Fault location optimization method and system for relay protection of new energy power grid

Abstract

The invention discloses a fault positioning optimization method and system for relay protection of a new energy power grid, and belongs to the technical field of fault positioning optimization. The system comprises a data acquisition module, an initial model generation module, a feature set optimization module, a hyper-parameter optimization module, a model optimization module and a protection control module. The particle swarm optimization algorithm is utilized to perform hyper-parameter global search and optimize the penalty coefficient and the kernel parameter, then local optimization is performed through the genetic algorithm, the penalty coefficient and the kernel parameter of the SVM fault judgment model are cooperatively optimized, the particle swarm global search and the local optimization capability of the genetic algorithm are fused, so that the hyper-parameter of the support vector machine is efficiently adapted, and the fault judgment accuracy is improved. The nonlinear fault classification robustness is remarkably improved, and the problems that in the prior art, the model parameter optimization efficiency is low, and local optimum is prone to occurring are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault location optimization, and in particular to a fault location optimization method and system for relay protection of a new energy power grid. Background Art

[0002] With the rapid increase in the penetration of renewable energy sources such as wind power and photovoltaics in the power system, the characteristics of grid faults have changed significantly. The high volatility and intermittency of renewable energy generation lead to complex operating conditions such as abnormal frequency fluctuations and transient power imbalances. This leads to problems such as limited short-circuit current amplitude, non-power frequency waveforms, and phase control.

[0003] Traditional relay protection relies on the principle of determining the phase difference direction of industrial frequency electrical quantities, and its adaptability fails in such scenarios: insufficient time domain feature capture capability leads to missed faults, frequency domain harmonic interference causes an increase in the misjudgment rate, and fixed threshold mechanisms are difficult to respond to dynamic topology changes.

[0004] While existing improvement solutions incorporate signal processing and machine learning technologies, they still face three bottlenecks: single-dimensional feature extraction fails to account for both transient and steady-state characteristics; inefficient model parameter optimization limits classification accuracy; and static discrimination rules are mismatched with the real-time state of the power grid. These issues degrade the accuracy of protection system operation in scenarios with a high proportion of renewable energy, seriously threatening the reliability of the power grid's security defenses. Summary of the Invention

[0005] The embodiments of the present application provide a fault location optimization method and system for new energy power grid relay protection, which solves the problems of low efficiency and susceptibility to local optimum in model parameter optimization in the prior art, and significantly improves the robustness of nonlinear fault classification.

[0006] The embodiment of the present application provides a fault location optimization method for relay protection of a new energy power grid, comprising: S1, collecting current signals and voltage signals of the new energy power grid in real time through a power grid monitoring device and performing signal denoising; S2, performs dual feature extraction in the time domain and frequency domain on the denoised current and voltage signals, generates time domain feature vectors and frequency domain feature vectors, and constructs an SVM fault discrimination model to obtain the fault location direction discrimination results of the new energy grid relay protection; S3, weighted fusion of time domain feature vectors, frequency domain feature vectors and fault location direction discrimination results to form multi-dimensional fusion features, and dimensionality reduction processing is performed on the multi-dimensional fusion features to generate an optimized feature set; S4: Build a support vector machine discrimination model based on the radial basis kernel function, use the particle swarm optimization algorithm to perform a global search for hyperparameters, optimize the penalty coefficient and kernel parameters, and then use the genetic algorithm for local tuning to collaboratively optimize the penalty coefficient and kernel parameters of the SVM fault discrimination model; S5: Input the optimized feature set into the trained SVM fault discrimination model, combine the three-level dynamic discrimination rules of time domain amplitude threshold, frequency domain frequency offset threshold and fusion feature threshold, and comprehensively output the fault direction judgment result; S6, generating a binary control signal according to the fault direction, and performing fault section isolation according to the binary control signal.

[0007] Furthermore, the time domain feature vector extraction specifically includes: Step 2.1, obtain the mother wavelet, the sliding window of the preset scale factor and translation factor; Step 2.2: Obtain the transient fluctuation characteristics of the current signal by continuous wavelet transform of the current signal, and record the amplitude of the current mutation point as the current time domain characteristic; Step 2.3, synchronously obtain the transient fluctuation characteristics of the voltage signal through wavelet transform of the voltage signal, and record the extreme value of the voltage fluctuation as the voltage time domain characteristic; Step 2.4: Construct a time domain feature vector based on the current time domain features and the voltage time domain features.

[0008] Furthermore, the continuous wavelet transform formula of the current signal is: ; in, is the transient fluctuation characteristic of the current signal, is the scale factor, is the translation factor, For my mother Xiaobo, is the current signal; The continuous wavelet transform formula of voltage signal is: ; in, is the transient fluctuation characteristic of the voltage signal, is a voltage signal.

[0009] Furthermore, the frequency domain feature vector extraction specifically includes: Step 3.1, performing fast Fourier transform on the current signal and the voltage signal to obtain the current spectrum and the voltage spectrum; Step 3.2: extract the actual value of the current fundamental frequency offset and the actual value of the voltage fundamental frequency offset, obtain the rated power frequency current and the rated power frequency voltage, and obtain the absolute current offset and the absolute voltage offset; Step 3.3, extract the amplitude of the main harmonic components of the current from the current spectrum and extract the amplitude of the main harmonic components of the voltage from the voltage spectrum; In step 3.4, a frequency domain feature vector is constructed based on the absolute current offset, the absolute voltage offset, the amplitude of the main harmonic component of the current, and the amplitude of the main harmonic component of the voltage.

[0010] Furthermore, the penalty coefficient and kernel parameters of the collaboratively optimized SVM fault discrimination model specifically include: Step 4.1, define the radial basis kernel function, initialize the penalty coefficient and kernel parameters; Step 4.2: Use particle swarm optimization to roughly adjust the parameters, initialize the particle swarm position, evaluate the fitness and calculate the cross-validation error of SVM; In step 4.3, the genetic algorithm is used to fine-tune the parameters: the optimal solution output by the particle swarm optimization is binary-encoded, a single-point crossover is performed according to the crossover probability, and the gene position is randomly flipped according to the mutation probability.

[0011] Furthermore, the signal denoising step includes: Apply wavelet threshold denoising to the current signal and voltage signal respectively: select Symlet wavelet basis function, perform 3-layer wavelet decomposition, and obtain approximate coefficients and the detail coefficient of the jth layer , j=1,2,3 represents the number of decomposition layers, and performs soft threshold processing to obtain the detail coefficient after threshold processing , after threshold processing Approximate coefficients with the original Reconstruct the signal to obtain the denoised current and voltage signals: ; in, is the adaptive threshold, , is the noise standard deviation, is the number of signal sampling points.

[0012] Furthermore, the weighted fusion to form multi-dimensional fusion features specifically includes: The time domain feature vector, frequency domain feature vector and fault location direction discrimination results are combined according to the weight coefficient to obtain the multi-dimensional fusion feature: ; in, is a multi-dimensional fusion feature. 、 、 is the weighting coefficient, , 、 、 They are the time domain eigenvector, frequency domain eigenvector and fault location direction discrimination results respectively.

[0013] Furthermore, the three-level dynamic discrimination rules include, in order: Time domain primary judgment: If the amplitude of the current mutation point is greater than the amplitude threshold or the voltage fluctuation extreme value is greater than the fluctuation threshold, the fault flag is triggered; Frequency domain secondary discrimination: If the absolute current offset is greater than the first offset threshold or the absolute voltage offset is greater than the second offset threshold, a fault is confirmed to have occurred; Final judgment of fusion features: If the calculated weighted output multi-dimensional fusion features are greater than the fusion threshold, the fault is confirmed and its direction is determined.

[0014] Furthermore, the protection control logic is: Step 8.1, define the control signal , is the fusion threshold, where Indicates an upstream failure. Indicates a downstream fault; Step 8.2, if The upstream circuit breaker is tripped. The upstream protection is blocked and the downstream protection device is activated.

[0015] The embodiment of the present application provides a fault location optimization system for new energy power grid relay protection, including: a data acquisition module, an initial model generation module, a feature set optimization module, a hyperparameter optimization module, a model optimization module, and a protection control module; The data acquisition module is used to collect the current signal and voltage signal of the new energy grid in real time through the grid monitoring device and perform signal denoising; The initial model generation module is used to perform dual feature extraction in the time domain and frequency domain on the denoised current signal and voltage signal, generate time domain feature vectors and frequency domain feature vectors, and construct an SVM fault discrimination model to obtain the fault location direction discrimination result of the new energy grid relay protection; The feature set optimization module is used to perform weighted fusion of the time domain feature vector, the frequency domain feature vector and the fault location direction discrimination result to form a multi-dimensional fusion feature, perform dimensionality reduction processing on the multi-dimensional fusion feature, and generate an optimized feature set; The hyperparameter optimization module is used to build a support vector machine discrimination model based on the radial basis kernel function, use the particle swarm optimization algorithm to perform a global search for hyperparameters, optimize the penalty coefficient and kernel parameters, and then use the genetic algorithm for local tuning to collaboratively optimize the penalty coefficient and kernel parameters of the SVM fault discrimination model; The model optimization module is used to input the optimized feature set into the trained SVM fault discrimination model, combine the three-level dynamic discrimination rules of time domain amplitude threshold, frequency domain frequency offset threshold and fusion feature threshold, and comprehensively output the fault direction judgment result; The protection control module is used to generate a binary control signal according to the fault direction and perform fault section isolation according to the binary control signal.

[0016] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Through the dual optimization algorithm relay parameter adjustment mechanism, the global search of particle swarm and the local optimization capabilities of genetic algorithm are integrated to efficiently adapt the hyperparameters of support vector machines, thereby significantly improving the robustness of nonlinear fault classification and solving the problem of low efficiency of model parameter optimization and easy to fall into local optimality.

[0017] 2. Through the collaborative extraction and fusion of multi-dimensional features in the time-frequency domain and dimensionality reduction, an interference-resistant composite fault feature set is constructed, thereby achieving accurate characterization of complex working conditions such as frequency fluctuations and harmonic distortion, solving the problem of insufficient discrimination accuracy caused by incomplete extraction of traditional single-dimensional features.

[0018] 3. Through the dynamic weighted multi-level threshold fusion rule, time domain mutations, frequency domain offsets and model output characteristics are integrated in real time to generate the judgment boundary of adaptive power grid fluctuations, solving the problem of high misjudgment rate of fixed threshold rules under dynamic conditions of new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a fault location optimization method for new energy grid relay protection provided in an embodiment of the present application; Figure 2 A structural diagram of a fault location optimization system for relay protection of a new energy power grid provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] The embodiments of the present application provide a fault location optimization method and system for relay protection of new energy power grids, which solves the problems of low efficiency and easy local optimum in model parameter optimization in the prior art. Through the dual optimization algorithm relay parameter adjustment mechanism, the global search of particle swarm and the local optimization capability of genetic algorithm are integrated to efficiently adapt the hyperparameters of support vector machines, thereby significantly improving the robustness of nonlinear fault classification.

[0021] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0022] like Figure 1 FIG2 is a flow chart of a fault location optimization method for relay protection of a new energy power grid provided by an embodiment of the present application. The method is applied to a fault location optimization system for relay protection of a new energy power grid. The method includes the following steps: S1, collecting current signals and voltage signals of the new energy power grid in real time through a power grid monitoring device and performing signal denoising; S2 performs dual feature extraction in both the time and frequency domains on the denoised current and voltage signals. Wavelet transform is first used to extract the transient fluctuation characteristics of the signals. Fast Fourier transform is then used to decompose the frequency components, generating time and frequency domain feature vectors. An SVM fault discrimination model is then constructed to determine the fault location and direction of the new energy grid relay protection. S3, weighted fusion of time domain feature vectors, frequency domain feature vectors and fault location direction discrimination results to form multi-dimensional fusion features, and dimensionality reduction processing is performed on the multi-dimensional fusion features to generate an optimized feature set; S4: Build a support vector machine discrimination model based on the radial basis kernel function, use the particle swarm optimization algorithm to perform a global search for hyperparameters, optimize the penalty coefficient and kernel parameters, and then use the genetic algorithm for local tuning to collaboratively optimize the penalty coefficient and kernel parameters of the SVM fault discrimination model; S5: Input the optimized feature set into the trained SVM fault discrimination model, combine the three-level dynamic discrimination rules of time domain amplitude threshold, frequency domain frequency offset threshold and fusion feature threshold, and comprehensively output the fault direction judgment result; S6, generating a binary control signal according to the fault direction, and performing fault section isolation according to the binary control signal.

[0023] Furthermore, the time domain feature vector extraction specifically includes: Step 2.1, obtain the mother wavelet, the sliding window of the preset scale factor and translation factor; Step 2.2: Obtain the transient fluctuation characteristics of the current signal through continuous wavelet transform of the current signal and record the amplitude of the current mutation point. As the current time domain characteristics ; Step 2.3, synchronously obtain the transient fluctuation characteristics of the voltage signal through wavelet transform of the voltage signal and record the extreme value of the voltage fluctuation As the voltage time domain characteristic ; Step 2.4: Construct the time domain feature vector based on the current time domain feature and voltage time domain feature .

[0024] Furthermore, the continuous wavelet transform formula of the current signal is: ; in, is the transient fluctuation characteristic of the current signal, is the scale factor, is the translation factor, For my mother Xiaobo, is the current signal; The continuous wavelet transform formula of voltage signal is: ; in, is the transient fluctuation characteristic of the voltage signal, is a voltage signal.

[0025] Furthermore, the frequency domain feature vector extraction specifically includes: Step 3.1, performing fast Fourier transform on the current signal and the voltage signal to obtain the current spectrum and the voltage spectrum; ; in, Grid signal The frequency domain representation, namely the spectrum, of the power grid signal includes current signal and voltage signal. is pi, is the frequency, e is the natural constant, j is the imaginary unit, and t is the time; Step 3.2: Extract the actual value of the current fundamental frequency offset and the actual value of the voltage fundamental frequency offset , and obtain the rated power frequency current and rated power frequency voltage , get the absolute current offset and voltage absolute offset ; Step 3.3, from the current spectrum Extract the amplitude of the main harmonic components of the current , Number the main harmonics of current from the voltage spectrum Extract the amplitude of the main harmonic components of voltage , The number of the main harmonics of voltage; Step 3.4: Construct the frequency domain feature vector according to the absolute current offset, the absolute voltage offset, the amplitude of the main harmonic component of the current, and the amplitude of the main harmonic component of the voltage. .

[0026] Furthermore, the penalty coefficient and kernel parameters of the collaboratively optimized SVM fault discrimination model specifically include: Given a training sample , i is the sample number, N is the total number of samples, where are input features, is the corresponding label, and the goal of SVM is to find a hyperplane: , so that the sample points satisfy: ,in, is the normal vector of the hyperplane, b is the bias term; Step 4.1, define the radial basis kernel function , initialize the penalty coefficient and kernel parameters ; Step 4.2: Use particle swarm optimization to perform coarse parameter adjustment and initialize the particle swarm position. ,express Combine and iteratively update particle speed , iterative update , , , is the inertia weight, is the learning factor, , is the optimal position of an individual, For the global optimal position, evaluate the fitness and calculate the cross-validation error of SVM ; Step 4.3: Use genetic algorithm to fine tune parameters: Binary code the optimal solution output by particle swarm optimization and select it according to the crossover probability. Perform single-point crossover, according to the mutation probability Randomly flip gene bits.

[0027] Furthermore, the signal denoising step includes: For current signal and voltage signals Apply wavelet threshold denoising separately: select Symlet wavelet basis function, perform 3-layer wavelet decomposition, and obtain approximate coefficients (low-frequency component, representing the main signal) and the detail coefficient of the jth layer (high-frequency component, representing noise and mutation), j=1,2,3 represents the number of decomposition layers, and soft threshold processing is performed to obtain the detail coefficient after threshold processing , after threshold processing Approximate coefficients with the original Reconstruct the signal to obtain the denoised current and voltage signals: ; in, is the adaptive threshold, , is the noise standard deviation, estimated by the median of the detail coefficients at the finest scale (j=1), is the number of signal sampling points.

[0028] Furthermore, the weighted fusion to form multi-dimensional fusion features specifically includes: The time domain feature vector, frequency domain feature vector and fault location direction discrimination results are combined according to the weight coefficient to obtain the multi-dimensional fusion feature: ; in, is a multi-dimensional fusion feature. 、 、 is the weighting coefficient, , 、 、 They are the time domain eigenvector, frequency domain eigenvector and fault location direction discrimination results respectively.

[0029] Furthermore, the three-level dynamic discrimination rules include, in order: Time domain primary judgment: If the amplitude of the current mutation point is greater than the amplitude threshold or the voltage fluctuation extreme value is greater than the fluctuation threshold, the fault flag is triggered; Frequency domain secondary discrimination: If the absolute current offset is greater than the first offset threshold or the absolute voltage offset is greater than the second offset threshold, a fault is confirmed to have occurred; Final judgment of fusion features: If the calculated weighted output multi-dimensional fusion features are greater than the fusion threshold, the fault is confirmed and its direction is determined.

[0030] Furthermore, the protection control logic is: Step 8.1, define the control signal , is the fusion threshold, where Indicates an upstream failure. Indicates a downstream fault; Step 8.2, if If , the upstream circuit breaker is tripped, and if , the upstream protection is locked and the downstream protection device is started.

[0031] False positive and missed positive rates are key indicators of relay protection system reliability and directly impact power grid security. While the optimization algorithm outperforms traditional methods in accuracy, its false positive and missed positive rates also warrant attention. To this end, this solution conducted a detailed evaluation of these rates under various fault scenarios. Table 1 shows the false positive and missed positive rates for the optimization algorithm and traditional methods under different fault types.

[0032] Table 1: False positive rate and missed positive rate of the optimized algorithm and traditional method under different fault types

[0033] Experimental results show that in short-circuit and ground fault scenarios, the model's false positive rates were 0.4% and 0.5%, respectively, and its missed positive rates were 0.3% and 0.4%, respectively. In overload and renewable energy fluctuation scenarios, the false positive and missed positive rates increased slightly, but remained within acceptable limits, at 0.6% and 0.7%, respectively. These results demonstrate that the optimized model effectively reduces false positives and missed positives, and maintains low false positive and missed positive rates even in complex power grid environments.

[0034] like Figure 2 As shown, it is a structural diagram of a fault location optimization system for new energy power grid relay protection provided by an embodiment of the present application. A fault location optimization system for new energy power grid relay protection provided by an embodiment of the present application includes: a data acquisition module, an initial model generation module, a feature set optimization module, a hyperparameter optimization module, a model optimization module, and a protection control module; The data acquisition module is used to collect the current signal and voltage signal of the new energy grid in real time through the grid monitoring device and perform signal denoising; The initial model generation module is used to perform dual feature extraction in the time domain and frequency domain on the denoised current and voltage signals. First, wavelet transform is used to extract the transient fluctuation characteristics of the signal. Then, the frequency components are decomposed by fast Fourier transform to generate time domain feature vectors and frequency domain feature vectors. In addition, an SVM fault discrimination model is constructed to obtain the fault location direction discrimination result of the new energy grid relay protection. The feature set optimization module is used to perform weighted fusion of the time domain feature vector, the frequency domain feature vector and the fault location direction discrimination result to form a multi-dimensional fusion feature, perform dimensionality reduction processing on the multi-dimensional fusion feature, and generate an optimized feature set; The hyperparameter optimization module is used to build a support vector machine discrimination model based on the radial basis kernel function, use the particle swarm optimization algorithm to perform a global search for hyperparameters, optimize the penalty coefficient and kernel parameters, and then use the genetic algorithm for local tuning to collaboratively optimize the penalty coefficient and kernel parameters of the SVM fault discrimination model; The model optimization module is used to input the optimized feature set into the trained SVM fault discrimination model, combine the three-level dynamic discrimination rules of time domain amplitude threshold, frequency domain frequency offset threshold and fusion feature threshold, and comprehensively output the fault direction judgment result; The protection control module is used to generate a binary control signal according to the fault direction and perform fault section isolation according to the binary control signal.

[0035] In summary, the dual optimization algorithm relay parameter adjustment mechanism of the embodiment of the present application integrates the global search of the particle swarm and the local optimization capabilities of the genetic algorithm, thereby efficiently adapting the hyperparameters of the support vector machine, thereby significantly improving the robustness of nonlinear fault classification.

[0036] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0037] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0038] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0039] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0040] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0041] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A fault location optimization method for new energy power grid relay protection, characterized in that: The following steps are involved: S1, collects the current and voltage signals of the new energy grid in real time through the grid monitoring device and performs signal denoising; S2, performs dual feature extraction in the time domain and frequency domain on the denoised current and voltage signals, generates time domain feature vectors and frequency domain feature vectors, and constructs an SVM fault discrimination model to obtain the fault location direction discrimination results of the new energy grid relay protection; S3, weighted fusion of time domain feature vectors, frequency domain feature vectors and fault location direction discrimination results to form multi-dimensional fusion features, and dimensionality reduction processing is performed on the multi-dimensional fusion features to generate an optimized feature set; S4: Build a support vector machine discrimination model based on the radial basis kernel function, use the particle swarm optimization algorithm to perform a global search for hyperparameters, optimize the penalty coefficient and kernel parameters, and then use the genetic algorithm for local tuning to collaboratively optimize the penalty coefficient and kernel parameters of the SVM fault discrimination model; S5: Input the optimized feature set into the trained SVM fault discrimination model, combine the three-level dynamic discrimination rules of time domain amplitude threshold, frequency domain frequency offset threshold and fusion feature threshold, and comprehensively output the fault direction judgment result; S6, generating a binary control signal according to the fault direction, and performing fault section isolation according to the binary control signal.

2. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The time domain feature vector extraction specifically includes: Step 2.1, obtain the mother wavelet, the sliding window of the preset scale factor and translation factor; Step 2.2: Obtain the transient fluctuation characteristics of the current signal by continuous wavelet transform of the current signal, and record the amplitude of the current mutation point as the current time domain characteristic; Step 2.3, synchronously obtain the transient fluctuation characteristics of the voltage signal through wavelet transform of the voltage signal, and record the extreme value of the voltage fluctuation as the voltage time domain characteristic; Step 2.4: Construct a time domain feature vector based on the current time domain features and the voltage time domain features.

3. A fault location optimization method for new energy grid relay protection according to claim 2, characterized in that: The continuous wavelet transform formula of the current signal is: ； in, is the transient fluctuation characteristic of the current signal, is the scale factor, is the translation factor, For my mother Xiaobo, is the current signal; The continuous wavelet transform formula of voltage signal is: ； in, is the transient fluctuation characteristic of the voltage signal, is a voltage signal.

4. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The frequency domain feature vector extraction specifically includes: Step 3.1, performing fast Fourier transform on the current signal and the voltage signal to obtain the current spectrum and the voltage spectrum; Step 3.2: extract the actual value of the current fundamental frequency offset and the actual value of the voltage fundamental frequency offset, obtain the rated power frequency current and the rated power frequency voltage, and obtain the absolute current offset and the absolute voltage offset; Step 3.3, extract the amplitude of the main harmonic components of the current from the current spectrum and extract the amplitude of the main harmonic components of the voltage from the voltage spectrum; In step 3.4, a frequency domain feature vector is constructed based on the absolute current offset, the absolute voltage offset, the amplitude of the main harmonic component of the current, and the amplitude of the main harmonic component of the voltage.

5. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The penalty coefficient and kernel parameters of the collaboratively optimized SVM fault discrimination model specifically include: Step 4.1, define the radial basis kernel function, initialize the penalty coefficient and kernel parameters; Step 4.2: Use particle swarm optimization to roughly adjust the parameters, initialize the particle swarm position, evaluate the fitness and calculate the cross-validation error of SVM; In step 4.3, the genetic algorithm is used to fine-tune the parameters: the optimal solution output by the particle swarm optimization is binary-encoded, a single-point crossover is performed according to the crossover probability, and the gene position is randomly flipped according to the mutation probability.

6. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The signal denoising step comprises: Apply wavelet threshold denoising to the current signal and voltage signal respectively: select Symlet wavelet basis function, perform 3-layer wavelet decomposition, and obtain approximate coefficients and the detail coefficient of the jth layer , j=1,2,3 represents the number of decomposition layers, and performs soft threshold processing to obtain the detail coefficient after threshold processing , after threshold processing Approximate coefficients with the original Reconstruct the signal to obtain the denoised current and voltage signals: ； in, is the adaptive threshold, , is the noise standard deviation, is the number of signal sampling points.

7. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The weighted fusion to form multi-dimensional fusion features specifically includes: The time domain feature vector, frequency domain feature vector and fault location direction discrimination results are combined according to the weight coefficient to obtain the multi-dimensional fusion feature: ； in, is a multi-dimensional fusion feature. 、 、 is the weighting coefficient, , 、 、 They are the time domain eigenvector, frequency domain eigenvector and fault location direction discrimination results respectively.

8. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The three-level dynamic discrimination rules include, in order: Time domain primary judgment: If the amplitude of the current mutation point is greater than the amplitude threshold or the voltage fluctuation extreme value is greater than the fluctuation threshold, the fault flag is triggered; Frequency domain secondary discrimination: If the absolute current offset is greater than the first offset threshold or the absolute voltage offset is greater than the second offset threshold, a fault is confirmed to have occurred; Final judgment of fusion features: If the calculated weighted output multi-dimensional fusion features are greater than the fusion threshold, the fault is confirmed and its direction is determined.

9. A fault location optimization method for new energy grid relay protection according to claim 1, characterized in that: The protection control logic is: Step 8.1, define the control signal , is the fusion threshold, where Indicates an upstream failure. Indicates a downstream fault; Step 8.2, if The upstream circuit breaker is tripped. The upstream protection is blocked and the downstream protection device is activated.

10. A fault location optimization system for new energy power grid relay protection, characterized in that: include: Data acquisition module, initial model generation module, feature set optimization module, hyperparameter optimization module, model optimization module, protection control module; The data acquisition module is used to collect the current signal and voltage signal of the new energy grid in real time through the grid monitoring device and perform signal denoising; The initial model generation module is used to perform dual feature extraction in the time domain and frequency domain on the denoised current signal and voltage signal, generate time domain feature vectors and frequency domain feature vectors, and construct an SVM fault discrimination model to obtain the fault location direction discrimination result of the new energy grid relay protection; The feature set optimization module is used to perform weighted fusion of the time domain feature vector, the frequency domain feature vector and the fault location direction discrimination result to form a multi-dimensional fusion feature, perform dimensionality reduction processing on the multi-dimensional fusion feature, and generate an optimized feature set; The hyperparameter optimization module is used to build a support vector machine discrimination model based on the radial basis kernel function, use the particle swarm optimization algorithm to perform a global search for hyperparameters, optimize the penalty coefficient and kernel parameters, and then use the genetic algorithm for local tuning to collaboratively optimize the penalty coefficient and kernel parameters of the SVM fault discrimination model; The model optimization module is used to input the optimized feature set into the trained SVM fault discrimination model, combine the three-level dynamic discrimination rules of time domain amplitude threshold, frequency domain frequency offset threshold and fusion feature threshold, and comprehensively output the fault direction judgment result; The protection control module is used to generate a binary control signal according to the fault direction and perform fault section isolation according to the binary control signal.

Images