Power line carrier channel noise prediction method and system considering type matching bias

Abstract

The application provides a power line carrier channel noise prediction method and system considering type matching deviation, and the method comprises the following steps: collecting original noise data of power line carrier channel noise; calculating the matching similarity between the original noise data and each noise type respectively, and determining the noise type; calculating the matching deviation of the corresponding noise type; inputting the original noise data into a pre-trained noise prediction model, and realizing the prediction of the power line carrier channel noise based on the output of the model; wherein the model is trained by a preset wavelet neural network based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation. Compared with the prior art, by identifying the type of the power line carrier channel noise, training the preset wavelet network model based on the matching deviation, realizing the prediction of the power line carrier channel noise, the type of the noise and the prediction model can be adapted, and the prediction accuracy and precision are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of noise identification, and in particular to a method and system for predicting power line carrier channel noise considering type matching deviation. Background Art

[0002] As the construction of new power systems continues to increase the demand for communications, power line carrier communication (PLC) has become an important means of achieving high-speed data transmission. PLC offers advantages such as wide coverage, security, and reliability. However, the increasing integration of renewable energy sources can generate complex noise from a large number of power electronic devices, impacting the quality of PLC.

[0003] Currently, existing technologies primarily observe PLC channel noise characteristics by establishing PLC channel noise models. However, these traditional modeling methods fail to consider noise classification, resulting in a mismatch between noise type and PLC channel noise prediction models, impacting noise prediction accuracy. Furthermore, they fail to consider the impact of noise type mismatch on PLC channel noise prediction, further impacting prediction accuracy. Summary of the Invention

[0004] The present invention provides a power line carrier channel noise prediction method and system taking into account type matching deviation, so as to solve the technical problem of how to improve the prediction precision and accuracy.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a power line carrier channel noise prediction method considering type matching deviation, including:

[0006] Collecting raw noise data of power line carrier channel noise;

[0007] Calculating the matching similarity between the original noise data and the historical data of each noise type respectively, and taking the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; calculating the matching deviation between the original noise data and the historical data of the corresponding noise type; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise and random pulse noise;

[0008] Inputting the original noise data into a pre-trained noise prediction model, and predicting the power line carrier channel noise based on the output of the noise prediction model;

[0009] The noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the historical data of the corresponding noise type, and is trained through a preset wavelet neural network.

[0010] As a preferred solution, the training process of the noise prediction model includes:

[0011] Obtaining a data training set; wherein the data training set includes noise samples and corresponding expected outputs;

[0012] Inputting the noise sample into the wavelet neural network to obtain the actual output of the network;

[0013] Calculating a wavelet neural network prediction error based on the actual output of the network and the expected output;

[0014] Based on the wavelet neural network prediction error and the matching deviation, the network weights and network thresholds of the wavelet neural network are iteratively updated until the prediction error is less than a preset error accuracy, thereby obtaining the trained noise prediction model.

[0015] As a preferred solution, the network weights and network thresholds of the wavelet neural network are specifically:

[0016]

[0017]

[0018] Among them, μ i (t) is the learning rate parameter of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔW i (t) is the weight update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔB i (t) is the threshold update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, B i (t) is the wavelet neural network threshold matrix of the t-th time slot, W i (t) is the weight matrix of the wavelet neural network in the t-th time slot, e i (t) is the wavelet neural network prediction error adapted to the i-th noise type in the t-th time slot.

[0019] As a preferred solution, the actual output of the network is specifically:

[0020]

[0021] Among them, y i,n (t) is the nth output of the wavelet neural network adapted to the i-th noise type in the t-th time slot, x k,i (t) is the kth neuron input of the wavelet neural network input layer adapted to the i-th noise type in the t-th time slot, w ks,i(t) is the weight between the input layer neurons and the hidden layer neurons in the tth time slot, b s,i (t) is the hidden layer threshold of the t-th time slot, w sn,i (t) is the weight between the hidden layer neurons and the output layer neurons in the tth time slot, b n,i (t) is the output layer threshold of the t-th time slot, a s,i (t) is the scaling factor of the hidden layer in the t-th time slot, and f is the wavelet basis function.

[0022] As a preferred solution, the calculation of the matching similarity between the original noise data and the historical data of each noise type is specifically as follows:

[0023] Calculate according to the following formula:

[0024]

[0025] Among them, D i (t) is the matching similarity between the original noise data of the t-th time slot and the i-th noise type, e i (t-1) is the historical prediction error of the wavelet neural network adapted to the i-th noise type, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, α i (t) is the weight parameter of the matching deviation of the i-th noise type in the t-th time slot, β i (t) is the weight parameter of the historical prediction error of the wavelet neural network adapted to the i-th noise type in the t-th time slot.

[0026] Accordingly, an embodiment of the present invention further provides a power line carrier channel noise prediction system considering type matching deviation, comprising an acquisition module, a noise type identification module and a prediction module; wherein,

[0027] The acquisition module is used to collect original noise data of the power line carrier channel noise;

[0028] The noise type identification module is configured to respectively calculate the matching similarity between the original noise data and the historical data of each noise type, and use the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; calculate the matching deviation between the original noise data and the historical data of the corresponding noise type; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise, and random pulse noise;

[0029] The prediction module is configured to input the original noise data into a pre-trained noise prediction model, and predict the power line carrier channel noise based on the output of the noise prediction model;

[0030] The noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the historical data of the corresponding noise type, and is trained through a preset wavelet neural network.

[0031] As a preferred solution, the training process of the noise prediction model includes:

[0032] Obtaining a data training set; wherein the data training set includes noise samples and corresponding expected outputs;

[0033] Inputting the noise sample into the wavelet neural network to obtain the actual output of the network;

[0034] Calculating a wavelet neural network prediction error based on the actual output of the network and the expected output;

[0035] Based on the wavelet neural network prediction error and the matching deviation, the network weights and network thresholds of the wavelet neural network are iteratively updated until the prediction error is less than a preset error accuracy, thereby obtaining the trained noise prediction model.

[0036] As a preferred solution, the network weights and network thresholds of the wavelet neural network are specifically:

[0037]

[0038]

[0039] Among them, μ i (t) is the learning rate parameter of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔW i (t) is the weight update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔB i (t) is the threshold update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, B i (t) is the wavelet neural network threshold matrix of the t-th time slot, W i (t) is the weight matrix of the wavelet neural network in the t-th time slot, e i (t) is the wavelet neural network prediction error adapted to the i-th noise type in the t-th time slot.

[0040] As a preferred solution, the actual output of the network is specifically:

[0041]

[0042] Among them, y i,n(t) is the nth output of the wavelet neural network adapted to the i-th noise type in the t-th time slot, x k,i (t) is the kth neuron input of the wavelet neural network input layer adapted to the i-th noise type in the t-th time slot, w ks,i (t) is the weight between the input layer neurons and the hidden layer neurons in the tth time slot, b s,i (t) is the hidden layer threshold of the t-th time slot, w sn,i (t) is the weight between the hidden layer neurons and the output layer neurons in the tth time slot, b n,i (t) is the output layer threshold of the t-th time slot, a s,i (t) is the scaling factor of the hidden layer in the t-th time slot, and f is the wavelet basis function.

[0043] As a preferred solution, the noise type identification module calculates the matching similarity between the original noise data and the historical data of each noise type, specifically:

[0044] The noise type identification module performs calculations according to the following formula:

[0045]

[0046] Among them, D i (t) is the matching similarity between the original noise data of the t-th time slot and the i-th noise type, e i (t-1) is the historical prediction error of the wavelet neural network adapted to the i-th noise type, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, α i (t) is the weight parameter of the matching deviation of the i-th noise type in the t-th time slot, β i (t) is the weight parameter of the historical prediction error of the wavelet neural network adapted to the i-th noise type in the t-th time slot.

[0047] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0048] An embodiment of the present invention provides a power line carrier channel noise prediction method and system that considers type matching deviation. The power line carrier channel noise prediction method includes: collecting original noise data of the power line carrier channel noise; calculating the matching similarity between the original noise data and the historical data of each noise type respectively, and taking the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; calculating the matching deviation between the original noise data and the corresponding noise type historical data; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise and random pulse noise; inputting the original noise data into a pre-trained noise prediction model, and realizing the prediction of the power line carrier channel noise based on the output of the noise prediction model; wherein the noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the corresponding noise type historical data, and is trained by a preset wavelet neural network. Compared with the existing technology, the present application identifies the type of power line carrier channel noise and trains a preset wavelet network model based on the matching deviation between the original noise data and the corresponding noise type historical data to realize the prediction of power line carrier channel noise. It can adapt the noise type and the prediction model, effectively improving the precision and accuracy of the prediction.

[0049] Furthermore, during the training process of the noise prediction model, the network weights and network thresholds are iteratively updated through the prediction error and matching deviation. That is, the influence of the type matching deviation in the network weight and threshold update process is taken into account, and the learning rate of the wavelet neural network is adaptively updated. Therefore, the prediction accuracy of the wavelet neural network is continuously improved through training, and the performance of the obtained noise prediction model is further improved.

[0050] Furthermore, the noise type matching deviation and wavelet neural network historical prediction error are used to describe the matching similarity between the power line carrier channel noise and the noise type. When the noise type matching deviation and the wavelet neural network historical prediction error are smaller, the noise type matching similarity is greater; conversely, the smaller the noise type matching similarity is, the accurate identification of the noise type is achieved, and the degree of adaptation between the noise type and the power line carrier channel noise prediction model is further improved, thereby improving the accuracy of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 : A flow chart of an embodiment of a power line carrier channel noise prediction method provided by the present invention taking into account type matching deviation.

[0052] Figure 2 : A flow chart of another embodiment of a power line carrier channel noise prediction method provided by the present invention taking into account type matching deviation.

[0053] Figure 3 : A structural diagram of an embodiment of a power line carrier channel noise prediction system provided by the present invention taking into account type matching deviation. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0055] Embodiment one:

[0056] Please refer to Figure 1 and Figure 2 , Figure 1 and Figure 2 A power line carrier channel noise prediction method considering type matching deviation provided in an embodiment of the present invention includes steps S1 to S3; wherein,

[0057] Step S1: collecting original noise data of the power line carrier channel noise.

[0058] Step S2, respectively calculating the matching similarity between the original noise data and the historical data of each noise type, and taking the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; calculating the matching deviation between the original noise data and the historical data of the corresponding noise type; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise and random pulse noise.

[0059] Step S3: input the original noise data into a pre-trained noise prediction model, and predict the power line carrier channel noise based on the output of the noise prediction model; wherein the noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type, and the matching deviation between the original noise data and the historical data of the corresponding noise type, and is trained by a preset wavelet neural network.

[0060] In this embodiment, step S2 takes into account the complex noise environment of the power line carrier channel in a power electronics environment with a high proportion of renewable energy access. A large number of power electronic devices generate complex noise, which in turn seriously affects the quality of power line communication. Based on the different noise sources, noise can be divided into five categories: colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise, and random pulse noise.

[0061] Preferably, the calculating of the matching similarity between the original noise data and the historical data of each noise type is specifically as follows:

[0062] Calculate according to the following formula:

[0063]

[0064] Among them, D i (t) is the matching similarity between the original noise data of the t-th time slot and the i-th noise type, e i (t-1) is the historical prediction error of the wavelet neural network adapted to the i-th noise type, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, α i (t) is the weight parameter of the matching deviation of the i-th noise type in the t-th time slot, β i (t) is the weight parameter of the wavelet neural network historical prediction error adapted to the i-th noise type in the t-th time slot. In the embodiment of the present application, the matching similarity between the noise data and the noise type is described by the noise type matching deviation and the wavelet neural network historical prediction error. When the noise type matching deviation and the wavelet neural network historical prediction error are smaller, the noise type matching similarity is greater, and the noise data tends to be identified as the noise type at this time; conversely, the smaller the noise type matching similarity, the lower the tendency to the corresponding type. Thus, based on the matching similarity between the original noise data and the historical data of each noise type (the above five noise types), the noise type corresponding to the highest matching similarity is identified as the noise type of the power line carrier channel noise.

[0065] Furthermore, for the above step S3, the noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the corresponding noise type historical data, and is trained through a preset wavelet neural network.

[0066] The preset wavelet neural network includes an input layer, a hidden layer, and an output layer. This embodiment can construct different wavelet neural networks for different noise types. For the subsequent calculation process, the important parameters of the network can be defined first, specifically: k i (t) is the number of neurons in the input layer of the wavelet neural network adapted to the i-th noise type in the t-th time slot, s i (t) is the number of neurons in the hidden layer. In addition, there is a network parameter P of the wavelet neural network adapted to the i-th noise type in the t-th time slot. i (W i (t),B i (t)), where W i(t) is the wavelet neural network weight matrix of the t-th time slot, B i (t) is the threshold matrix of the wavelet neural network in the t-th time slot.

[0067] Use the wavelet neural network model as the basic model and train it:

[0068] First, a data training set is obtained; wherein the data training set includes noise samples and corresponding expected outputs; the noise samples are input into the wavelet neural network to obtain the actual output of the network;

[0069] Specifically, the actual output of the network is:

[0070]

[0071] Among them, y i,n (t) is the nth output of the wavelet neural network adapted to the i-th noise type in the t-th time slot, x k,i (t) is the kth neuron input of the wavelet neural network input layer adapted to the i-th noise type in the t-th time slot, w ks,i (t) is the weight between the input layer neurons and the hidden layer neurons in the tth time slot, b s,i (t) is the hidden layer threshold of the t-th time slot, w sn,i (t) is the weight between the hidden layer neurons and the output layer neurons in the tth time slot, b n,i (t) is the output layer threshold of the t-th time slot, a s,i (t) is the scaling factor of the hidden layer in the t-th time slot, and f is the wavelet basis function.

[0072] According to the actual output of the network and the expected output calculated by formula (2), the wavelet neural network prediction error is further calculated:

[0073]

[0074] Among them, e i (t) is the prediction error of the wavelet neural network adapted to the i-th noise type in the t-th time slot, The nth expected output of the wavelet neural network adapted to the i-th noise type for the t-th time slot.

[0075] Based on the wavelet neural network prediction error and the matching deviation, the network weight and network threshold of the wavelet neural network are iteratively updated.

[0076] Specifically, the network weights and network thresholds of the wavelet neural network are:

[0077]

[0078]

[0079] Among them, μ i (t) is the learning rate parameter of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔW i (t) is the weight update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔB i (t) is the threshold update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot. The embodiment of the present application improves the learning rate by matching the deviation so that the learning rate is inversely proportional to the matching deviation. That is, the larger the matching deviation, the lower the learning rate, thereby achieving adaptive update of the learning rate to improve the prediction accuracy of the wavelet neural network.

[0080] Until the wavelet neural network prediction error at the t-th time slot is less than a preset error precision emin(t), the wavelet neural network is determined to have met the convergence condition, and the trained noise prediction model is obtained. The raw noise data is input into the trained noise prediction model, and the power line carrier channel noise is predicted based on the trained noise prediction model to obtain a corresponding prediction result.

[0081] Accordingly, refer to Figure 3 The embodiment of the present invention further provides a power line carrier channel noise prediction system considering type matching deviation, including an acquisition module 101, a noise type identification module 102 and a prediction module 103; wherein,

[0082] The acquisition module 101 is used to collect original noise data of the power line carrier channel noise;

[0083] The noise type identification module 102 is configured to respectively calculate the matching similarity between the original noise data and the historical data of each noise type, and use the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; and calculate the matching deviation between the original noise data and the historical data of the corresponding noise type; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise, and random pulse noise;

[0084] The prediction module 103 is configured to input the original noise data into a pre-trained noise prediction model, and predict the power line carrier channel noise based on the output of the noise prediction model;

[0085] The noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the historical data of the corresponding noise type, and is trained through a preset wavelet neural network.

[0086] As a preferred implementation, the training process of the noise prediction model includes:

[0087] Obtaining a data training set; wherein the data training set includes noise samples and corresponding expected outputs;

[0088] Inputting the noise sample into the wavelet neural network to obtain the actual output of the network;

[0089] Calculating a wavelet neural network prediction error based on the actual output of the network and the expected output;

[0090] Based on the wavelet neural network prediction error and the matching deviation, the network weights and network thresholds of the wavelet neural network are iteratively updated until the prediction error is less than a preset error accuracy, thereby obtaining the trained noise prediction model.

[0091] As a preferred embodiment, the network weights and network thresholds of the wavelet neural network are specifically as follows:

[0092]

[0093]

[0094] Among them, μ i (t) is the learning rate parameter of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔW i (t) is the weight update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, ΔB i (t) is the threshold update amount of the wavelet neural network adapted to the i-th noise type in the t-th time slot, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, B i (t) is the wavelet neural network threshold matrix of the t-th time slot, W i (t) is the weight matrix of the wavelet neural network in the t-th time slot, e i (t) is the wavelet neural network prediction error adapted to the i-th noise type in the t-th time slot.

[0095] As a preferred embodiment, the actual output of the network is specifically:

[0096]

[0097] Among them, y i,n (t) is the nth output of the wavelet neural network adapted to the i-th noise type in the t-th time slot, x k,i (t) is the kth neuron input of the wavelet neural network input layer adapted to the i-th noise type in the t-th time slot, w ks,i(t) is the weight between the input layer neurons and the hidden layer neurons in the tth time slot, b s,i (t) is the hidden layer threshold of the t-th time slot, w sn,i (t) is the weight between the hidden layer neurons and the output layer neurons in the tth time slot, b n,i (t) is the output layer threshold of the t-th time slot, a s,i (t) is the scaling factor of the hidden layer in the t-th time slot, and f is the wavelet basis function.

[0098] As a preferred embodiment, the noise type identification module 102 calculates the matching similarity between the original noise data and the historical data of each noise type, specifically:

[0099] The noise type identification module 102 performs calculations according to the following formula:

[0100]

[0101] Among them, D i (t) is the matching similarity between the original noise data of the t-th time slot and the i-th noise type, e i (t-1) is the historical prediction error of the wavelet neural network adapted to the i-th noise type, error i (t) is the matching deviation of the i-th noise type in the t-th time slot, α i (t) is the weight parameter of the matching deviation of the i-th noise type in the t-th time slot, β i (t) is the weight parameter of the historical prediction error of the wavelet neural network adapted to the i-th noise type in the t-th time slot.

[0102] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0103] An embodiment of the present invention provides a power line carrier channel noise prediction method and system that considers type matching deviation. The power line carrier channel noise prediction method includes: collecting original noise data of the power line carrier channel noise; calculating the matching similarity between the original noise data and the historical data of each noise type respectively, and taking the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; calculating the matching deviation between the original noise data and the corresponding noise type historical data; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise and random pulse noise; inputting the original noise data into a pre-trained noise prediction model, and realizing the prediction of the power line carrier channel noise based on the output of the noise prediction model; wherein the noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the corresponding noise type historical data, and is trained by a preset wavelet neural network. Compared with the existing technology, the present application identifies the type of power line carrier channel noise and trains a preset wavelet network model based on the matching deviation between the original noise data and the corresponding noise type historical data to realize the prediction of power line carrier channel noise. It can adapt the noise type and the prediction model, effectively improving the precision and accuracy of the prediction.

[0104] Furthermore, during the training process of the noise prediction model, the network weights and network thresholds are iteratively updated through the prediction error and matching deviation. That is, the influence of the type matching deviation in the network weight and threshold update process is taken into account, and the learning rate of the wavelet neural network is adaptively updated. Therefore, the prediction accuracy of the wavelet neural network is continuously improved through training, and the performance of the obtained noise prediction model is further improved.

[0105] Furthermore, the noise type matching deviation and wavelet neural network historical prediction error are used to describe the matching similarity between the power line carrier channel noise and the noise type. When the noise type matching deviation and the wavelet neural network historical prediction error are smaller, the noise type matching similarity is greater; conversely, the smaller the noise type matching similarity is, the accurate identification of the noise type is achieved, and the degree of adaptation between the noise type and the power line carrier channel noise prediction model is further improved, thereby improving the accuracy of the prediction results.

[0106] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for predicting power line carrier channel noise considering type matching deviation, characterized in that: include: Collecting raw noise data of power line carrier channel noise; Calculating the matching similarities between the original noise data and the historical data of each noise type respectively, and taking the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; Calculating the matching deviation between the original noise data and the historical data of the corresponding noise type; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise and random pulse noise; Inputting the original noise data into a pre-trained noise prediction model, and predicting the power line carrier channel noise based on the output of the noise prediction model; The noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the historical data of the corresponding noise type, and is trained by a preset wavelet neural network; The training process of the noise prediction model includes: Obtaining a data training set; wherein the data training set includes noise samples and corresponding expected outputs; Inputting the noise sample into the wavelet neural network to obtain the actual output of the network; Calculating a wavelet neural network prediction error based on the actual output of the network and the expected output; Iteratively updating the network weights and network thresholds of the wavelet neural network based on the wavelet neural network prediction error and the matching deviation until the prediction error is less than a preset error accuracy, thereby obtaining the trained noise prediction model; The network weights and network thresholds of the wavelet neural network are specifically: ； ； in, For the Time slot adaptation The learning rate parameters of the wavelet neural network with different noise types, For the Time slot adaptation The weight update amount of the wavelet neural network with different noise types, For the Time slot adaptation The threshold update amount of the wavelet neural network with different noise types, For the Time slot Matching deviation of noise types, For the The wavelet neural network threshold matrix of time slots, For the The time-slot wavelet neural network weight matrix, For the Time slot adaptation Wavelet neural network prediction error of different noise types; The actual output of the network is: ； in, For the Time slot adaptation Wavelet neural network with different noise types outputs, For the Time slot adaptation The input layer of the wavelet neural network with different noise types neuron input, For the The weights between the input layer neurons and the hidden layer neurons for each time slot, For the The hidden layer threshold of time slots, For the The weights between the hidden layer neurons and the output layer neurons in each time slot, For the The output layer threshold of the time slot, For the The scaling factor of the hidden layer of each time slot, is the wavelet basis function; The wavelet neural network prediction error is: ； in, For the Time slot adaptation The prediction error of the wavelet neural network for different noise types, For the Time slot adaptation Wavelet neural network with different noise types Expected output.

2. A method for predicting power line carrier channel noise considering type matching deviation according to claim 1, characterized in that: The calculation of the matching similarity between the original noise data and the historical data of each noise type is specifically as follows: Calculate according to the following formula: ； in, For the The original noise data of the time slot is compared with the The matching similarity of the noise types, To adapt to the The historical prediction error of the wavelet neural network with different noise types, For the Time slot Matching deviation of noise types, For the Time slot The weight parameters of the noise type matching deviation, For the Time slot adaptation The weight parameters of the historical prediction errors of the wavelet neural network for different noise types.

3. A power line carrier channel noise prediction system considering type matching deviation, characterized in that: It includes acquisition module, noise type recognition module and prediction module; among them, The acquisition module is used to collect original noise data of the power line carrier channel noise; The noise type identification module is configured to respectively calculate the matching similarity between the original noise data and the historical data of each noise type, and use the noise type corresponding to the highest matching similarity as the noise type of the power line carrier channel noise; calculate the matching deviation between the original noise data and the historical data of the corresponding noise type; wherein the noise types include colored background noise, narrowband noise, power frequency asynchronous periodic pulse noise, power frequency synchronous periodic pulse noise, and random pulse noise; The prediction module is configured to input the original noise data into a pre-trained noise prediction model, and predict the power line carrier channel noise based on the output of the noise prediction model; The noise prediction model is based on the wavelet neural network prediction error of the corresponding noise type and the matching deviation between the original noise data and the historical data of the corresponding noise type, and is trained by a preset wavelet neural network; The training process of the noise prediction model includes: Obtaining a data training set; wherein the data training set includes noise samples and corresponding expected outputs; Inputting the noise sample into the wavelet neural network to obtain the actual output of the network; Calculating a wavelet neural network prediction error based on the actual output of the network and the expected output; Iteratively updating the network weights and network thresholds of the wavelet neural network based on the wavelet neural network prediction error and the matching deviation until the prediction error is less than a preset error accuracy, thereby obtaining the trained noise prediction model; The network weights and network thresholds of the wavelet neural network are specifically: ； ； in, For the Time slot adaptation The learning rate parameters of the wavelet neural network with different noise types, For the Time slot adaptation The weight update amount of the wavelet neural network with different noise types, For the Time slot adaptation The threshold update amount of the wavelet neural network with different noise types, For the Time slot Matching deviation of noise types, For the The wavelet neural network threshold matrix of time slots, For the The time-slot wavelet neural network weight matrix, For the Time slot adaptation Wavelet neural network prediction error of different noise types; The actual output of the network is: ； in, For the Time slot adaptation Wavelet neural network with different noise types outputs, For the Time slot adaptation The input layer of the wavelet neural network with different noise types neuron input, For the The weights between the input layer neurons and the hidden layer neurons for each time slot, For the The hidden layer threshold of time slots, For the The weights between the hidden layer neurons and the output layer neurons in each time slot, For the The output layer threshold of the time slot, For the The scaling factor of the hidden layer of each time slot, is the wavelet basis function; The wavelet neural network prediction error is: ； in, For the Time slot adaptation The prediction error of the wavelet neural network for different noise types, For the Time slot adaptation Wavelet neural network with different noise types Expected output.

4. A power line carrier channel noise prediction system considering type matching deviation according to claim 3, characterized in that: The noise type identification module calculates the matching similarity between the original noise data and the historical data of each noise type, specifically: The noise type identification module performs calculations according to the following formula: ； in, For the The original noise data of the time slot is compared with the The matching similarity of the noise types, To adapt to the The historical prediction error of the wavelet neural network with different noise types, For the Time slot Matching deviation of noise types, For the Time slot The weight parameters of the noise type matching deviation, For the Time slot adaptation The weight parameters of the historical prediction errors of the wavelet neural network for different noise types.

Images