Instantaneous interference elimination system and method based on artificial intelligence modeling for power line network

Abstract

The invention relates to the technical field of power line communication, discloses an instantaneous interference elimination system and method based on artificial intelligence modeling, and is used for solving the problem of instantaneous interference caused by plugging of electrical equipment. The system comprises an instantaneous interference detector, an artificial intelligence database and an instantaneous interference canceller. The detector divides signals through a power divider, carries out delay processing on one path, stores samples after analog-to-digital conversion on the other path, dynamically compares a sampling value with a threshold value, judges interference and preliminarily recovers interference samples. A database adopts a least square support vector machine (LSSVM) and echo state network (ESN) double-level model, the LSSVM generates a linear prediction result through a radial basis kernel function, and the ESN corrects a nonlinear error by using a reserve pool and outputs a high-precision interference waveform. The canceller is matched with a database template to generate a reverse waveform (time sequence deviation is less than or equal to 1ns), and interference attenuation is greater than or equal to 20dB after superposition. The system supports dynamic parameter adjustment, such as delay adaptation, threshold value sliding window update and ESN incremental learning. Through testing, the interference detection rate is greater than or equal to 98%, the signal-to-noise ratio is increased by greater than or equal to 15dB, and the error rate is reduced to 10 <-6 >. According to the invention, through accurate modeling and real-time offset, the reliability of power line communication is significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power line communication (PLC), and particularly to a method for eliminating instantaneous interference caused by the insertion, removal, or on / off state switching of electrical equipment in a power line network. Background Art

[0002] Power line communication is a key technology for constructing a smart grid, which uses existing low-voltage power lines for information transmission. During the power line communication process, it is often affected by a large number of noise and interference signals. Among them, the instantaneous interference caused by the insertion, removal, or on / off state switching of electrical equipment is particularly prominent. These interference signals carry a large amount of instantaneous energy, posing a serious threat to the quality of power line communication.

[0003] To address this problem, currently, the adopted solutions include, for example, improving the topology of the power line network to reduce the impact of the insertion and removal of electrical equipment on the overall network; or using signal processing technologies such as filtering and noise reduction to weaken the intensity of instantaneous interference. However, these solutions are either highly complex to implement or have limited effects, and it is difficult to meet the growing demand for power line communication. Summary of the Invention

[0004] To overcome the deficiencies of the above-mentioned prior art, the present invention provides an instantaneous interference detection and elimination device based on artificial intelligence modeling for eliminating instantaneous interference caused by the insertion, removal, or on / off state switching of electrical equipment in a power line communication network. The device is located at the front end of the power line receiving communication device, and can capture and eliminate instantaneous interference to avoid its impact on the quality of power line communication.

[0005] During operation, the device can detect instantaneous interference in real time and preliminarily recover the instantaneous interference from the received signal. By comparing with the instantaneous interference template in the interference library, the closest instantaneous interference template is found, and a value opposite to it is generated, so as to cancel the instantaneous interference in the received signal.

[0006] The most critical design of the instantaneous interference elimination device includes a power splitter, an instantaneous interference detector, an artificial intelligence-based instantaneous interference database, and an instantaneous interference eliminator.

[0007] Power splitter: First, the received signal is sent to the power splitter, which divides the received signal into two signals with equal power. One of the signals is subjected to delay processing, and the other signal is subjected to analog-to-digital (A / D) conversion, and the sampled samples of the received signal after this A / D conversion are stored. Then, within the transmission time of a certain symbol, when some sampled values exceed a set threshold (which can be set by the waveform of the transmitted signal), it is considered that there is instantaneous interference in the received signal.

[0008] Delay device: Delays the signal output from the power divider to make it consistent in time with the signal from the other branch.

[0009] Instantaneous interference detector: First, the standard transmission signal is subtracted from the received signal (for example, when transmitting BPSK modulated symbols, there are two different transmission waveform templates corresponding to the two symbols 0 and 1, and one of them can be selected at will, because the waveform amplitude of the signal template is much smaller than the order of magnitude of the instantaneous interference), and the sample value of the initial judgment of the instantaneous interference is obtained. The initial judgment instantaneous interference samples are stored and sent to the instantaneous interference database based on artificial intelligence to obtain the accurate final instantaneous interference.

[0010] Artificial intelligence-based instantaneous interference database: This library uses the data output by the instantaneous interference detector as input and outputs the final instantaneous interference through an artificial intelligence algorithm.

[0011] The training data of this database mainly includes two categories: ① The sampling values ​​of historical instantaneous interference generated by the plugging or unplugging operations of a large number of electrical devices at different distances from these devices; ② The sampling values ​​of historical instantaneous interference generated by the plugging or unplugging operations of a single electrical device at different distances from the device.

[0012] The double-level comprehensive prediction model of least squares support vector machine and echo state network is used for the above training library to improve the accuracy of the restored instantaneous interference. The principle is:

[0013] (1) Least square support vector machine (LSSVM)

[0014] LSSVM is an extension of support vector machine. This method replaces the inequality constraints in SVM with equality constraints, so that the solution of the original quadratic programming problem is converted into the solution of a linear equation system. The optimization problem can be expressed as

[0015]

[0016] Formula (1) can be solved according to the Lagrange multiplier method and KKT conditions.

[0017] (2) Echo State Network (ESN)

[0018] ESN is a new type of recursive neural network suitable for the prediction of nonlinear time series. Its network structure is as follows: Figure 2 As shown in the figure, u(n), y(n) and x(n) represent input samples, output samples and internal state vectors respectively; W in , W intr , W out , W backThey respectively represent the input weight matrix, the internal weight matrix, the output weight matrix, and the output feedback matrix.

[0019] The ESN uses a large-scale coefficient network to establish the mapping from low-dimensional input samples to a high-dimensional state space, and uses the linear regression method to train and obtain the output connection weights, which simplifies the network training. The update formulas for its state vector and output vector are

[0020] x(n + 1) = f(W in u(n + 1) + W intr x(n) + W back y(n)) (2)

[0021] y(n + 1) = f out (W out [u(n + 1), x(n + 1)]) (3)

[0022] In the formulas: f and f out are respectively the activation functions of the reservoir and the output layer.

[0023] Based on the data in the instantaneous interference database and the initially judged instantaneous interference sample values, the least squares support vector machine is used to predict the final instantaneous interference, and the result is denoted as P 1 , and the process of predicting the final instantaneous interference is as follows:

[0024] 1) Establish a least squares support vector machine model, map the obtained training samples to a high-dimensional kernel space, use the radial basis function as the kernel function, and use the Lagrange multiplier method to determine the optimal parameters of the least squares support vector machine model, thereby obtaining the least squares support vector machine model;

[0025]

[0026] κ(x, x i ) = φ(x) T φ(x i ) = exp(-||x - x i || 2 / 2σ 2 )

[0027] In the formulas, φ(x) is the mapping function that maps x to a high-dimensional kernel space; b is the least squares vector machine model parameter; α i is the Lagrange multiplier; c is the regularization parameter; ω is the weight vector; y i is the i-th output vector of the training sample; ε i is the slack variable; κ(x, x i ) is the kernel function; σ is the kernel width, and its value is

[0028] 2) Predict the final instantaneous interference: Input the initially judged instantaneous interference sample value S(m) = (s 1 (m), …, s N (m)) T into the least squares support vector machine model to obtain the predicted value P 1 of the final instantaneous interference. The calculation formula for the predicted value P 1 of the final instantaneous interference is as follows:

[0029]

[0030] Based on the initially judged instantaneous interference sample value data, use the echo state network to predict the final instantaneous interference, and the result is denoted as P 2 . The process of predicting the final instantaneous interference is as follows:

[0031] 1) Initialize the parameters of the echo state network: Randomly determine the dimension K of the reservoir nodes, and randomly generate a reservoir state matrix W with a spectral radius less than 1 K×K , input matrix and output feedback weight matrix . Among them, once the reservoir state matrix W K ×K , input matrix and output feedback weight matrix are determined, they remain unchanged during the training and testing phases;

[0032] 2) Based on the training sample database, input the training samples into the echo state network to calculate and record the reservoir state U(n) = (u 1 (n), …, u K (n)) T of the echo state network. The calculation formula for the reservoir state U of the echo state network is:

[0033]

[0034] In the formula, f is the activation function of the reservoir unit, and it is selected as the hyperbolic tangent function;

[0035] f(*) = tanh(*);

[0036] 3) Calculate the output weight matrix according to the reservoir state U and the output vector Y. The calculation formula for the output weight matrix of the echo state network is:

[0037]

[0038] In the formula, is the pseudo-inverse of [X(n), U(n), Y(n - 1)];

[0039] 4) Predict the final instantaneous interference. Based on the trained echo state network and the initially judged instantaneous interference sample values, predict the final instantaneous interference P 2 ; The predicted result P of the final instantaneous interference 2 is:

[0040]

[0041] V 2 (i) = W out [S(i), U(n + i), V 2 (i - 1)] i = 1, …, m.

[0042] Integrate the prediction result P of the least squares support vector machine 1 and the prediction result P of the echo state network 2 , and obtain the final prediction result P, which is the prediction model; The prediction model is:

[0043] P = 0.5P 1 + 0.5P 2 .

[0044] Based on the above training library, according to the initially judged instantaneous interference input, the final instantaneous interference will be output.

[0045] Instantaneous interference eliminator: Generate the opposite value of the discrete value of the waveform of the template that is closest to the detected instantaneous interference output by the instantaneous interference library data within the duration of the instantaneous interference.

[0046] Adder: Add the sampled values of the received signal and the opposite value (negative value) of the discrete value of the final instantaneous interference waveform correspondingly to cancel the instantaneous interference.

[0047] After this device, the reception can be carried out according to the normal power line reception steps.

[0048] The present invention aims at the instantaneous interference caused by the plugging and unplugging of electrical equipment, and recovers and eliminates it in two steps to avoid its interference with the transmitted information. The present invention uses artificial intelligence means to train the instantaneous interference generated under different numbers of electrical equipment and at different distances from the electrical equipment, and establishes an accurate instantaneous interference template library. This library can accurately restore the instantaneous interference in the received signal, so as to eliminate it from the received signal. The present invention effectively improves the influence of the instantaneous interference caused by the plugging and unplugging of electrical equipment in the power line network by establishing an accurate instantaneous interference template library and accurately restoring the instantaneous interference signal, and improves the transmission reliability of the power line communication system. Description of the Drawings

[0049] Figure 1Block diagram of a system for eliminating instantaneous interference based on artificial intelligence modeling for a power line network according to the present invention;

[0050] Figure 2 Block diagram of the ESN network structure adopted by the instantaneous interference library; Specific implementation manners

[0051] The technical solution of the present invention will be further limited below in conjunction with the accompanying drawings and embodiments, but the protection scope of the present invention should not be limited thereby.

[0052] The instantaneous interference elimination system of this embodiment is deployed in the front section of the power line receiving device, such as Figure 1 shown, and includes an instantaneous interference detector, an instantaneous interference database based on artificial intelligence, and an instantaneous interference eliminator.

[0053] The instantaneous interference detector is used to detect instantaneous interference in the power line received signal in real time;

[0054] The instantaneous interference data module is connected to the instantaneous interference detector and is used to generate a final instantaneous interference template according to the initially judged instantaneous interference samples, and it adopts a two - level comprehensive prediction model of least - squares support vector machine (LSSVM) and echo state network (ESN);

[0055] The instantaneous interference eliminator is connected to the instantaneous interference database and is used to generate a reverse waveform according to the final instantaneous interference template and superimpose it with the received signal to cancel the instantaneous interference.

[0056] Among them, the working process of the instantaneous interference detector is as follows:

[0057] Signal segmentation: The received signal is first sent to a power divider, and the power divider evenly divides it into two signals with equal power.

[0058] Signal processing: One of the signals is subjected to delay processing, and the other signal is subjected to analog - to - digital (A / D) conversion to convert the analog signal into a digital signal. The converted digital signal is sampled and stored to form a sampling sample set.

[0059] Interference detection: During the transmission of the signal, the instantaneous interference detector checks the sampling values in real time. When some sampling values exceed a preset threshold within the transmission time of a certain symbol, the instantaneous interference detector determines that there is instantaneous interference in the received signal. This threshold is usually set according to the waveform characteristics of the transmitted signal.

[0060] Initial Judgment and Storage: Once a transient interference is detected, the transient interference detector subtracts the received signal from the standard transmission signal (such as one of the waveform templates corresponding to the 0 and 1 symbols modulated by BPSK) to obtain the sampled values of the transient interference for initial judgment. These initially judged interference samples are stored and sent to an artificial intelligence-based transient interference database for further analysis.

[0061] The artificial intelligence-based transient interference database receives the output data from the transient interference detector as input and processes it through artificial intelligence algorithms to output the final transient interference waveform. The training data of this database mainly comes from two categories: ① Sampled values of historical transient interferences generated when plugging or unplugging a large number of electrical devices at different positions from these devices; ② Sampled values of historical transient interferences generated when plugging or unplugging a single electrical device at different positions from this device. To improve the accuracy of transient interference recovery, this database adopts a two-level comprehensive prediction model of the Least Square Support Vector Machine (LSSVM) and the Echo State Network (ESN).

[0062] LSSVM is an extension of the Support Vector Machine (SVM). It replaces the inequality constraint in SVM with an equality constraint, converting the original quadratic programming problem into the solution of a system of linear equations. Its optimization problem can be expressed as

[0063]

[0064] Equation (1) can be solved according to the Lagrange multiplier method and the KKT conditions.

[0065] ESN is a new type of recurrent neural network suitable for the prediction of non-linear time series. Its network structure includes an input layer, a reservoir (also known as the internal state vector layer), and an output layer. ESN uses a large-scale coefficient network to establish the mapping from low-dimensional input samples to high-dimensional state space and obtains the output connection weights through linear regression training. This method simplifies the network training process while maintaining good prediction performance. In ESN, the input samples enter the reservoir through the input weight matrix, multiply with the internal weight matrix to generate the internal state vector. Then, the internal state vector obtains the output vector through the output weight matrix and the output feedback matrix. The activation functions of the reservoir and the output layer are used to introduce non-linear characteristics to enhance the prediction ability of the network. The update formulas for the state vector and the output vector are

[0066] x(n + 1) = f(W in u(n + 1) + W intr x(n) + W back y(n)) (2)

[0067] y(n + 1) = f out (W out [u(n + 1), x(n + 1)]) (3)

[0068] In the formula: f and f out are the activation functions of the reservoir and the output layer respectively, u(n), y(n) and x(n) represent the input sample, output sample and internal state vector respectively; W in , W intr , W out , W back represent the input weight matrix, internal weight matrix, output weight matrix and output feedback matrix respectively.

[0069] Instantaneous interference eliminator: The instantaneous interference template that is closest to the detected instantaneous interference output from the instantaneous interference library data generates sampling values with waveforms opposite to that template within the duration of the instantaneous interference, and adds them corresponding to each sampling value of the received signal to cancel the instantaneous interference.

Claims

1. A transient interference elimination system based on artificial intelligence modeling for power line networks, characterized in that: include: The instantaneous interference detector is used to divide the received signal into two signals of equal power, perform delay processing on one of the signals, perform analog-to-digital conversion on the other signal and store the sampled samples; during the transmission of the signal, when the sampled value exceeds the preset threshold value, it is determined that there is instantaneous interference, and the sampled value of the instantaneous interference is preliminarily restored from the received signal; An instantaneous interference database based on artificial intelligence is used to receive the instantaneous interference samples initially judged by the instantaneous interference detector, and process the instantaneous interference samples through a double-level comprehensive prediction model of a least squares support vector machine (LSSVM) and an echo state network (ESN) to improve the accuracy of instantaneous interference recovery and output a final instantaneous interference waveform; The instantaneous interference canceller is used to receive the final instantaneous interference from the instantaneous interference database, generate sampling values ​​opposite to the instantaneous interference waveform within the duration of the instantaneous interference, and add the sampling values ​​of the opposite waveform to the corresponding sampling values ​​of the received signal, thereby canceling the instantaneous interference.

2. The instantaneous interference elimination device according to claim 1, characterized in that: The instantaneous interference detector comprises: A power divider is used to evenly divide the received signal into two signals; A delay processor, used for delaying one of the signals; An analog-to-digital converter, used for performing analog-to-digital conversion on another signal; A storage unit, used for storing the sampled samples after analog-to-digital conversion; A comparator, used to compare the sampled value with a preset threshold value to determine whether there is instantaneous interference; The adder is used to subtract the standard transmission signal from the received signal to obtain a preliminary judged instantaneous interference sampling value.

3. The instantaneous interference elimination device according to claim 1, characterized in that: In the two-level comprehensive prediction model, LSSVM is used to generate preliminary prediction results by solving a linear equation group; ESN corrects the nonlinear time series through a recursive neural network and outputs the final instantaneous interference waveform.

4. The instantaneous interference elimination device according to claim 1, characterized in that: The instantaneous interference canceller comprises: A template matching unit, used for searching the template closest to the detected instantaneous interference in the instantaneous interference database; A waveform generating unit, used for generating sampling values ​​of an opposite waveform according to the matched template; The adding unit is used to add the sampling values ​​of the opposite waveform to the sampling values ​​of the received signal accordingly, so as to offset the instantaneous interference.

5. The system according to claim 1, characterized in that The training data of the instantaneous interference database based on artificial intelligence includes: (a) Historical instantaneous interference samples generated by plugging and unplugging multiple electrical devices at different locations; (b) Historical instantaneous interference samples generated by plugging and unplugging a single electrical device at different locations.

6. A method for eliminating instantaneous interference in a power line network, characterized in that: The following steps are involved: Step S1: Split the received signal into two paths, perform delay processing on one path, and perform analog-to-digital conversion and storage on the other path; Step S2: Detect whether the sampling value exceeds the preset threshold value. If so, it is determined that there is instantaneous interference, and the instantaneous interference sample is preliminarily restored; Step S3: predict the sample through the LSSVM and ESN two-level model and output the final instantaneous interference waveform; Step S4: Generate a sampling value that is inverse to the final waveform and superimpose it with the received signal to eliminate instantaneous interference.

7. The method according to claim 6, characterized in that In step S3, LSSVM maps samples to a high-dimensional space through a radial basis kernel function, and corrects nonlinear time series errors in combination with a recursive neural network of ESN.

8. The method according to claim 6, characterized in that In step S4, the generation of the reverse sampling value includes: (a) Match the closest instantaneous interference template in the database; (b) Generate reverse waveform sampling values ​​with opposite phases and equal amplitudes according to the template.

Images