Ultra-high voltage high harmonic frequency measurement method, device and electronic equipment

Abstract

The present disclosure relates to the technical field of power system frequency measurement, and discloses a method, device, and electronic device for measuring ultra-high voltage high harmonic frequency, comprising: dynamically selecting the optimal window function type, window length, and sliding step size for the power system operating state based on the time-varying frequency signal characteristics of the power system using an adaptive window information design algorithm based on reinforcement learning; performing a short-time Fourier transform on the time-varying frequency signal based on the selected window function type, window length, and sliding step size, retaining the positive frequency components and deleting the negative frequency components during the short-time Fourier transform process, and converting the time-varying frequency signal from a time domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain; determining the spectral line position based on the two-dimensional time-frequency matrix, and determining the ultra-high voltage high harmonic frequency based on the main spectral line position, the sampling rate for the time-varying frequency signal, and the selected window length. The above method can improve the accuracy of the measurement results of the ultra-high voltage high harmonic frequency.

Description

Technical Field

[0001] The present disclosure relates to the technical field of power system frequency measurement, for example, to a method, device and electronic equipment for measuring ultra-high voltage high-order harmonic frequency. Background Art

[0002] With the development of power systems, especially the widespread adoption of ultra-high voltage transmission systems, accurate measurement of higher-order harmonic frequencies in power systems has become increasingly important. High-order harmonics not only affect power quality but can also cause equipment damage and communication interference. Therefore, accurate and real-time monitoring of these high-order harmonic frequencies is crucial to ensuring stable power system operation.

[0003] Existing methods for measuring high-order harmonic frequencies primarily include the traditional Fast Fourier Transform (FFT) and its variants, such as the Short-Time Fourier Transform (STFT). However, these methods suffer from several significant limitations in practical applications: While the traditional FFT method can provide high spectral resolution, due to its global nature, it cannot effectively capture the time-varying characteristics of the signal. This means it is not well suited to the rapidly changing high-order harmonic frequencies in power systems. The STFT method can alleviate this problem to a certain extent by introducing a window function, but the choice of window length directly affects the trade-off between time and frequency resolution. A shorter window improves time resolution but reduces spectral resolution, and vice versa. This trade-off makes it difficult to achieve both high accuracy and real-time performance.

[0004] Furthermore, in real signal processing, due to the symmetry of the Fourier transform, positive and negative frequency components always appear in pairs. Although simple filtering can be used to remove negative frequency components, this step increases computational complexity and, in turn, time consumption. Interpolation algorithms (such as three-line interpolation) are often required. However, existing interpolation methods often rely on a fixed window function and lack adaptability. This results in unstable interpolation accuracy under different operating conditions, making it difficult to ensure consistent, high-precision measurement results. Furthermore, the interpolation algorithm itself increases the computational burden, further extending processing time, especially when frequent parameter adjustments are required to adapt to dynamically changing power systems.

[0005] In summary, although the existing technology can meet the needs of high-order harmonic frequency measurement to a certain extent, it still has obvious shortcomings in the following aspects:

[0006] Both traditional FFT and STFT methods often require long computation times while maintaining high spectral resolution. This is particularly problematic when processing large amounts of data or high-frequency, dynamically changing signals. The use of fixed window functions and interpolation algorithms limits the accuracy of frequency estimation, making it difficult to achieve sustained, stable, and high-precision measurements, particularly in complex and changing power system environments. The lack of adaptive and intelligent methods prevents automatic optimization of measurement parameters based on the changing power system state, making it difficult to achieve optimal overall measurement performance. Summary of the Invention

[0007] In order to provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. The summary is not an extensive review, nor is it intended to identify key / critical elements or delineate the scope of protection of these embodiments, but rather serves as a prelude to the detailed description that follows.

[0008] The embodiments of the present disclosure provide a method, device and electronic equipment for measuring ultra-high voltage high harmonic frequencies, which can improve the accuracy of measurement results of ultra-high voltage high harmonic frequencies in power systems with low time consumption.

[0009] According to a first aspect of the present disclosure, a method for measuring ultra-high voltage high harmonic frequency is provided, the method comprising:

[0010] According to the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the optimal window function type, window length and sliding step size for the current power system operating state.

[0011] Based on the selected window function type, window length and sliding step size, a short-time Fourier transform is performed on the time-varying frequency signal, positive frequency components and negative frequency components in the transform result are identified during the short-time Fourier transform process, the positive frequency components are retained and the negative frequency components are deleted, thereby converting the time-varying frequency signal from a time domain representation into a two-dimensional time-frequency matrix represented in the time-frequency domain;

[0012] The spectral line positions are determined based on a two-dimensional time-frequency matrix, and the ultra-high voltage high harmonic frequencies of the power system are determined based on the main spectral line positions, the sampling rate of the time-varying frequency signal, and the selected window length.

[0013] In some embodiments, based on the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the optimal window function type, window length, and sliding step size for the current power system operating state, including:

[0014] Based on the time-varying frequency signal characteristics of the power system, an intelligent agent with perception capabilities is designed and deployed, and a set of window function types, window lengths, and sliding step sizes are defined. The intelligent agent can collect the power system frequency signal from the operating environment in real time.

[0015] Construct a policy gradient network as the core decision-making unit of the intelligent agent;

[0016] Construct an evaluation network, which receives spectrum information as input and evaluates the results of the agent's actions. Actions are defined as selecting different window function types, window lengths, and sliding steps.

[0017] The following window information selection process is performed iteratively:

[0018] The policy gradient network is used as the core decision-making unit of the intelligent agent. The intelligent agent takes the current high-order harmonic frequency distribution state of the power system as input, and uses the softmax function to perform sampling based on the high-order harmonic frequency distribution state to select an action from multiple optional actions.

[0019] The agent processes the frequency signal of the power system according to the selected action and selects the corresponding window function type, window length and sliding step size;

[0020] The evaluation network receives the results of the agent's actions and generates reward values ​​to feed back to the agent.

[0021] The agent calculates the gradient and updates the policy gradient network based on the current state of the high-order harmonic frequency distribution, the selected action, and the resulting reward value;

[0022] When the policy gradient network reaches a convergence state, the window function type, window length, and sliding step size corresponding to the currently selected action are determined as the optimal window function type, window length, and sliding step size for the current power system operation state;

[0023] When the policy gradient network has not reached the convergence state, the next window information selection process is continued.

[0024] In some embodiments, the high-order harmonic frequency distribution state includes the frequency component distribution, amplitude variation trend, and time-frequency correlation information between adjacent moments of the current signal.

[0025] In some embodiments, based on the selected window function type, window length, and sliding step size, a short-time Fourier transform is performed on the time-varying frequency signal, positive frequency components and negative frequency components in the transform result are identified during the short-time Fourier transform process, the positive frequency components are retained, and the negative frequency components are deleted, thereby converting the time-varying frequency signal from a time-domain representation into a two-dimensional time-frequency matrix represented in the time-frequency domain, including:

[0026] A shiftable finite-length window function based on the selected window function type and window length is used to truncate the time-varying frequency signal in the time domain, and a Fourier transform is performed on the windowed time-varying frequency signal at each window position to obtain the corresponding local spectrum characteristics;

[0027] During the short-time Fourier transform process, positive and negative frequency components in the transform result are identified, the positive frequency components are retained, and the negative frequency components are deleted;

[0028] The window function is moved sequentially according to the selected sliding step size, and the time axis of the entire time-varying frequency signal is gradually traversed. The local spectral features under each window position are combined into a spectral result sequence, and a two-dimensional time-frequency matrix is ​​constructed. Each column of the two-dimensional time-frequency matrix represents the segmented spectral result sequence obtained after windowing the signal at different time positions.

[0029] In some embodiments, the local spectrum feature is obtained by the following formula:

[0030] ,

[0031] Indicates the time point at which the window function slides along the time axis, represents the angular frequency, Indicates at a point in time At, angular frequency The local spectrum characteristics under represents the time domain variable, represents a time-varying frequency signal, Represents the window function centered at time point τ, and the two-dimensional time-frequency matrix consists of all Combination composition.

[0032] In some embodiments, determining a spectral line position based on a two-dimensional time-frequency matrix, and determining an ultra-high voltage high harmonic frequency of a power system based on a main spectral line position, a sampling rate for a time-varying frequency signal, and a selected window length, includes:

[0033] Perform a peak search operation on each column of the two-dimensional time-frequency matrix to identify the spectral line index with the largest energy as the integer part of the main spectral line position;

[0034] The elements in the two-dimensional time-frequency matrix are squared to enhance the peak contrast, and the spectral component with the local maximum amplitude, the spectral component with the local second maximum amplitude and the spectral component with the local third maximum amplitude are extracted therefrom;

[0035] Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component and the local third largest amplitude spectral component, combined with the polynomial fitting parameters corresponding to the selected window function, the decimal offset of the main spectral line position is calculated;

[0036] The ultra-high voltage high harmonic frequencies of the power system are determined based on the integer part and fractional offset of the main spectral line position, the sampling rate of the time-varying frequency signal, and the selected window length.

[0037] In some embodiments, the fractional offset of the main spectral line position is calculated based on the spectral component with the local maximum amplitude, the spectral component with the local second largest amplitude, and the spectral component with the local third largest amplitude, in combination with the polynomial fitting parameters corresponding to the selected window function, including:

[0038] Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component, and the local third largest amplitude spectral component, the interpolation coefficient is calculated using the following formula: ,in, represents the interpolation coefficient, represents the spectral component with the local maximum amplitude, represents the local sub-largest spectral component, Represents the spectral component with the third largest local amplitude;

[0039] Calculates the fractional offset of the main spectral line position based on the interpolation coefficients and the polynomial fit parameters corresponding to the selected window function.

[0040] In some embodiments, the ultra-high voltage high harmonic frequency of the power system is calculated using the following formula: ,in, Indicates the ultra-high voltage high harmonic frequency of the power system, Indicates the main line position, represents the integer part of the main spectral line position, Indicates the decimal offset of the main spectral line position, represents the sampling rate for time-varying frequency signals, Indicates the selected window length.

[0041] According to a second aspect of the present disclosure, there is provided an ultra-high voltage high harmonic frequency measurement device, comprising: a received signal processing module configured to: extract an effective received sound wave signal containing useful components within a target bandwidth from an original received sound wave signal of a target receiver, the target bandwidth being the frequency range of the sound wave signal emitted by a target sound source;

[0042] a window information selection module configured to dynamically select the optimal window function type, window length, and sliding step size for the current power system operating state using an adaptive window information design algorithm based on reinforcement learning, based on the time-varying frequency signal characteristics of the power system;

[0043] The frequency signal conversion module is configured to: perform a short-time Fourier transform on the time-varying frequency signal based on the selected window function type, window length, and sliding step size; identify positive frequency components and negative frequency components in the transformation result during the short-time Fourier transform process; retain the positive frequency components and delete the negative frequency components, thereby converting the time-varying frequency signal from a time-domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain;

[0044] The harmonic frequency determination module is configured to: determine the spectral line position based on the two-dimensional time-frequency matrix, and determine the ultra-high voltage high harmonic frequency of the power system based on the main spectral line position, the sampling rate for the time-varying frequency signal and the selected window length.

[0045] According to a third aspect of the present disclosure, an electronic device is provided, comprising a processor and a memory storing program instructions, wherein the processor is configured to execute the ultra-high voltage high harmonic frequency measurement method provided by the first aspect of the present disclosure when running the program instructions.

[0046] The ultra-high voltage high harmonic frequency measurement method, device, and electronic device provided by the embodiments of the present disclosure can achieve the following technical effects:

[0047] The ultra-high voltage high-order harmonic frequency measurement method provided by the embodiment of the present disclosure can dynamically adjust the window function type, window length and sliding step size according to the real-time operating status of the power system, thereby improving the time response capability while ensuring a high frequency resolution. In addition, only the positive frequency components are retained in the transformation result of the short-time Fourier transform process, and the complexity of subsequent processing is reduced by ignoring the negative frequency components in the short-time Fourier transform result; and the spectral line position is determined by a two-dimensional time-frequency matrix, and the ultra-high voltage high-order harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate for the time-varying frequency signal and the selected window length, thereby further improving the frequency estimation accuracy. Ultimately, it is achieved to improve the accuracy of the measurement results of the ultra-high voltage high-order harmonic frequency in the power system under the premise of low time consumption.

[0048] The foregoing general description and the following description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] One or more embodiments are exemplarily described by corresponding drawings. These exemplary descriptions and drawings do not limit the embodiments. Elements with the same reference numerals in the drawings are shown as similar elements. The drawings do not constitute a scale limitation. In addition,

[0050] Figure 1 This is a flow chart of a method for measuring ultra-high voltage high-order harmonic frequency provided by an embodiment of the present disclosure;

[0051] Figure 2is a flow chart of an adaptive window information design algorithm provided by an embodiment of the present disclosure;

[0052] Figure 3 is a structural diagram of a policy gradient network provided by an embodiment of the present disclosure;

[0053] Figure 4 1 is a flow chart of another method for measuring ultra-high voltage high harmonic frequency provided by an embodiment of the present disclosure;

[0054] Figure 5 1 is a flow chart of another method for measuring ultra-high voltage high harmonic frequency provided by an embodiment of the present disclosure;

[0055] Figure 6 1 is a structural diagram of an ultra-high voltage high harmonic frequency measurement device provided by an embodiment of the present disclosure;

[0056] Figure 7 is a schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0057] In order to be able to understand the features and technical content of the embodiments of the present disclosure in more detail, the implementation of the embodiments of the present disclosure is described in detail below in conjunction with the accompanying drawings. The accompanying drawings are for reference only and are not used to limit the embodiments of the present disclosure. In the following technical description, for the sake of convenience of explanation, a full understanding of the disclosed embodiments is provided through multiple details. However, one or more embodiments can still be implemented without these details. In other cases, to simplify the drawings, well-known structures and devices can be simplified for display.

[0058] In the description and claims of the embodiments of the present disclosure, as well as in the accompanying drawings, the terms "first," "second," and the like are used to distinguish similar items and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate to describe the embodiments of the present disclosure herein. In addition, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions.

[0059] Unless otherwise stated, the term "plurality" means two or more.

[0060] In the embodiment of the present disclosure, the character " / " indicates that the preceding and following objects are in an "or" relationship. For example, A / B means: A or B.

[0061] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0062] The term "correspondence" may refer to an association relationship or a binding relationship. The correspondence between A and B means that there is an association relationship or a binding relationship between A and B.

[0063] The present disclosure provides a method for measuring ultra-high voltage high harmonic frequency. Figure 1 As shown, the ultra-high voltage high harmonic frequency measurement method includes:

[0064] S101 , according to the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the window function type, window length, and sliding step size that are optimal for the current power system operating state.

[0065] In an embodiment of the present disclosure, a time-varying frequency signal of a power system can be received and discretely sampled to determine the characteristics of the time-varying frequency signal. Then, based on an adaptive window information design algorithm, the optimal window function type, window length, and sliding step size for the current power system operating state are dynamically selected. The adaptive window information design algorithm is obtained using a reinforcement learning (RL) algorithm. By dynamically sensing the time-frequency characteristics of the time-varying frequency signal, the adaptive window information design algorithm can achieve coordinated adaptive matching of the window function type, window length, and sliding step size parameters.

[0066] S102, based on the selected window function type, window length and sliding step size, performs short-time Fourier transform on the time-varying frequency signal, identifies the positive frequency components and negative frequency components in the transformation result during the short-time Fourier transform process, retains the positive frequency components and deletes the negative frequency components, thereby converting the time-varying frequency signal from the time domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain.

[0067] In the disclosed embodiments, the short-time Fourier transform (STFT) process retains only positive frequency components, ignoring negative frequency components in the STFT result, reducing the complexity of subsequent processing. The time-varying frequency signal is converted into a two-dimensional time-frequency matrix represented in the time-frequency domain, effectively revealing the mathematical essence of the signal's time-frequency characteristics and laying the theoretical foundation for the subsequent construction of a fast solution algorithm.

[0068] S103, determining the spectral line position based on the two-dimensional time-frequency matrix, and determining the ultra-high voltage high harmonic frequency of the power system based on the main spectral line position, the sampling rate for the time-varying frequency signal, and the selected window length.

[0069] The ultra-high voltage high-order harmonic frequency measurement method provided by the embodiment of the present disclosure can dynamically adjust the window function type, window length and sliding step size according to the real-time operating status of the power system, thereby improving the time response capability while ensuring a high frequency resolution. In addition, only the positive frequency components are retained in the transformation result of the short-time Fourier transform process, and the complexity of subsequent processing is reduced by ignoring the negative frequency components in the short-time Fourier transform result; and the spectral line position is determined by a two-dimensional time-frequency matrix, and the ultra-high voltage high-order harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate for the time-varying frequency signal and the selected window length, thereby further improving the frequency estimation accuracy. Ultimately, it is achieved to improve the accuracy of the measurement results of the ultra-high voltage high-order harmonic frequency in the power system under the premise of low time consumption.

[0070] In some embodiments, based on the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the optimal window function type, window length, and sliding step size for the current power system operating state, including:

[0071] Based on the time-varying frequency signal characteristics of the power system, an intelligent agent with perception capabilities is designed and deployed, and a set of window function types, window lengths, and sliding step sizes are defined. The intelligent agent can collect the frequency signals of the power system from the operating environment in real time. A policy gradient (PG) network is constructed as the core decision-making unit of the intelligent agent. An evaluation network is constructed, which receives spectrum information as input and evaluates the results of the actions performed by the intelligent agent. The actions are defined as the selection of different window function types, window lengths, and sliding step sizes.

[0072] The following window information selection process is performed iteratively:

[0073] The policy gradient network is used as the core decision-making unit of the intelligent agent. The intelligent agent takes the current high-order harmonic frequency distribution state of the power system as input, and uses the softmax function to perform sampling based on the high-order harmonic frequency distribution state to select an action from multiple optional actions.

[0074] The agent processes the frequency signal of the power system according to the selected action and selects the corresponding window function type, window length and sliding step size;

[0075] The evaluation network receives the results of the agent's actions and generates reward values ​​to feed back to the agent.

[0076] The agent calculates the gradient and updates the policy gradient network based on the current state of the high-order harmonic frequency distribution, the selected action, and the resulting reward value;

[0077] When the policy gradient network reaches a convergence state, the window function type, window length, and sliding step size corresponding to the currently selected action are determined as the optimal window function type, window length, and sliding step size for the current power system operation state;

[0078] When the policy gradient network has not reached the convergence state, the next window information selection process is continued.

[0079] The adaptive window information design algorithm used in this disclosure, based on reinforcement learning (RL), belongs to the field of reinforcement learning, a key branch of machine learning. This method focuses on achieving goal-oriented learning decisions through interaction with the environment and is suitable for complex systems with dynamically changing characteristics.

[0080] In the reinforcement learning system, state, action, and reward constitute the core loop of the interaction between the agent and the environment. Specifically:

[0081] State: Typically used as a model input to characterize the current system operating status. In this solution, the state represents the distribution of ultra-high voltage high-order harmonic frequencies in the power system, including key information such as signal frequency components, amplitude variation trends, and time-frequency correlations between adjacent moments.

[0082] Action: The decision made by the agent based on the current state, which is the output of the model. In this solution, the action is defined as the selection of the window function type, window length, and sliding step size.

[0083] Reward: Feedback from the environment regarding the results of an agent's actions is the core basis for model learning and optimization strategies. Reward values ​​reflect the quality of action choices, and the agent continuously adjusts its decision-making strategy by maximizing long-term cumulative rewards.

[0084] In the RL system, an agent with autonomous learning ability observes the current state at each time step. , and select actions based on the current strategy Interact with the external environment; the environment is then updated to the new state , and feedback the reward value to the agent Through the continuous iteration of this process, the agent gradually learns the optimal strategy to maximize long-term performance.

[0085] To formally model this decision-making process, this paper abstracts the ultra-high voltage high-harmonic frequency measurement problem into a Markov decision process (MDP). MDPs are a standard framework widely used in reinforcement learning, clearly describing the dynamic interactions between an agent and its environment. Almost all reinforcement learning tasks can be modeled within this framework.

[0086] An MDP is defined by a five-tuple (S, A, T, R, γ), where:

[0087] State space S: defined as a finite set , including various high-order harmonic frequency distribution states that may appear in the power system;

[0088] Action space A: defined as a finite set , represents all possible actions that the agent may take during its interaction with the power system;

[0089] The state transition T is defined as When in state Application Action When , the probability of the system transitioning from one high-order harmonic frequency state to another state is determined by the probability distribution of the state generated by the transition function T. Obviously , and ;

[0090] Reward function R: defined as R:S×A×S→R, which provides immediate feedback for a specific state transition;

[0091] Discount Factor : Used to measure the degree of discount of future rewards. , the agent only focuses on the immediate reward, showing "myopic" behavior; as As it approaches 1, the agent will place more emphasis on potential future gains.

[0092] In the adaptive window information design algorithm provided in the disclosed embodiment, the specific definitions are as follows:

[0093] Different types of signals are defined as ∈State;

[0094] The selection of window function type, window length, and sliding step size are defined as:

[0095] : Window function set;

[0096] : Window length set;

[0097] : Sliding step set;

[0098] The evaluation accuracy of the evaluation network (E-NET) on the signal spectrum is defined as ∈Reward.

[0099] Window function set Includes the following commonly used window function types:

[0100] Rectangular window, hanning, hamming, Blackman(), Blackman-Harris, 4-term minimum sidelobe Nuttall, 4-term 5th order Nuttall, RV(Ⅲ)-5, RV(Ⅰ)-5.

[0101] Window length set Set to:

[0102] {200,250,300,350,400,450,500,550,600,650,700,750,800}.

[0103] Sliding step set Set to:

[0104] {100,125,150,175,200,225,250,275,300,325,350,375,400}.

[0105] Sliding step set It can also be set to: {100,150,200,240,300,400}.

[0106] Figure 2 The overall flow chart of the adaptive window information design algorithm is shown. Figure 3 The specific structure of the policy gradient network is shown. In addition, the system is also equipped with multiple functional modules, including state extraction module, reward module, state module, training data and verification data module, etc.

[0107] Combine Figure 2 and Figure 3 As shown in Figure 2, the intelligent agent is the decision-making core of the entire system, responsible for interacting with the power system and receiving the current state As input, the agent, based on a policy gradient network, samples through a softmax function and selects one action from multiple candidate actions. Its decision directly affects the subsequent trajectory of the system and serves as a key bridge between state perception and action execution.

[0108] The Policy Gradient Network (PGN) is designed to gradually compress the spatial dimensions of the input while increasing the number of feature maps, ultimately forming a decision layer with a classification output. Convolutional layers extract local features from the input sequence, while pooling layers reduce the data dimensionality. These alternating convolution and pooling operations gradually shorten the sequence length. Finally, fully connected layers integrate local features, enabling effective modeling of long sequences.

[0109] The state extraction module is responsible for generating spectrum information. A spectrum is a representation of a signal in the time-frequency domain. It clearly reflects the temporal trends of the signal's frequency components, providing critical support for subsequent evaluation. For example, in audio processing, the spectrum can be used to analyze frequency characteristics at different moments in time. Part of the generated spectrum is used to evaluate network input, while the remaining part is used in subsequent steps such as state updates.

[0110] The evaluation network uses information such as the spectrum as input to assess the effectiveness of the agent's actions, calculates a reward, and provides feedback to the agent. The reward reflects the quality of the action. The evaluation network also outputs metrics such as accuracy, which are used to measure the overall system performance and provide guidance for the agent's learning direction.

[0111] The reward module is responsible for returning reward values, generated by the evaluation network. This is a key feedback signal in the agent's learning process. The agent's goal is to maximize long-term cumulative rewards by continuously adjusting its strategy to find the optimal action combination.

[0112] The state module is used to record the operating state of the system at a certain moment, including the current signal characteristics and other information required for the agent to make decisions. Make an action decision; after executing the action, the system state will be updated to the new state The accurate representation and timely update of the state are of great significance to the decision-making of the intelligent agent. It not only reflects the dynamic evolution process of the system, but also provides the basis for the intelligent agent to perceive the environment and make decisions.

[0113] The training and validation data modules provide the data sources required for training and evaluation. Training data is obtained through sampling and used for agent learning and strategy optimization. Validation data, also obtained through sampling, is used to evaluate the model's generalization ability during training, detect problems such as overfitting, and ensure the system's stability and effectiveness in real-world applications.

[0114] The workflow of the entire system is as follows:

[0115] During the system initialization phase, the agent obtains the initial state from the power system and initializes the strategy network and evaluation network.

[0116] Entering the information interaction stage, the agent Select an action And execute, the environment returns to the new state Reward Value ;

[0117] The evaluation network evaluates the value of the new state and feeds back the reward to the agent;

[0118] Based on the current state, action, and reward information, the system calculates the gradient and updates the policy network, enabling the agent to make better decisions in the future.

[0119] The above process is repeated until the policy gradient network converges. At this point, the higher the reward value obtained by the agent, the better the selected set of action parameters ( , , ) is more suitable for the current situation.

[0120] Through the above mechanism, the present invention realizes the intelligent optimization selection of window function type, window length and sliding step size, which significantly improves the accuracy and efficiency of high-order harmonic frequency measurement.

[0121] In some embodiments, the high-order harmonic frequency distribution state includes the frequency component distribution, amplitude variation trend, and time-frequency correlation information between adjacent moments of the current signal.

[0122] By capturing and analyzing the frequency component distribution of the current signal, it is possible to accurately identify the various frequency components present in the power system, particularly those higher harmonics that affect power quality. Analysis based on frequency component distribution can more precisely locate and quantify the presence and intensity of individual harmonics, thereby improving the accuracy of overall frequency measurement. Monitoring the changing trends of signal amplitudes helps understand the real-time operating conditions of the power system. Using amplitude trends as input features, the intelligent agent can more quickly respond to changes in system status and select window function parameters that best suit the current conditions, thereby improving the measurement process's adaptability to dynamic changes. Considering the time-frequency correlation between adjacent moments can help better understand the temporal characteristics of the signal and its temporal evolution. By analyzing time-frequency correlation, it is possible to optimize temporal resolution while maintaining high spectral resolution. For example, during periods of rapid signal fluctuations, a shorter window length is selected to improve temporal resolution; during periods of relatively stable signals, a longer window length is selected to achieve higher spectral resolution. This adaptive adjustment significantly improves measurement accuracy and efficiency.

[0123] In some embodiments, based on the selected window function type, window length and sliding step size, a short-time Fourier transform is performed on the time-varying frequency signal, and in the short-time Fourier transform process, positive frequency components and negative frequency components in the transform result are identified, the positive frequency components are retained and the negative frequency components are deleted, thereby converting the time-varying frequency signal from a time domain representation into a two-dimensional time-frequency matrix represented in the time-frequency domain, including: using a translatable finite-length window function based on the selected window function type and window length to perform time domain truncation on the time-varying frequency signal, and performing a Fourier transform on the windowed time-varying frequency signal at each window position to obtain corresponding local spectral features; in the short-time Fourier transform process, positive frequency components and negative frequency components in the transform result are identified, the positive frequency components are retained and the negative frequency components are deleted; the window function is moved sequentially according to the selected sliding step size, and the time axis of the entire time-varying frequency signal is gradually traversed, and the local spectral features at each window position are combined into a spectral result sequence to construct a two-dimensional time-frequency matrix, wherein each column of the two-dimensional time-frequency matrix represents a segmented spectral result sequence obtained after windowing the signal at different time positions.

[0124] Through reinforcement learning, the optimal window length and sliding step size are dynamically selected. When the signal changes dramatically, a shorter window and small step size are used to improve the time resolution; in the stable phase, a longer window is used to improve the frequency resolution. This achieves an adaptive balance between time resolution and frequency resolution, significantly improving the dynamic response capability and applicability of the measurement system, especially for the rapid changes of ultra-high voltage and high-order harmonics in power systems. The negative frequency components are directly ignored, and only the positive frequency part is retained for subsequent analysis, which reduces the data dimension and the amount of calculation, improves the speed and real-time performance of frequency estimation, and avoids interpolation errors caused by negative frequency interference. The local spectrum results at each window position are organized into a two-dimensional time-frequency matrix, with each column corresponding to the spectrum information of a time segment. A structured time-frequency representation model is constructed to clearly reflect the time-frequency evolution process of the signal.

[0125] The present disclosure provides another method for measuring ultra-high voltage high harmonic frequency. Figure 4 As shown, the ultra-high voltage high harmonic frequency measurement method includes:

[0126] S401 , according to the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the window function type, window length, and sliding step size that are optimal for the current power system operating state.

[0127] S402, using a shiftable finite-length window function based on the selected window function type and window length to perform time-domain truncation on the time-varying frequency signal, and performing Fourier transform on the windowed time-varying frequency signal at each window position to obtain corresponding local spectrum features.

[0128] S403 , identifying positive frequency components and negative frequency components in the transformation result during the short-time Fourier transform process, retaining the positive frequency components and deleting the negative frequency components.

[0129] S404, sequentially moving the window function according to the selected sliding step size, gradually traversing the time axis of the entire time-varying frequency signal, combining the local spectrum features at each window position into a spectrum result sequence, and constructing a two-dimensional time-frequency matrix.

[0130] Each column of the two-dimensional time-frequency matrix represents a sequence of segmented spectrum results obtained after windowing the signal at different time positions.

[0131] S405 , determining the spectral line position based on the two-dimensional time-frequency matrix, and determining the ultra-high voltage high harmonic frequency of the power system based on the main spectral line position, the sampling rate for the time-varying frequency signal, and the selected window length.

[0132] In some embodiments, the local spectrum feature is obtained by the following formula:

[0133] ,

[0134] Indicates the time point at which the window function slides along the time axis, represents the angular frequency, Indicates at a point in time At, angular frequency The local spectrum characteristics under represents the time domain variable, represents a time-varying frequency signal, Represents the window function centered at time point τ, and the two-dimensional time-frequency matrix consists of all Combination composition.

[0135] In some embodiments, the spectral line position is determined based on a two-dimensional time-frequency matrix, and the ultra-high voltage high-order harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate for the time-varying frequency signal, and the selected window length, including: performing a peak search operation on each column of the two-dimensional time-frequency matrix, identifying the spectral line index with the largest energy, as the integer part of the main spectral line position; squaring the elements in the two-dimensional time-frequency matrix to enhance the peak contrast, and extracting the local maximum amplitude spectral component, the local second-largest amplitude spectral component, and the local third-largest amplitude spectral component therefrom; based on the local maximum amplitude spectral component, the local second-largest amplitude spectral component, and the local third-largest amplitude spectral component, combined with the polynomial fitting parameters corresponding to the selected window function, calculating the decimal offset of the main spectral line position; based on the integer part and decimal offset of the main spectral line position, the sampling rate for the time-varying frequency signal, and the selected window length, the ultra-high voltage high-order harmonic frequency of the power system is determined.

[0136] The disclosed embodiment performs a peak search on a two-dimensional time-frequency matrix to identify the integer positions of the main spectral lines, and combines the three-line interpolation algorithm with the polynomial fitting parameters related to the window function to further calculate the fractional offset of the main spectral line, thereby achieving high-precision measurement of ultra-high voltage high-order harmonic frequencies in the power system. This method not only significantly improves the accuracy and stability of frequency estimation, but also enhances the system's adaptability to a variety of window functions and real-time processing efficiency by introducing a square enhancement mechanism and structured matrix processing. Therefore, this technology is particularly suitable for the dynamic monitoring needs of high-precision, low-latency high-order harmonic frequencies in complex power systems.

[0137] The present disclosure provides a method for measuring ultra-high voltage high harmonic frequency. Figure 5 As shown, the ultra-high voltage high harmonic frequency measurement method includes:

[0138] S501 , according to the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the window function type, window length, and sliding step size that are optimal for the current power system operating state.

[0139] S502, based on the selected window function type, window length and sliding step size, performs short-time Fourier transform on the time-varying frequency signal, identifies the positive frequency components and negative frequency components in the transformation result during the short-time Fourier transform process, retains the positive frequency components and deletes the negative frequency components, thereby converting the time-varying frequency signal from the time domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain.

[0140] S503 , performing a peak search operation on each column of the two-dimensional time-frequency matrix, identifying the spectrum line index with the maximum energy as the integer part of the main spectrum line position.

[0141] S504 , square the elements in the two-dimensional time-frequency matrix to enhance peak contrast, and extract the spectral component with the local maximum amplitude, the spectral component with the local second largest amplitude, and the spectral component with the local third largest amplitude.

[0142] S505 , calculating the decimal offset of the main spectral line position based on the local maximum amplitude spectral component, the local second largest amplitude spectral component and the local third largest amplitude spectral component in combination with the polynomial fitting parameters corresponding to the selected window function.

[0143] S506 , determining the ultra-high voltage higher harmonic frequency of the power system based on the integer part and the fractional offset of the main spectrum line position, the sampling rate for the time-varying frequency signal, and the selected window length.

[0144] In some embodiments, after the elements in the two-dimensional time-frequency matrix are squared, the expression of the two-dimensional time-frequency matrix is ​​as follows:

[0145] .

[0146] Represents the two-dimensional time-frequency matrix table after square processing. Typically represents time-dependent coefficients, such as signal characteristics within a time window. Typically represents a frequency-dependent coefficient, such as the amplitude of a frequency component.

[0147] In some embodiments, the fractional offset of the main spectral line position is calculated based on the spectral component with the local maximum amplitude, the spectral component with the local second largest amplitude, and the spectral component with the local third largest amplitude, in combination with the polynomial fitting parameters corresponding to the selected window function, including:

[0148] Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component, and the local third largest amplitude spectral component, the interpolation coefficient is calculated using the following formula: ,in, represents the interpolation coefficient, represents the spectral component with the local maximum amplitude, represents the local sub-largest spectral component, Represents the spectral component with the third largest local amplitude; calculates the fractional offset of the main spectral line position based on the interpolation coefficients and the polynomial fitting parameters corresponding to the selected window function.

[0149] Table 1 provides the interpolation coefficients corresponding to different window functions in a specific scenario. These coefficients are used to calculate the decimal offset of the main spectral line position, helping to improve the accuracy and stability of frequency measurement.

[0150] .

[0151] In some embodiments, the ultra-high voltage high harmonic frequency of the power system is calculated using the following formula: ,in, Indicates the ultra-high voltage high harmonic frequency of the power system, Indicates the main line position, represents the integer part of the main spectral line position, Indicates the decimal offset of the main spectral line position, represents the sampling rate for time-varying frequency signals, Indicates the selected window length.

[0152] Combine Figure 6 As shown, the present disclosure provides an ultra-high voltage high harmonic frequency measurement device 1000 , which includes a window information selection module 1001 , a frequency signal conversion module 1002 and a harmonic frequency determination module 1003 .

[0153] The window information selection module 1001 is configured to dynamically select the optimal window function type, window length and sliding step size for the current power system operating state using an adaptive window information design algorithm based on reinforcement learning according to the time-varying frequency signal characteristics of the power system.

[0154] The frequency signal conversion module 1002 is configured to: perform a short-time Fourier transform on the time-varying frequency signal based on the selected window function type, window length and sliding step size, identify the positive frequency components and negative frequency components in the transformation result during the short-time Fourier transform process, retain the positive frequency components and delete the negative frequency components, thereby converting the time-varying frequency signal from the time domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain.

[0155] The harmonic frequency determination module 1003 is configured to determine the spectral line position based on the two-dimensional time-frequency matrix, and determine the ultra-high voltage high harmonic frequency of the power system based on the main spectral line position, the sampling rate for the time-varying frequency signal and the selected window length.

[0156] The ultra-high voltage high-order harmonic frequency measurement device provided by the embodiment of the present disclosure can dynamically adjust the window function type, window length and sliding step size according to the real-time operating status of the power system, thereby improving the time response capability while ensuring a high frequency resolution. In addition, by ignoring the negative frequency components in the short-time Fourier transform results, the complexity of subsequent processing is reduced; and the spectral line position is determined by a two-dimensional time-frequency matrix, and the ultra-high voltage high-order harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate for the time-varying frequency signal and the selected window length, thereby further improving the frequency estimation accuracy. Ultimately, it is achieved to improve the accuracy of the measurement results of the ultra-high voltage high-order harmonic frequency in the power system under the premise of low time consumption.

[0157] In some embodiments, the window information selection module 1001 is configured to:

[0158] Based on the time-varying frequency signal characteristics of the power system, an intelligent agent with perception capabilities is designed and deployed, and a set of window function types, window lengths, and sliding step sizes are defined. The intelligent agent can collect the power system frequency signal from the operating environment in real time.

[0159] Construct a policy gradient network as the core decision-making unit of the intelligent agent;

[0160] Construct an evaluation network, which receives spectrum information as input and evaluates the results of the agent's actions. Actions are defined as selecting different window function types, window lengths, and sliding steps.

[0161] The following window information selection process is performed iteratively:

[0162] The policy gradient network is used as the core decision-making unit of the intelligent agent. The intelligent agent takes the current high-order harmonic frequency distribution state of the power system as input, and uses the softmax function to perform sampling based on the high-order harmonic frequency distribution state to select an action from multiple optional actions.

[0163] The agent processes the frequency signal of the power system according to the selected action and selects the corresponding window function type, window length and sliding step size;

[0164] The evaluation network receives the results of the agent's actions and generates reward values ​​to feed back to the agent.

[0165] The agent calculates the gradient and updates the policy gradient network based on the current state of the high-order harmonic frequency distribution, the selected action, and the resulting reward value;

[0166] When the policy gradient network reaches a convergence state, the window function type, window length, and sliding step size corresponding to the currently selected action are determined as the optimal window function type, window length, and sliding step size for the current power system operation state;

[0167] When the policy gradient network has not reached the convergence state, the next window information selection process is continued.

[0168] In some embodiments, the high-order harmonic frequency distribution state includes the frequency component distribution, amplitude variation trend, and time-frequency correlation information between adjacent moments of the current signal.

[0169] In some embodiments, the frequency signal conversion module 1002 is configured to:

[0170] A shiftable finite-length window function based on the selected window function type and window length is used to truncate the time-varying frequency signal in the time domain, and a Fourier transform is performed on the windowed time-varying frequency signal at each window position to obtain the corresponding local spectrum characteristics;

[0171] During the short-time Fourier transform process, positive and negative frequency components in the transform result are identified, the positive frequency components are retained, and the negative frequency components are deleted;

[0172] The window function is moved sequentially according to the selected sliding step size, and the time axis of the entire time-varying frequency signal is gradually traversed. The local spectral features under each window position are combined into a spectral result sequence, and a two-dimensional time-frequency matrix is ​​constructed. Each column of the two-dimensional time-frequency matrix represents the segmented spectral result sequence obtained after windowing the signal at different time positions.

[0173] In some embodiments, the local spectrum feature is obtained by the following formula:

[0174] ;

[0175] Indicates the time point at which the window function slides along the time axis, represents the angular frequency, Indicates at a point in time At, angular frequency The local spectrum characteristics under represents the time domain variable, represents a time-varying frequency signal, Represents the window function centered at time point τ, and the two-dimensional time-frequency matrix consists of all Combination composition.

[0176] In some embodiments, the harmonic frequency determination module 1003 is configured to:

[0177] Perform a peak search operation on each column of the two-dimensional time-frequency matrix to identify the spectral line index with the largest energy as the integer part of the main spectral line position;

[0178] The elements in the two-dimensional time-frequency matrix are squared to enhance the peak contrast, and the spectral component with the local maximum amplitude, the spectral component with the local second maximum amplitude and the spectral component with the local third maximum amplitude are extracted therefrom;

[0179] Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component and the local third largest amplitude spectral component, combined with the polynomial fitting parameters corresponding to the selected window function, the decimal offset of the main spectral line position is calculated;

[0180] The ultra-high voltage high harmonic frequencies of the power system are determined based on the integer part and fractional offset of the main spectral line position, the sampling rate of the time-varying frequency signal, and the selected window length.

[0181] In some embodiments, the harmonic frequency determination module 1003 is configured to:

[0182] Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component, and the local third largest amplitude spectral component, the interpolation coefficient is calculated using the following formula: ,in, represents the interpolation coefficient, represents the spectral component with the local maximum amplitude, represents the local sub-largest spectral component, Represents the spectral component with the third largest local amplitude;

[0183] Calculates the fractional offset of the main spectral line position based on the interpolation coefficients and the polynomial fit parameters corresponding to the selected window function.

[0184] In some embodiments, the ultra-high voltage high harmonic frequency of the power system is calculated using the following formula: ,in, Indicates the ultra-high voltage high harmonic frequency of the power system, Indicates the main line position, represents the integer part of the main spectral line position, Indicates the decimal offset of the main spectral line position, represents the sampling rate for time-varying frequency signals, Indicates the selected window length.

[0185] Combine Figure 7 As shown, an embodiment of the present disclosure provides an electronic device 2000, which includes a processor 2001 and a memory 2002. Optionally, the electronic device 2000 may further include a communication interface 2003 and a bus 2004. The processor 2001, the communication interface 2003, and the memory 2002 may communicate with each other through the bus 2004. The communication interface 2003 may be used for information transmission. The processor 2001 may call the logic instructions in the memory 2002 to execute the ultra-high voltage high harmonic frequency measurement method of the above embodiment. In addition, the logic instructions in the above memory 2002 may be implemented in the form of a software functional unit and may be stored in a computer-readable storage medium when sold or used as an independent product.

[0186] Memory 2002, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as the program instructions / modules corresponding to the methods in the embodiments of the present disclosure. Processor 2001 executes the program instructions / modules stored in memory 2002 to perform functional applications and data processing, thereby implementing the ultra-high voltage high-order harmonic frequency measurement method in the above-described embodiments.

[0187] The memory 2002 may include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function; the data storage area may store data generated based on the use of the terminal device. Furthermore, the memory 2002 may include high-speed random access memory and non-volatile memory.

[0188] The above description and the accompanying drawings sufficiently illustrate the embodiments of the present disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, process and other changes. The embodiments represent only possible variations. Unless expressly required, individual components and functions are optional, and the order of operations may vary. Portions and features of some embodiments may be included in or replace portions and features of other embodiments. Moreover, the terms used in this application are only used to describe the embodiments and are not used to limit the claims. As used in the description of the embodiments and claims, the singular forms "a", "an" and "the" are intended to also include the plural forms unless the context clearly indicates otherwise. Similarly, the term "and / or" as used in this application means any and all possible combinations of one or more of the associated listings. In addition, when used in this application, the term "comprise" and its variations "comprises" and / or comprising refer to the presence of stated features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof. In the absence of further restrictions, an element defined by the sentence "comprising a..." does not exclude the presence of other identical elements in the process, method or device that includes the element. In this article, each embodiment may focus on the differences from other embodiments, and the same and similar parts between the embodiments can be referred to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts can be referred to the description of the method part.

[0189] The flowcharts and block diagrams in the accompanying drawings show the possible implementation architectures, functions and operations of the systems, methods and computer program products according to the embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical functions. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified function or action, or may be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A method for measuring ultra-high voltage high harmonic frequency, characterized in that: include: According to the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the optimal window function type, window length and sliding step size for the current power system operating state. Based on the selected window function type, window length and sliding step size, a short-time Fourier transform is performed on the time-varying frequency signal, positive frequency components and negative frequency components in the transform result are identified during the short-time Fourier transform process, the positive frequency components are retained and the negative frequency components are deleted, thereby converting the time-varying frequency signal from a time domain representation into a two-dimensional time-frequency matrix represented in the time-frequency domain; The spectral line position is determined based on the two-dimensional time-frequency matrix. The ultra-high voltage high harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate of the time-varying frequency signal and the selected window length. The spectral line position is determined based on the two-dimensional time-frequency matrix, and the ultra-high voltage high harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate of the time-varying frequency signal, and the selected window length, including: Perform a peak search operation on each column of the two-dimensional time-frequency matrix to identify the spectral line index with the largest energy as the integer part of the main spectral line position; The elements in the two-dimensional time-frequency matrix are squared to enhance the peak contrast, and the spectral component with the local maximum amplitude, the spectral component with the local second maximum amplitude and the spectral component with the local third maximum amplitude are extracted therefrom; Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component and the local third largest amplitude spectral component, combined with the polynomial fitting parameters corresponding to the selected window function, the decimal offset of the main spectral line position is calculated; The ultra-high voltage high harmonic frequencies of the power system are determined based on the integer part and fractional offset of the main spectral line position, the sampling rate of the time-varying frequency signal, and the selected window length.

2. The method for measuring ultra-high voltage high harmonic frequency according to claim 1, wherein: According to the time-varying frequency signal characteristics of the power system, an adaptive window information design algorithm based on reinforcement learning is used to dynamically select the optimal window function type, window length, and sliding step size for the current power system operating state, including: Based on the time-varying frequency signal characteristics of the power system, an intelligent agent with perception capabilities is designed and deployed, and a set of window function types, window lengths, and sliding step sizes are defined. The intelligent agent can collect the power system frequency signal from the operating environment in real time. Construct a policy gradient network as the core decision-making unit of the intelligent agent; Construct an evaluation network, which receives spectrum information as input and evaluates the results of the agent's actions. Actions are defined as selecting different window function types, window lengths, and sliding steps. The following window information selection process is performed iteratively: The policy gradient network is used as the core decision-making unit of the intelligent agent. The intelligent agent takes the current high-order harmonic frequency distribution state of the power system as input, and uses the softmax function to perform sampling based on the high-order harmonic frequency distribution state to select an action from multiple optional actions. The agent processes the frequency signal of the power system according to the selected action and selects the corresponding window function type, window length and sliding step size; The evaluation network receives the results of the agent's actions and generates reward values ​​to feed back to the agent. The agent calculates the gradient and updates the policy gradient network based on the current state of the high-order harmonic frequency distribution, the selected action, and the resulting reward value; When the policy gradient network reaches a convergence state, the window function type, window length, and sliding step size corresponding to the currently selected action are determined as the optimal window function type, window length, and sliding step size for the current power system operation state; When the policy gradient network has not reached the convergence state, the next window information selection process is continued.

3. The method for measuring ultra-high voltage high harmonic frequency according to claim 2, wherein: The high-order harmonic frequency distribution state includes the frequency component distribution of the current signal, the amplitude change trend, and the time-frequency correlation information between adjacent moments.

4. The method for measuring ultra-high voltage high harmonic frequency according to claim 1, wherein: Based on the selected window function type, window length, and sliding step size, a short-time Fourier transform is performed on the time-varying frequency signal. During the short-time Fourier transform process, the positive and negative frequency components in the transform result are identified, the positive frequency components are retained, and the negative frequency components are deleted, thereby converting the time-varying frequency signal from the time domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain, including: A shiftable finite-length window function based on the selected window function type and window length is used to truncate the time-varying frequency signal in the time domain, and a Fourier transform is performed on the windowed time-varying frequency signal at each window position to obtain the corresponding local spectrum characteristics; During the short-time Fourier transform process, positive and negative frequency components in the transform result are identified, the positive frequency components are retained, and the negative frequency components are deleted; The window function is moved sequentially according to the selected sliding step size, and the time axis of the entire time-varying frequency signal is gradually traversed. The local spectral features under each window position are combined into a spectral result sequence, and a two-dimensional time-frequency matrix is ​​constructed. Each column of the two-dimensional time-frequency matrix represents the segmented spectral result sequence obtained after windowing the signal at different time positions.

5. The method for measuring ultra-high voltage high harmonic frequency according to claim 4, characterized in that: The local spectrum characteristics are obtained by the following formula: ； Indicates the time point at which the window function slides along the time axis, represents the angular frequency, Indicates at a point in time At, angular frequency The local spectrum characteristics under represents the time domain variable, represents a time-varying frequency signal, Represents the window function centered at time point τ, and the two-dimensional time-frequency matrix consists of all Combination composition.

6. The method for measuring ultra-high voltage high harmonic frequency according to claim 1, wherein: Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component and the local third largest amplitude spectral component, combined with the polynomial fitting parameters corresponding to the selected window function, the decimal offset of the main spectral line position is calculated, including: Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component, and the local third largest amplitude spectral component, the interpolation coefficient is calculated using the following formula: ,in, represents the interpolation coefficient, represents the spectral component with the local maximum amplitude, represents the local sub-largest spectral component, Represents the spectral component with the third largest local amplitude; Calculates the fractional offset of the main spectral line position based on the interpolation coefficients and the polynomial fit parameters corresponding to the selected window function.

7. The method for measuring ultra-high voltage high harmonic frequency according to claim 1, wherein: The ultra-high voltage high harmonic frequency of the power system is calculated by the following formula: ,in, Indicates the ultra-high voltage high harmonic frequency of the power system, Indicates the main line position, represents the integer part of the main spectral line position, Indicates the decimal offset of the main spectral line position, represents the sampling rate for time-varying frequency signals, Indicates the selected window length.

8. An ultra-high voltage high harmonic frequency measuring device, characterized in that: include: The receiving signal processing module is configured to extract an effective received sound wave signal containing useful components within a target bandwidth from an original received sound wave signal of a target receiver, where the target bandwidth is a frequency range of the sound wave signal emitted by the target sound source; a window information selection module configured to dynamically select the optimal window function type, window length, and sliding step size for the current power system operating state using an adaptive window information design algorithm based on reinforcement learning, based on the time-varying frequency signal characteristics of the power system; The frequency signal conversion module is configured to: perform a short-time Fourier transform on the time-varying frequency signal based on the selected window function type, window length, and sliding step size; identify positive frequency components and negative frequency components in the transformation result during the short-time Fourier transform process; retain the positive frequency components and delete the negative frequency components, thereby converting the time-varying frequency signal from a time-domain representation to a two-dimensional time-frequency matrix represented in the time-frequency domain; The harmonic frequency determination module is configured to: determine the spectral line position based on the two-dimensional time-frequency matrix, determine the ultra-high voltage high harmonic frequency of the power system based on the main spectral line position, the sampling rate of the time-varying frequency signal and the selected window length, The spectral line position is determined based on the two-dimensional time-frequency matrix, and the ultra-high voltage high harmonic frequency of the power system is determined based on the main spectral line position, the sampling rate of the time-varying frequency signal, and the selected window length, including: Perform a peak search operation on each column of the two-dimensional time-frequency matrix to identify the spectral line index with the largest energy as the integer part of the main spectral line position; The elements in the two-dimensional time-frequency matrix are squared to enhance the peak contrast, and the spectral component with the local maximum amplitude, the spectral component with the local second maximum amplitude and the spectral component with the local third maximum amplitude are extracted therefrom; Based on the local maximum amplitude spectral component, the local second largest amplitude spectral component and the local third largest amplitude spectral component, combined with the polynomial fitting parameters corresponding to the selected window function, the decimal offset of the main spectral line position is calculated; The ultra-high voltage high harmonic frequencies of the power system are determined based on the integer part and fractional offset of the main spectral line position, the sampling rate of the time-varying frequency signal, and the selected window length.

9. An electronic device comprising a processor and a memory storing program instructions, characterized in that: The processor is configured to execute the ultra-high voltage high harmonic frequency measurement method according to any one of claims 1 to 7 when running program instructions.

Images