An ultrahigh voltage alternating current transmission line audible noise probability prediction method and system

Abstract

The application belongs to the technical field of electromagnetic environment of power transmission line, and particularly relates to an audible noise probability prediction method and system for an ultra-high voltage alternating current power transmission line, comprising: obtaining field detection data of audible noise of the ultra-high voltage alternating current power transmission line; performing denoising processing on the obtained field detection data based on a wavelet threshold method to obtain denoised field detection data; and predicting audible noise probability of the ultra-high voltage alternating current power transmission line according to the denoised field detection data and a long short-term memory neural network prediction model. The audible noise probability is predicted through wavelet threshold denoising and a long short-term memory neural network, so that the credibility and accuracy of prediction are improved.

Description

Technical Field

[0001] This invention belongs to the field of electromagnetic environment technology for power transmission lines, specifically relating to a method and system for predicting the probability of audible noise in ultra-high voltage AC transmission lines. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In recent years, the electromagnetic environment of ultra-high voltage (UHV) transmission lines has increasingly attracted public attention. The electromagnetic environment problems of transmission lines are mainly caused by corona discharge, a form of discharge that occurs in a highly non-uniform field. Transmission lines have a small radius of curvature, and the conductor surface often has burrs and defects, resulting in a high degree of electric field non-uniformity near the conductor surface. Corona discharge occurs when the conductor voltage reaches a certain level. The corona discharge process generates electromagnetic environment problems such as radio interference and audible noise, which have become key technical challenges in UHV power transmission.

[0004] According to the inventor, current research on audible noise in ultra-high voltage transmission lines mainly relies on traditional detection methods, and the prediction of audible noise probability can only depend on traditional prediction formulas. This has the following shortcomings:

[0005] First, traditional testing methods are often affected by high wind speeds and background noise, which not only limits the ideal testing period but also fails to fully guarantee accuracy.

[0006] Secondly, most existing studies on audible noise prediction rely on empirical formulas, but these formulas are based on long-term observations of low-voltage lines and therefore contain errors in predicting audible noise for ultra-high-voltage lines. Summary of the Invention

[0007] To address the aforementioned issues, this invention proposes a method and system for predicting the probability of audible noise in ultra-high voltage AC transmission lines. By employing wavelet threshold denoising and a long short-term memory neural network, the probability prediction of audible noise is improved, thereby enhancing the reliability and accuracy of the prediction.

[0008] According to some embodiments, the first aspect of the present invention provides a method for predicting the probability of audible noise in ultra-high voltage AC transmission lines, employing the following technical solution:

[0009] A method for predicting the probability of audible noise in ultra-high voltage AC transmission lines, comprising:

[0010] Obtain on-site detection data of audible noise from ultra-high voltage AC transmission lines;

[0011] The acquired field detection data is denoised using the wavelet thresholding method to obtain denoised field detection data.

[0012] Based on the denoised on-site detection data and the long short-term memory neural network prediction model, the probability of audible noise in ultra-high voltage AC transmission lines is predicted.

[0013] Furthermore, during the denoising process, high-frequency noise signals in the acquired field detection data are processed by setting wavelet coefficient thresholds. Field detection data larger than the wavelet coefficient threshold are shrunk and retained, while field detection data smaller than the wavelet coefficient threshold are removed, thereby eliminating background noise in the acquired field detection data.

[0014] As a further technical limitation, the wavelet coefficient threshold function ω j,k for Where n is the wavelet decomposition level, s is the constraint constant, and λ is the threshold.

[0015] Furthermore, fuzzy tomography is used to optimize the parameters of the long short-term memory neural network prediction model.

[0016] Furthermore, the specific process of optimizing parameters using the fuzzy tomography method is as follows: construct a fuzzy judgment matrix, calculate the weight of each element in the constructed fuzzy judgment matrix using the summation and averaging method to obtain the element weight matrix; perform consistency verification on the constructed fuzzy judgment matrix, and when the consistency verification is satisfied, calculate the global weight of the obtained element weight matrix, sort the obtained global weights, and achieve parameter optimization.

[0017] As a further technical limitation, in the process of predicting the probability of audible noise in UHV AC transmission lines, a prediction interval is constructed based on the fluctuation range of the predicted audible noise values ​​and the nonparametric probability density estimation.

[0018] Furthermore, the process of constructing the prediction interval is as follows: the predicted point values ​​of the audible noise are divided into multiple segments at equal intervals; the prediction error of each segment is obtained, wherein the prediction error is the difference between the predicted value and the actual value; the probability density curve of the prediction error is determined by the kernel density estimation method; the upper and lower limits of the corresponding probability distribution are determined according to the probability density function curve of each segment; and the prediction interval is constructed by combining the point prediction values ​​at the corresponding time.

[0019] According to some embodiments, a second aspect of the present invention provides an audible noise probability prediction system for ultra-high voltage AC transmission lines, employing the following technical solution:

[0020] A system for predicting the probability of audible noise in ultra-high voltage AC transmission lines includes:

[0021] The acquisition module is configured to acquire on-site detection data of audible noise from UHV AC transmission lines;

[0022] The processing module is configured to denoise the acquired field detection data based on the wavelet thresholding method to obtain denoised field detection data.

[0023] The prediction module is configured to predict the probability of audible noise in UHV AC transmission lines based on denoised field detection data and a long short-term memory neural network prediction model.

[0024] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium, employing the following technical solution:

[0025] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for predicting the probability of audible noise in ultra-high voltage AC transmission lines as described in the first aspect of the present invention.

[0026] According to some embodiments, the fourth aspect of the present invention provides an electronic device, which adopts the following technical solution:

[0027] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the method for predicting the probability of audible noise in ultra-high voltage AC transmission lines as described in the first aspect of the present invention.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] This invention uses wavelet threshold denoising to optimize audible noise detection data of lines containing environmental background noise, removing the interference of background noise, making the detection data more accurate, and making the data prediction results more precise.

[0030] This invention uses a long short-term memory neural network model improved from a recurrent neural network to predict audible noise in ultra-high voltage AC transmission lines. It takes into account the influence of multiple factors and leverages the advantages of the recurrent neural network model, making the data prediction results more accurate.

[0031] This invention uses fuzzy hierarchical analysis to screen out the main factors affecting the audible noise of UHV AC transmission lines, which can more scientifically and rationally predict the audible noise of the lines.

[0032] This invention uses nonparametric probability density estimation, which can range the predicted values ​​and improve the reliability of the predicted values. Attached Figure Description

[0033] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0034] Figure 1 This is a flowchart of the method for predicting the probability of audible noise in ultra-high voltage AC transmission lines in Embodiment 1 of the present invention;

[0035] Figure 2 This is an overall prediction flowchart in Embodiment 1 of the present invention;

[0036] Figure 3 This is a schematic diagram of the RNN structure and its hidden layer units in Embodiment 1 of the present invention;

[0037] Figure 4 This is a schematic diagram of the long short-term memory neural network memory unit structure in Embodiment 1 of the present invention;

[0038] Figure 5 This is a structural block diagram of the audible noise probability prediction system for ultra-high voltage AC transmission lines in Embodiment 2 of the present invention. Detailed Implementation

[0039] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0040] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0041] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0042] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0043] Example 1

[0044] Embodiment 1 of this invention introduces a method for predicting the probability of audible noise in ultra-high voltage AC transmission lines.

[0045] like Figure 1 and Figure 2 The method for predicting the probability of audible noise in an ultra-high voltage AC transmission line, as shown, includes:

[0046] Obtain on-site detection data of audible noise from ultra-high voltage AC transmission lines;

[0047] The acquired field detection data is denoised using the wavelet thresholding method to obtain denoised field detection data.

[0048] Based on the denoised on-site detection data and the long short-term memory neural network prediction model, the probability of audible noise in ultra-high voltage AC transmission lines is predicted.

[0049] Audible noise background interference removal for ultra-high voltage AC transmission lines

[0050] The basic principle of wavelet thresholding denoising is to set a threshold to process high-frequency noise signals. Line audible noise signals with wavelet coefficients greater than this threshold are contracted and preserved; background noise with wavelet coefficients less than this threshold is removed. A one-dimensional signal model containing background noise can be represented as:

[0051] f(t)=s(t)+n(t) (1)

[0052] Where s(t) is the audible noise signal; n(t) is the ambient background noise with variance σ², following the N(0, σ²) rule. 2 The basic idea of ​​wavelet threshold denoising is as follows: ① Select wavelet basis functions to decompose the signal f(t) into multiple layers to obtain wavelet coefficients; ② Apply a certain threshold function to obtain new wavelet coefficients; ③ Reconstruct the wavelet to obtain the denoised signal.

[0053] After signal decomposition, the wavelet coefficients need to be thresholded. Commonly used thresholding functions include hard thresholding and soft thresholding. The mathematical expression for the hard thresholding function is:

[0054]

[0055] in, These are the estimated values ​​of the wavelet coefficients; ω j,k λ represents the wavelet coefficients of the j-th layer; λ is the threshold.

[0056] However, both soft and hard thresholding functions have inherent drawbacks: the hard thresholding function is discontinuous at ±λ, which can lead to a pseudo-Gibbs phenomenon, meaning that there will be vibrational effects during signal reconstruction; compared to the hard thresholding function, the soft thresholding function is continuous at ±λ, but its reconstruction accuracy is insufficient, resulting in a constant deviation. To improve the shortcomings of both soft and hard thresholding methods, and to make the thresholding function continuous at ±λ while improving reconstruction accuracy and reducing constant deviation, this patent selects an improved new thresholding function, the mathematical model of which is shown in equation (3):

[0057]

[0058] Where n is the wavelet decomposition level, and s is a constraint constant, whose value is controlled between 0 and 1; the soft and hard levels of the threshold function can be improved according to the value of s, and the threshold function can be flexibly processed with a variable threshold mode.

[0059] A higher number of signal decomposition levels is generally beneficial for signal-to-noise separation. However, for signal reconstruction, the reconstruction error also increases with the number of decomposition levels. Typically, the number of decomposition levels n for a signal of length M is:

[0060]

[0061] The wavelet threshold denoising effect is measured by the signal-to-noise ratio (SNR) and the root mean square error (RMSE).

[0062]

[0063]

[0064] Where f(i) is the audible noise signal detected by the original UHV AC transmission line; This refers to the audible noise signal after removing background noise. Generally speaking, the smaller the root mean square error and the higher the signal-to-noise ratio, the better the denoising effect.

[0065] Long Short-Term Memory Neural Network Prediction Model:

[0066] Recurrent Neural Networks (RNNs), compared to backpropagation (BP) neural networks and fully connected neural networks, possess memory capabilities. This is primarily achieved by utilizing their hidden layer units to pass information processed in previous time steps to the next time step for output computation. In other words, the current output of a sequence is related to its own input and also to the output of the previous sequence, thus enabling the consideration of the temporal correlation of audible noise from UHV AC transmission lines. The RNN structure and its hidden layer unit expansion are as follows: Figure 3 As shown in the figure, x is the input unit, h is the hidden unit, y is the output unit, the lower corner represents the state at each time step, and U, W, and V are the weight matrices between each layer.

[0067] In RNNs, the loop is manifested in the fact that, for time t: the performance of h(t) at time t is not only determined by the input x at that moment. t The decision is also subject to h t-1 The impact of this is that the parameter matrix W is a matrix that is shared cyclically throughout the time steps. Therefore, when the training time steps are large, W may tend to 0 or become infinitely large, causing gradient vanishing or exploding problems. Ultimately, the network cannot be trained and cannot achieve long-term memory, resulting in the short-term memory problem of RNNs.

[0068] When establishing a predictive model for audible noise of UHV AC transmission lines, since the audible noise data collected is a long-term data set, directly using an RNN model would lead to long-term dependence on the transmitted information, easily resulting in gradient vanishing or exploding, making it difficult to train the collected data well and achieving good predictive results.

[0069] Long-short-term memory neural networks (LSTMs) effectively address long-term dependencies in information, avoiding gradient vanishing or exploding. Compared to traditional RNNs, they structurally improve the memory units, such as... Figure 4 As shown, three gates are used to control the contents of the memory unit, namely input gate i t (Input Gate), Output Gate t (Output Gate), Forget Gate t (Forget Gate). The input gate controls how much information can be passed to the current unit, the forget gate controls how much information from the previous sequence is retained and discarded, and the output gate controls how much information from the current time step can be passed to the next unit. The calculations are as follows:

[0070] i t =σ(W i [h t-1 ,x t ]+b i (6)

[0071] f t =σ(W f [h t-1 ,x t ]+b f (7)

[0072] o t =σ(W0[h t-1 ,x t ]+b0) (8)

[0073] Where σ is the Sigmoid activation function; h t-1 The hidden layer state at time t-1; x t The input at time t; W i W f W and W0 are the weights of the input layer, hidden layer, and output layer, respectively; b i b f b0 and b0 are the biases of the input layer, hidden layer, and output layer, respectively.

[0074] The LSTM employs three gates to compute partial information retention and forgetting, allowing its memory units to store information over a longer period, thus mitigating the gradient explosion or vanishing problem. Research indicates that LSTM outperforms RNNs in multivariate time series prediction. Therefore, this embodiment uses an LSTM model trained on audible noise data collected from UHV AC transmission lines to predict the audible noise levels.

[0075] Model parameter optimization

[0076] Numerous factors influence the audible noise of UHV AC transmission lines. These factors can be broadly categorized into two types: the first is the influence of the line's structure, design, and construction itself; the second is the influence of external conditions such as the atmosphere and environment. The first type of influence primarily manifests in conductor parameters and arrangement, which can be controlled through design, and the audible noise level generally stabilizes after the line has been built and operated for a period of time. The second type of influence includes factors such as ground conductivity, air pressure, relative humidity, ultraviolet radiation intensity, wind speed, rainfall, and fog concentration, affecting the audible noise over a wider range. Before incorporating these characteristic factors into the LSTM model training, optimization can be performed to improve the neural network's prediction performance. Secondly, ranking the importance of each influencing factor is crucial. Using fuzzy hierarchical analysis to select parameters that significantly impact the target value greatly aids model training and prevents overfitting. For the already trained audible noise prediction model, combining the optimized parameters with the greatest impact on audible noise provides more accurate guidance for subsequent detection operations.

[0077] The analytic hierarchy process (AHP) based on fuzzy trigonometric functions can incorporate uncertain information, eliminate subjective errors to some extent, and construct a judgment matrix that better conforms to objective logic; specifically:

[0078] Step 1: Construct a fuzzy judgment matrix.

[0079] Based on expert experience, pairwise comparisons are performed on factors within the same layer to obtain a triangular fuzzy judgment matrix. Recorded as:

[0080]

[0081] Among them, (l ij ,m ij ,u ij ) represents the fuzzy quantization relationship between factors i and j, and the matrix is ​​used to represent this relationship. (l) ij ,m ij ,u ij ) and (l ji ,m ji ,uji They are reciprocals; ij u ij They are respectively (l ij ,m ij ,u ij The lower and upper bounds of the support for factors i and j, m ij For (l) ij ,m ij ,u ij The median value supported by factors i and j, and l ij ≤m ij ≤u ij .

[0082] Step 2: Weight calculation.

[0083] The weight of each element in the fuzzy judgment matrix is ​​calculated using the summation and averaging method, resulting in the element weight matrix H. i as follows:

[0084]

[0085] To eliminate the influence of expert subjectivity on the judgment matrix, an adjustment coefficient θ is introduced. θ > 0.5 indicates that the judge has a pessimistic personality or is making judgments with negative emotions; θ < 0.5 indicates that the judge is positive. The expected value E(H) is calculated. i After normalization, the generated weight vector is:

[0086]

[0087]

[0088] Among them, l i m i u i They are l ij m ij u ij The normalized value; E is the factor weight matrix H i Expected value; E L The factor weight matrix H i In the interval [l i m i The average expected value within ]; E R The factor weight matrix H i In the interval [m i u i The average expected value within ]; w i This is the factor weight vector.

[0089] Step 3: Consistency check.

[0090] When determining the diagonal elements of matrix A, i.e., i = j, it is certain that lij =l ji m ij =m ji and u ij =u ji When i ≠ j, we have l ji ≤m ji ≤u ji The weights of each element in A must satisfy the following condition: ji ≤(w i / w j )≤u ji Introducing matrices intermediate quantity m ji Calculate the ratio function G of the weights. ji (w i / w j Define the consistency check index for matrix A:

[0091]

[0092]

[0093] Where, when γ>e –1 =0.3679 indicates that the factor weights meet the consistency check; γ=1 indicates that the evaluation matrix is ​​completely consistent. The larger the value of γ, the better the consistency. γ refers to the fuzzy matrix. Consistency verification metrics.

[0094] Step 4: Global Weights of Factors When the judgment matrices of each factor layer satisfy the consistency check, calculate the global weights of the sub-factors. Let... The weights of the first-level criteria layer to the target layer; The weights of the first-level criterion layer to its sub-criterion layers are then represented by the global weights. for:

[0095]

[0096] The weights of the calculated influencing factors are sorted, and the factors with the greatest influence are selected.

[0097] Interval forecasting

[0098] The audible noise of ultra-high voltage AC transmission lines exhibits strong randomness and volatility, which means that even the most accurate prediction models will contain errors. Deterministic prediction methods do not comprehensively describe the magnitude and regularity of audible noise. Therefore, in addition to obtaining the predicted audible noise value, it is also necessary to determine the fluctuation range of that predicted value and construct a prediction interval. Interval prediction of audible noise uses a probability interval to represent the difference between the actual and predicted values ​​of the audible noise, i.e., the prediction error. This error is then combined with the already obtained point prediction values ​​of the audible noise to obtain the prediction interval.

[0099] Based on the probability distribution of the predicted value and prediction error of audible noise from UHV AC transmission lines, a prediction interval for audible noise is constructed. Let the predicted value be P, and the probability distribution function of the prediction error be F(·). Then, the prediction interval at a confidence level of 1-α is expressed as:

[0100] [P+F -1 (α1),P+F -1 (α2)] (15)

[0101] Where α1 and α2 are the upper and lower limits of the probability distribution, respectively, α1 = α / 2, α2 = 1 – α / 2.

[0102] Kernel density estimation (KDE) is a nonparametric estimation method. KDE uses each data point and bandwidth as parameters for a kernel function, resulting in several kernel functions. These kernel functions are then superimposed to obtain the kernel density estimation function. After normalization, the probability density function of the kernel density is obtained, which is used to determine the probability distribution of the prediction error. The probability density function obtained using KDE is:

[0103]

[0104] Where N is the number of samples; h is the window width; K(·) is the kernel function; x i Let be the sample value of the i-th prediction error. The kernel function chosen is the Epanechnikov function, with the following formula:

[0105]

[0106] in, This is for estimating the probability density of the sample.

[0107] Based on the above results, the prediction interval is constructed using the following steps:

[0108] Step 1: Divide the predicted point values ​​of the audible noise into multiple segments at equal intervals;

[0109] Step 2: Obtain the prediction error for each segment, i.e., the difference e between the predicted value and the actual value. i =y i -o i The probability density curve of the prediction error is determined using the kernel density estimation method.

[0110] Step 3: Determine the corresponding α1 and α2 point values ​​according to the probability density function curve of each segment;

[0111] Step 4: Calculate the α / 2 and 1-α / 2 quantiles of the corresponding curve, and construct the prediction interval by combining the predicted values ​​of the points at the corresponding times.

[0112] This embodiment employs wavelet thresholding to remove environmental background noise interference from the audible noise detection data of UHV AC transmission lines, making the prediction results more accurate. A long short-term memory neural network model is used to train and predict the audible noise detection data of UHV AC transmission lines. Fuzzy hierarchical analysis is used to screen out the main influencing factors of audible noise in UHV AC transmission lines, enabling a more scientific and reasonable prediction of the line's audible noise. Based on nonparametric probability density estimation, the predicted values ​​are intervalized to improve the reliability of the predicted values.

[0113] Example 2

[0114] Embodiment 2 of the present invention introduces an audible noise probability prediction system for ultra-high voltage AC transmission lines.

[0115] like Figure 5 The audible noise probability prediction system for ultra-high voltage AC transmission lines shown includes:

[0116] The acquisition module is configured to acquire on-site detection data of audible noise from UHV AC transmission lines;

[0117] The processing module is configured to denoise the acquired field detection data based on the wavelet thresholding method to obtain denoised field detection data.

[0118] The prediction module is configured to predict the probability of audible noise in UHV AC transmission lines based on denoised field detection data and a long short-term memory neural network prediction model.

[0119] The detailed steps are the same as those of the UHV AC transmission line audible noise probability prediction method provided in Example 1, and will not be repeated here.

[0120] Example 3

[0121] Embodiment 3 of the present invention provides a computer-readable storage medium.

[0122] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for predicting the probability of audible noise in ultra-high voltage AC transmission lines as described in Embodiment 1 of the present invention.

[0123] The detailed steps are the same as those of the UHV AC transmission line audible noise probability prediction method provided in Example 1, and will not be repeated here.

[0124] Example 4

[0125] Embodiment 4 of the present invention provides an electronic device.

[0126] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the method for predicting the probability of audible noise in ultra-high voltage AC transmission lines as described in Embodiment 1 of the present invention.

[0127] The detailed steps are the same as those of the UHV AC transmission line audible noise probability prediction method provided in Example 1, and will not be repeated here.

[0128] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A method of audible noise probability prediction for an extra-high voltage AC power transmission line, characterized in that, The method comprises the following steps: obtaining field detection data of audible noise of an ultra-high voltage AC transmission line; performing denoising processing on the obtained field detection data based on a wavelet threshold method to obtain denoised field detection data; in the process of denoising processing, setting a wavelet coefficient threshold to process high-frequency noise signals of the obtained field detection data, performing shrinkage and reservation processing on field detection data greater than the wavelet coefficient threshold, and eliminating field detection data less than the wavelet coefficient threshold to remove background noise of the obtained field detection data; The wavelet coefficient threshold function For ; wherein, is a wavelet decomposition level, is a constraint constant, is a threshold threshold; predicting the audible noise probability of the ultra-high voltage AC transmission line according to the denoised field detection data and a long short-term memory neural network prediction model.

2. A method of audible noise probability prediction for a UHV AC power transmission line as claimed in claim 1, characterized in that, The long short-term memory neural network prediction model parameters are optimized by using a fuzzy chromatography analysis method.

3. A method of audible noise probability prediction for a UHV AC power transmission line as claimed in claim 2, characterized in that, The specific process of optimizing the parameters by using the fuzzy chromatography analysis method is as follows: a fuzzy judgment matrix is constructed, the proportion of each element in the constructed fuzzy judgment matrix is calculated by using a summation average method to obtain an element weight matrix; the constructed fuzzy judgment matrix is subjected to consistency checking, when the consistency checking is satisfied, the global weight of the obtained element weight matrix is calculated, the obtained global weight is sorted, and the optimization of the parameters is realized.

4. A method of audible noise probability prediction for a UHV AC power transmission line as defined in claim 1, characterized in that, In the process of predicting the audible noise probability of the ultra-high voltage AC transmission line, a prediction interval is constructed according to the fluctuation range of the predicted value of the audible noise and a non-parametric probability density estimation.

5. A method of audible noise probability prediction for a UHV AC power transmission line as defined in claim 4, characterized in that, The process of constructing the prediction interval is as follows: the point prediction value of the audible noise obtained by prediction is equally divided into multiple sections; the prediction error of each section is obtained, the prediction error being the difference between the predicted value and the actual value; the probability density curve of the prediction error is determined by using a kernel density estimation method; the upper limit and the lower limit in the corresponding probability distribution are determined according to the probability density function curve of each section; and the prediction interval is constructed in combination with the point prediction value at the corresponding time.

6. An audible noise probability prediction system for an extra-high voltage AC power transmission line, characterized by The method comprises the following steps: an obtaining module configured to obtain field detection data of audible noise of an ultra-high voltage AC transmission line; a processing module configured to perform denoising processing on the obtained field detection data based on a wavelet threshold method to obtain denoised field detection data; in the process of denoising processing, setting a wavelet coefficient threshold to process high-frequency noise signals of the obtained field detection data, performing shrinkage and reservation processing on field detection data greater than the wavelet coefficient threshold, and eliminating field detection data less than the wavelet coefficient threshold to remove background noise of the obtained field detection data; The wavelet coefficient threshold function For ; wherein, is a wavelet decomposition level, is a constraint constant, is a threshold threshold; a prediction module configured to predict the audible noise probability of the ultra-high voltage AC transmission line according to the denoised field detection data and a long short-term memory neural network prediction model.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to realize the steps of the method for predicting the audible noise probability of the ultra-high voltage AC transmission line according to any one of claims 1-5.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the steps of the method for predicting the audible noise probability of the ultra-high voltage AC transmission line according to any one of claims 1-5.

Images