reactive power flow.

[0100] In a specific embodiment, first, the control terminal calls the model construction module to build a deep neural network model suitable for the analysis scenario. The input layer dimension of the model is consistent with the dimension of the correlation feature matrix M, which is used to receive the preprocessed correlation feature matrix data.

[0101] Specifically, as Figure 2As shown, the model contains 4 hidden layers, adopts an alternating structure of fully connected layers and convolutional layers: the first hidden layer is a fully connected layer, containing 2n 2 neurons, and the activation function adopts LeakyReLU (the slope parameter is set to 0.01), which realizes global feature mapping of the input matrix through the formula H1 = LeakyReLU (W1 M + b1), where W1 is the weight matrix of the layer, and b1 is the bias vector; the second hidden layer is a convolutional layer, using 32 3x3 convolutional kernels, with a step size of 1, which extracts local key features through the formula , where represents convolution operation. The third and fourth hidden layers respectively repeat the structure of fully connected layers and convolutional layers, and the number of neurons and the convolutional kernel parameters decrease in proportion to realize deep compression and abstraction of features. The output layer of the model is a fully connected layer, with the same output dimension as the input layer, which is used to output the predicted correlation feature matrix

[0102] After the model structure is built, the control terminal reads the optimal parameter combination Z best obtained by the Zebra optimization algorithm from the parameter storage module, and configures it into the parameters of each layer of the model according to the corresponding dimensions. Subsequently, the control terminal retrieves the preprocessed correlation feature matrix data from the feature database, divides it into a training set (used for model parameter iterative update) and a validation set (used for monitoring model generalization ability) according to a ratio of 8:2, and sets the training parameters: batch size is set to 32, initial learning rate η = 0.002, and maximum training rounds E max = 100.

[0103] During training, the control terminal uses a hybrid loss function to guide parameter adjustment, and the loss function is defined as:

[0104]

[0105] , where is the mean square error;

[0106] is the mean absolute error;

[0107] λ1 = 0.7, λ2 = 0.3 are weight coefficients.

[0108] The control terminal calculates the gradient of the loss function with respect to each layer parameter through the back propagation algorithm, updates the parameters according to the formula , and calculates the validation loss of the model using the validation set after each training round. When the validation loss does not decrease for 10 consecutive rounds, the control terminal stops training and saves the current model parameters to obtain the final trained correlation feature analysis model.

[0109] ​By the above network structure design with full connection layer and convolution layer alternately, the advantages of the two levels can be fully given. The full connection layer performs global feature mapping on the input associated feature matrix, which can capture global dependencies such as power transmission across nodes. The convolution layer focuses on extracting local key features in the matrix, such as the reactive power interaction characteristics between adjacent nodes, realizing all-round learning of complex association patterns. The optimal parameters obtained by zebra optimization are configured into the model, effectively avoiding the problems of low training efficiency or falling into local optimum caused by random initialization of parameters, greatly speeding up the model convergence speed. The training process based on the mixed loss function takes into account both the square term and the absolute value term of the prediction error, suppressing abnormal value interference while improving overall prediction accuracy, ultimately enabling the model to accurately reproduce the correlation characteristics of multi-source random load reactive power flow, providing a reliable tool for subsequent feature analysis and power system decision support.

[0110] Step S5, based on the trained model, the multi-source random load reactive power flow correlation characteristics are analyzed and the results are output, which is the key link of converting the analysis ability of the model into actual decision support. The reactive power flow law is intuitively presented through the combination of quantification and visualization.

[0111] In a specific embodiment, first, the control terminal loads the trained and optimized correlation characteristic analysis model, retrieves the latest collected and preprocessed multi-source load data from the real-time database, and generates a real-time correlation characteristic matrix M real as input data of the model according to the method of constructing the correlation characteristic matrix in step S2.

[0112] In the model running stage, the control terminal inputs M real into the deep neural network and calculates the predicted correlation characteristic matrix through forward propagation. Based on , two layers of feature analysis are carried out: first, the reactive power flow relationship analysis between nodes is carried out, and the numerical value and sign of the matrix element are extracted. A positive value indicates positive flow from node i to node j, and a negative value indicates reverse flow. Combined with the power system topology, a dynamic reactive power flow map is generated to intuitively show the reactive power transmission path and strength between nodes. Second, voltage parameter correlation analysis is carried out by comparing the change trend of with the corresponding node , identifying the key threshold interval, for example, when per unit value, the fluctuation amplitude exceeds 50%, it can be determined that the voltage interval is a reactive power flow sensitive area.

[0113] In the quantitative index calculation link, the control terminal calculates two types of core indexes: the reactive power transmission coefficient κ ij between nodes, and the calculation formula is:

[0114]

[0115] in, For matrix The maximum value is used to standardize the coefficient to the [0, 1] interval. The larger the value, the stronger the impact of the voltage change of node i on the reactive power of node j.

[0116] Voltage stability index ζ i The calculation formula is:

[0117]

[0118] Where, ζ i The value range is [0, 1]. The closer it is to 1, the less the voltage of the node is affected by the reactive power fluctuations of other nodes, and the better the stability.

[0119] In the results output phase, the control terminal generates three types of results through the visualization module, such as... Figure 3 and 4 As shown: The dynamic reactive power flow diagram uses arrow thickness to represent transmission intensity, and red and blue to distinguish the flow direction; the voltage-reactive power correlation curve uses the horizontal axis as the horizontal axis. per-unit value, with the vertical axis representing the per-unit value. Eigenvalues ​​are marked with different colors for curves at different nodes; the quantification index table includes the κ value for each node. ij Ranking and ζ i The specific values ​​are shown in the table below:

[0120] Table 1. Quantitative Indicators of Reactive Power Flow Correlation Characteristics of Multi-Source Random Loads

[0121] Node Voltage stability index ζ i ]]> Maximum reactive power transfer coefficient Corresponding node Node 1 0.5652 0.9265 Node 1 Node 2 0.5611 0.9105 Node 1 Node 3 0.8447 0.9927 Node 2 Node 4 0.7433 0.8416 Node 1 Node 5 0.7660 0.8966 Node 3 Node 6 0.5482 0.9830 Node 8 Node 7 0.9368 0.9747 Node 7 Node 8 0.5993 0.9358 Node 6

[0122] These results are displayed in real time through the terminal interface and can be exported as PDF reports or Excel spreadsheets, and simultaneously stored in the historical database for subsequent traceability and analysis.

[0123] To sum up, the technical scheme of the present application aims at the problems in the prior art that the traditional method is difficult to adapt to the characteristics of multi-source random load, the parameter optimization of intelligent optimization algorithm lacks pertinence, and the model feature extraction capability is limited, etc., by collecting active power, reactive power, voltage amplitude and phase angle parameters and preprocessing, the integrity and consistency of the data are ensured, laying a high-quality data foundation for subsequent analysis; the reactive flow correlation feature matrix constructed quantifies the complex correlation between nodes, realizing the structured expression of abstract relationship; when the zebra optimization algorithm is used to optimize the model parameters, the exploration, utilization and information interaction strategies are designed by combining the collected data characteristics, the global optimality and efficiency of parameter optimization are improved, and local optimization is avoided; combined with the deep neural network model with alternating full connection layer and convolution layer, global and local correlation features can be captured at the same time, and the model's extraction and analysis capability for complex reactive flow characteristics is significantly enhanced by training with a hybrid loss function. The overall scheme realizes effective fusion of multi-source load data, accurate optimization of model parameters and deep mining of correlation features, greatly improving the accuracy, timeliness and reliability of power system reactive flow analysis, and providing stronger decision support for system stable operation and economic dispatch.

[0124] In an exemplary embodiment, a computer device, which can be a server or a terminal, is provided, and an internal structure diagram of the computer device can be as shown in Figure 5 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store multi-source random load reactive flow correlation feature data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement a multi-source random load reactive flow correlation feature analysis method based on zebra optimization.

[0125] Those skilled in the art can understand that Figure 5 The structure shown in the above

[0126] In an example embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0127] In an example embodiment, a computer readable storage medium is provided, storing a computer program, the computer program implementing the steps in the above method embodiments when executed by a processor.

[0128] In an example embodiment, a computer program product is provided, including a computer program, the computer program implementing the steps in the above method embodiments when executed by a processor.

[0129] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0130] It can be understood by those skilled in the art that all or part of the processes in the above method embodiments can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0131] The database involved in each of the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in each of the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like, without being limited thereto.

[0132] The technical features of the above embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above embodiments are not described, but it should be considered that any combination of the technical features is within the scope of the present disclosure, as long as there is no contradiction.

[0133] The principles and implementation manners of the present application are described by using specific examples herein, and the above embodiments are only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In summary, the content of the present description should not be understood as a limitation of the present application.

Claims

1. A multi-source random load reactive power flow correlation feature analysis method based on zebra optimization, characterized in that, The method comprises the following steps: Collecting real-time active power, reactive power, voltage amplitude and voltage phase angle of each load node in the multi-source random load and preprocessing; Based on the preprocessed active power, reactive power, voltage amplitude and voltage phase angle, a reactive power flow correlation feature matrix is constructed; The parameters of the correlation feature analysis model are optimized by using the zebra optimization algorithm; The reactive power flow correlation feature analysis model is constructed and trained combined with the optimized parameters; Based on the trained model, the reactive power flow correlation feature of the multi-source random load is analyzed and the result is output.

2. The method of claim 1, wherein, The collection of real-time active power, reactive power, voltage amplitude and voltage phase angle of each load node in the multi-source random load and preprocessing comprises: Real-time collection of active power, reactive power, voltage amplitude and voltage phase angle of each load node by power monitoring equipment; The collected active power, reactive power, voltage amplitude and voltage phase angle are respectively standardized; The standardized data are denoised by wavelet transform, and the data with missing values are interpolated and filled by K nearest neighbor algorithm.

3. The method of claim 2, wherein, The reactive power flow correlation feature matrix is constructed based on the preprocessed active power, reactive power, voltage amplitude and voltage phase angle, comprising: Based on the preprocessed active power, reactive power, voltage amplitude and voltage phase angle of each load node, an n*n order reactive power flow correlation feature matrix M is constructed; The element in the i-th row and j-th column of the matrix M is denoted as M ij The reactive flow correlation characteristic between the i-th load node and the j-th load node is represented by the following formula: where P i and P j are the pre-processed active power of the ith and jth load nodes, respectively. Q i and Q j Qi and Qj are the pre-processed reactive power of the i-th and j-th load nodes, respectively. V i and V j are the pre-processed voltage magnitudes of the i-th and j-th load nodes, respectively. θ i and θ j are the pre-processed voltage phase angles of the i-th and j-th load nodes, respectively.

4. The method of claim 3, wherein, The parameters of the correlation feature analysis model are optimized by using the zebra optimization algorithm, comprising: Determine the optimization range of model parameters and generate an initial zebra population; According to the target of correlation feature analysis, a fitness function is constructed to evaluate the optimization effect of individuals in the population; The optimal parameters are found by population iterative updating mechanism until the preset optimization termination condition is met.

5. The method of claim 4, wherein, The determination of the optimization range of model parameters and the generation of initial zebra population are as follows: According to the physical meaning and value constraint of the parameters in the correlation feature analysis model, the optimization interval of each parameter is determined; Randomly generate N zebra individuals in the interval, and the position vector of each individual corresponds to a set of model parameters. The parameter dimension is related to the dimension of the correlation feature matrix.

6. The method of claim 5, wherein, According to the target of correlation feature analysis, a fitness function is constructed to evaluate the optimization effect of individuals in the population, comprising: Minimizing the deviation between the actual value of the correlation feature matrix and the predicted value based on the current parameters of the model to construct the fitness function; By calculating the fitness value of each zebra individual, the optimization effect of the parameter combination represented by the individual is evaluated. The smaller the fitness value is, the better the optimization effect is.

7. The method of claim 6, wherein, The optimal parameters are found by population iterative updating mechanism until the preset optimization termination condition is met, comprising: During the population iteration process, the individuals update their positions based on the characteristic information of the collected data through exploration, utilization and information interaction strategy; The preset optimization termination conditions include the maximum number of iterations and the convergence threshold of the fitness value. When any of the conditions is met, the iteration is stopped and the parameters corresponding to the current optimal individual are output.

8. The method of claim 7, wherein, The construction of the reactive power flow correlation feature analysis model and the training combined with the optimized parameters comprise: A network model suitable for analyzing the correlation characteristics of reactive power flow of multi-source random load is constructed, the model comprising multiple hidden layers for feature extraction and processing; Optimal parameters obtained by the zebra optimization algorithm are configured to the model; The model is trained using the preprocessed correlation characteristic data to improve the analysis accuracy of the model.

9. The method of claim 8, wherein, The hidden layers are alternately arranged in a structure of fully connected layers and convolutional layers; The fully connected layers are used for global feature mapping of the input correlation characteristic matrix, and the convolutional layers are used for extracting local key features in the correlation characteristic matrix, so as to realize deep mining of the complex correlation characteristics of the reactive power flow through multiple layers of alternating processing.

10. The method of claim 9, wherein, The trained model is used to analyze the correlation characteristics of the reactive power flow of the multi-source random load and output results, including: The trained model is used to analyze the flow relationship of the reactive power among nodes, the correlation characteristics of the voltage parameters and the reactive power flow; A reactive power flow diagram, a voltage-reactive power relationship curve and quantitative indexes are output, the quantitative indexes including a reactive power transmission coefficient among nodes and a voltage stability index.

Images