Flow prediction method and device of power distribution communication network, terminal equipment and storage medium

Abstract

The invention discloses a flow prediction method and device for a power distribution communication network, terminal equipment and a storage medium. The method comprises the following steps: acquiring historical time sequence flow data of a power distribution communication network to be predicted; inputting the historical time sequence traffic data into a preset traffic prediction model, so that the traffic prediction model performs feature extraction on the historical time sequence traffic data through an internal time convolutional neural network to obtain time sequence traffic features, and inputting the time sequence traffic features into an internal long-short-term memory network to obtain time sequence traffic data; and determining the predicted flow of the power distribution communication network to be predicted in a future time period by the long short-term memory network according to the time sequence flow characteristics. By implementing the embodiment of the invention, the accuracy of flow prediction of the power distribution communication network can be improved.

Description

Technical Field

[0001] The present invention relates to the field of flow prediction, and in particular to a flow prediction method, device, terminal equipment and storage medium for a power distribution communication network. Background Art

[0002] With the rapid development of the Internet, various online applications have emerged in an endless stream, network behavior has become more complex, and network traffic has also shown an explosive growth trend. This has brought huge challenges to network operators in managing and maintaining websites. In order to meet these challenges, operators need to adopt effective strategies to optimize network resources to improve network stability and reliability. However, determining the appropriate amount of network resources is a difficult task, which requires finding a balance between insufficient and excessive resource supply. If network resources are insufficiently supplied, it may cause network congestion, affect network performance, and cause user dissatisfaction; if resources are oversupplied, it may cause a large amount of network resources to be wasted, increasing the operator's operating costs. Network traffic is a key indicator for measuring network operation load and an important basis for evaluating network performance. By monitoring and analyzing network traffic, operators can more accurately evaluate the operation status of the network and make more reasonable resource allocation decisions.

[0003] In the prior art, convolutional neural networks (CNN) are usually used to predict network traffic. Convolutional neural networks can extract local features and global dependencies from traffic data to perform traffic prediction. However, traditional CNNs cannot capture long-term dependencies in sequences and are not accurate enough in predicting network traffic. Summary of the invention

[0004] Embodiments of the present invention provide a flow prediction method, apparatus, terminal equipment and storage medium for a power distribution communication network, which can improve the accuracy of flow prediction for the power distribution communication network.

[0005] An embodiment of the present invention provides a method for predicting flow of a power distribution communication network, comprising: obtaining historical time series flow data of the power distribution communication network to be predicted;

[0006] The historical time series flow data is input into a preset flow prediction model, so that the flow prediction model extracts features from the historical time series flow data through a built-in time convolutional neural network to obtain time series flow features, and the time series flow features are input into a built-in long short-term memory network, so that the long short-term memory network determines the predicted flow of the distribution communication network to be predicted in a future time period based on the time series flow features.

[0007] Further, after determining the predicted flow of the power distribution communication network to be predicted in a future period, the method includes:

[0008] Determine the low flow period based on the predicted flow in the future period;

[0009] When the traffic trough period arrives, the maintenance plan of the power distribution communication network to be predicted is executed, and the power distribution communication network to be predicted is subjected to off-peak maintenance.

[0010] Furthermore, the training of the traffic prediction model includes:

[0011] Obtain the outbound and inbound traffic of each device port in a historical period of the power distribution communication network to obtain a time series traffic sample;

[0012] The time series traffic samples are supplemented with missing values, subjected to logarithmic processing, and normalized and converted into supervised data to obtain training samples;

[0013] The training samples are input into a traffic prediction model provided with a time convolutional neural network and a long short-term memory network for training until the loss function converges.

[0014] Furthermore, the time series traffic samples are supplemented with missing values, logarithmically processed, and normalized and converted into supervised data to obtain training samples, including:

[0015] For the time series traffic sample, supplement the missing values ​​to obtain a filled time series traffic sample;

[0016] Perform logarithmic processing on the filled time series traffic samples to obtain logarithmically converted time series traffic samples;

[0017] The time series traffic samples after logarithmic transformation are normalized to obtain normalized time series traffic samples;

[0018] For the normalized time series traffic samples, based on the sliding window strategy, starting from the left side of the normalized time series traffic samples, the window is gradually slid to the right according to the time step to generate several input sequences and the target value corresponding to each input sequence. The target value is used as the label of the corresponding input sequence to generate training samples; among which, the target value of each input sequence is the traffic value at the next time point of the last time point in the input sequence.

[0019] Furthermore, the missing values ​​of the time series traffic samples are supplemented by the following formula:

[0020]

[0021] in, It represents the missing value of the flow data of the distribution communication network at day i, hour j, and minute k; It represents the flow data of the power distribution communication network at the i-1 day, the jth hour, and the kth minute; It represents the traffic data of the power distribution communication network at the i+1th day, the jth hour, and the kth minute.

[0022] Furthermore, the filled time series traffic samples are logarithmically processed by the following formula:

[0023] x′=log(1+x);

[0024] Among them, x is the traffic data in the filled time series traffic sample, and x' is the traffic data obtained after logarithmic transformation.

[0025] Furthermore, the logarithmically transformed time series traffic samples are normalized using the following formula:

[0026]

[0027] Among them, x * The traffic data in the time series traffic sample after normalization, x min is the minimum value of the flow data in the time series flow sample after logarithmic transformation, x man It is the maximum value of the flow data in the time series flow sample after logarithmic transformation.

[0028] Based on the above method embodiment, the present invention provides a corresponding device embodiment;

[0029] An embodiment of the present invention provides a flow prediction device for a power distribution communication network, comprising: a data acquisition module and a model prediction module;

[0030] The data acquisition module is used to acquire the historical time series flow data of the power distribution communication network to be predicted;

[0031] The model prediction module is used to input the historical time series flow data into a preset flow prediction model, so that the flow prediction model can extract features from the historical time series flow data through a built-in time convolutional neural network to obtain time series flow features, and input the time series flow features into a built-in long short-term memory network, so that the long short-term memory network can determine the predicted flow of the distribution communication network to be predicted in a future time period based on the time series flow features.

[0032] Based on the above method embodiment, the present invention provides another embodiment;

[0033] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the flow prediction method of the distribution communication network as described in any method embodiment of the present invention.

[0034] Based on the above method embodiment, the present invention provides another embodiment:

[0035] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the flow prediction method for the distribution communication network as described in any method embodiment of the present invention.

[0036] The following beneficial effects are achieved by implementing the embodiments of the present invention:

[0037] The embodiment of the present invention provides a flow prediction method, apparatus, terminal device and storage medium for a distribution communication network. The method first obtains the historical time series flow data of the distribution communication network to be predicted, and then inputs the time series flow data into a preset flow prediction model. The flow prediction model is provided with a time convolutional neural network and a long short-term memory network. The time series flow features are extracted from the historical time series flow data by the time convolutional neural network, and then the predicted flow of the distribution communication network to be predicted in a future period is determined by the long short-term memory network according to the time series flow features. Compared with the prior art, the present invention combines the time convolutional neural network with the long short-term memory network, making full use of the feature extraction capability of the time convolutional neural network and the long-term dependency capture capability of the long short-term memory network, so as to capture the long-term dependency in the sequence and thus improve the accuracy of flow prediction of the distribution communication network. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of a method for traffic prediction of a power distribution communication network provided in one embodiment of the present invention.

[0039] Figure 2 It is a structural diagram of a flow prediction device for a power distribution communication network in one embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] like Figure 1 As shown, a flow prediction method for a power distribution communication network provided by an embodiment of the present invention includes the following steps:

[0042] Step S1: Obtain historical time series flow data of the power distribution communication network to be predicted.

[0043] Specifically, the above historical time series traffic data includes the outflow and inflow of each device port of the power distribution communication network to be predicted within a historical period.

[0044] Step S2: input the historical time series flow data into a preset flow prediction model, so that the flow prediction model extracts features from the historical time series flow data through a built-in time convolutional neural network to obtain time series flow features, and inputs the time series flow features into a built-in long short-term memory network, so that the long short-term memory network determines the predicted flow of the distribution communication network to be predicted in a future time period based on the time series flow features.

[0045] The following is a detailed description of the traffic prediction model:

[0046] In a preferred embodiment, the training of the traffic prediction model includes:

[0047] Obtain the outbound and inbound traffic of each device port in the power distribution communication network within a historical period to obtain a time series traffic sample;

[0048] The time series traffic samples are supplemented with missing values, subjected to logarithmic processing, and normalized and converted into supervised data to obtain training samples;

[0049] The training samples are input into a traffic prediction model provided with a time convolutional neural network and a long short-term memory network for training until the loss function converges.

[0050] In a preferred embodiment, the time series traffic samples are supplemented with missing values, logarithmically processed, and normalized and converted into supervised data to obtain training samples, including:

[0051] For the time series traffic sample, supplement the missing values ​​to obtain a filled time series traffic sample;

[0052] Perform logarithmic processing on the filled time series traffic samples to obtain logarithmically converted time series traffic samples;

[0053] The time series traffic samples after logarithmic transformation are normalized to obtain normalized time series traffic samples;

[0054] For the normalized time series traffic samples, based on the sliding window strategy, starting from the left side of the normalized time series traffic samples, the window is gradually slid to the right according to the time step to generate several input sequences and the target value corresponding to each input sequence. The target value is used as the label of the corresponding input sequence to generate training samples; among which, the target value of each input sequence is the traffic value at the next time point of the last time point in the input sequence.

[0055] Specifically, when making traffic predictions, the missing values ​​are properly processed and filled. Here, the mean filling method is used. When data is missing, the mean of network traffic at the same time point on other days can be used as the filling value. Preferably, the missing values ​​of the time series traffic samples are supplemented by the following formula:

[0056]

[0057] in, It represents the missing value of the flow data of the distribution communication network at day i, hour j, and minute k; It represents the flow data of the power distribution communication network at the i-1 day, the jth hour, and the kth minute; It represents the traffic data of the power distribution communication network at the i+1th day, the jth hour, and the kth minute.

[0058] In order to reduce the order of magnitude of the data while retaining the basic characteristics of the time series data and making the distribution of the data closer to the normal distribution, the data is further logarithmically transformed after filling the missing values, which is beneficial for the analysis and prediction of time series.

[0059] Preferably, the filled time series traffic samples are logarithmically processed by the following formula:

[0060] x′=log(1+x);

[0061] Among them, x is the traffic data in the filled time series traffic sample, and x' is the traffic data obtained after logarithmic transformation.

[0062] Next, in order to optimize the performance of the model and ensure its stability and prediction accuracy, the minimum-maximum normalization strategy is used to normalize the time series traffic samples after logarithmic transformation. The mathematical expression of this normalization method is:

[0063]

[0064] Among them, x * The traffic data in the time series traffic sample after normalization, x min is the minimum value of the flow data in the time series flow sample after logarithmic transformation, x manIt is the maximum value of the flow data in the time series flow sample after logarithmic transformation.

[0065] In an optional embodiment, in order to effectively remove noise, especially for noise types with obvious nonlinear characteristics such as impulse noise and salt and pepper noise, the normalized time series traffic samples can be further denoised. Specifically, the median filtering method can be used to process outliers or extreme values ​​in the data. The median filtering method has good resistance to outliers or extreme values ​​in the data, so that the important edge features of the data can be better preserved, and it has a significant inhibitory effect on salt and pepper noise and impulse noise, preventing the loss of data details. The mathematical expression of median filtering can be expressed by the following formula:

[0066] in, represents the median value of the window data, m refers to the size of the window, and Med(·) is the function of the median filter.

[0067] In order to construct the data set required for supervised learning, we adopted a sliding window strategy, that is, starting from the left side of the data sequence, we gradually slide the window to the right according to the time step, so as to generate pairs of features and target values ​​one by one. Features refer to the input sequence extracted by the sliding window method, which will serve as the input of the model; and the target value refers to the flow value at the next time point, which will serve as the predicted output of the model. In this way, we can systematically convert the original time series data into a format suitable for supervised learning algorithm processing.

[0068] After the above data processing is performed on the time series traffic samples, training samples for model training can be obtained, and then the training samples are input into the traffic prediction model to perform traffic prediction.

[0069] In the present invention, the traffic prediction model is a comprehensive model based on long short-term neural memory neural network and time convolution neural network (TCN-LSTM model);

[0070] Based on the data format requirements of the temporal convolutional neural network, the input data must be converted into a three-dimensional vector with a shape of [samples, timesteps, features]. Samples is the number of samples in the data set, timesteps is the step size, that is, how many previous data the output data is associated with, and features is the dimension. Therefore, the original data set format needs to be reshaped to convert the two-dimensional matrix into the required three-dimensional matrix.

[0071] Temporal Convolutional Neural Network (TCN) combines some advanced technologies in CNN, such as causal convolution, dilated convolution and residual connections, which make TCN particularly suitable for processing sequence data.

[0072] The meaning of causal convolution is: given a filter F = (f1, f2, ..., f k ), and the time series X = (x1, x2, …, x n ), where k is the size of the filter, and n is the length of the sequence X, f i is the sequence X on x i The weight of x t The mathematical expression of causal convolution is:

[0073]

[0074] The dilated convolution can be t The convolution result is calculated by the expansion factor d at the position, and its mathematical expression is given by the following formula:

[0075]

[0076] Where d is the dilation factor. This approach allows convolutional networks to capture long-term dependencies without significantly increasing computational complexity, thus performing well in sequence modeling tasks.

[0077] In addition, the parameters of the temporal convolutional neural network (TCN) include the depth of the network (ie, the number of layers), the size of the convolution kernel, the number of convolution kernels per layer, the batch size during training, and the selected optimization algorithm. By optimizing these parameters, the prediction ability and operation speed of the model can be significantly improved.

[0078] The unit structure of the long short-term memory network (LSTM) is designed to include two parts: hidden state and cell state. The former is responsible for short-term memory, and the latter is responsible for long-term memory. The core innovation of LSTM is the introduction of gating. o The mechanism controls the flow of information, allowing the network to learn and remember long-term dependencies in time series. When analyzing the calculation process of the LSTM unit at time point t, we focus on the hidden state h at the previous moment. t-1 and cell state c t-1 , and the current input x t The specific calculation rules are given below.

[0079] f t =σ(w f [ht-1 ,x t ]+b f )

[0080] i t =σ(w i [h t-1 ,x t ]+b i )

[0081]

[0082] o t =σ(w o [h t-1 ,x t ]+b o )

[0083] o t =σ(w o [h t-1 ,x t ]+b o )

[0084] Among them, f, i, o and There are three types of gate vectors and candidate vectors, Tanh is the hyperbolic tangent function; w f , w o , w i , w c represents the weight vector; represents the dot product of the vector; b f , b o , b i , b c etc. represent the offset; σ represents the Sigmoid function.

[0085] In an optional embodiment, TCN is composed of two residual modules. The first residual module consists of two causal dilated convolutions with two convolution kernels of size 5, dilated convolution coefficients of 1, and 10 convolution kernels, while the dilated convolution of the second residual module is 2, and the other parameters are consistent with the first residual block. The output of the TCN layer can be calculated as follows:

[0086]

[0087] L 2 =ResidualBlock(L 1 ,5,2)

[0088] ResidualBlock(·) represents the residual module function of TCN, L 1 Representation sequence After the output of the first residual module, L 2 Represents the sequence L1 After the output of the second residual module. 2 As the input of LSTM, the LSTM model is used to obtain the long-term dependency of the sequence, thereby obtaining the output h of the last time step T ,h T It can be calculated by the LSTM part formula introduced above, and then the output h T As the input of the fully connected layer, the output z of this layer is obtained T .

[0089] z T =relu(vh T +bias)

[0090] Among them, v is the weight matrix, bias is the bias vector, and relu(·) represents the relu activation function.

[0091] Finally, z T As the input of the output layer, the traffic prediction value at the next moment is obtained

[0092]

[0093] Among them, u represents the weight parameter of the fully connected layer, q represents the bias parameter, and linear(·) represents the linear function.

[0094] In a preferred embodiment, after determining the predicted flow of the power distribution communication network to be predicted in a future period, the method includes:

[0095] Determine the low flow period based on the predicted flow in the future period;

[0096] When the traffic trough period arrives, the maintenance plan of the power distribution communication network to be predicted is executed, and the power distribution communication network to be predicted is subjected to off-peak maintenance.

[0097] In this embodiment, based on the predicted traffic in the future time period, the time period when the predicted traffic is lower than the preset value is regarded as the traffic low period. When the traffic low period arrives, the predicted distribution communication network is subjected to off-peak maintenance according to the preset maintenance plan, thereby optimizing the traffic flow path and reducing congestion.

[0098] Based on the above method embodiment of the present invention, a corresponding device embodiment is provided;

[0099] like Figure 2 As shown, another embodiment of the present invention provides a flow prediction device for a power distribution communication network, comprising: a data acquisition module and a model prediction module;

[0100] The data acquisition module is used to acquire the historical time series flow data of the power distribution communication network to be predicted.

[0101] The model prediction module is used to input the historical time series flow data into a preset flow prediction model, so that the flow prediction model can extract features from the historical time series flow data through a built-in time convolutional neural network to obtain time series flow features, and input the time series flow features into a built-in long short-term memory network, so that the long short-term memory network can determine the predicted flow of the distribution communication network to be predicted in a future time period based on the time series flow features.

[0102] It can be understood that the above-mentioned device item embodiments correspond to the method item embodiments of the present invention, and can implement the flow prediction method of the distribution communication network provided by any one of the above-mentioned method item embodiments of the present invention.

[0103] It should be noted that the device embodiments described above are merely schematic, wherein the units / modules described as separate components may or may not be physically separated, and the components displayed as units / modules may or may not be physical units / modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without creative work. The schematic diagram is merely an example of the device and does not constitute a limitation on the device, and may include more or fewer components than shown in the diagram, or a combination of certain components, or different components.

[0104] Based on the above method embodiment of the present invention, a corresponding terminal device embodiment is provided;

[0105] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the flow prediction method for the distribution communication network provided by any one of the above-mentioned method embodiments of the present invention.

[0106] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0107] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The device may include, but is not limited to, a processor and a memory. Those skilled in the art may understand that, for example, the device may also include an input / output device, a network access device, a bus, etc.

[0108] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the device, and various interfaces and lines are used to connect various parts of the entire device.

[0109] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0110] Another embodiment is provided based on the above method embodiment of the present invention. Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the traffic prediction method for the distribution communication network described in any one of the above embodiments of the present invention.

[0111] Wherein, the storage medium is a computer-readable storage medium, and the module / unit integrated in the device can be stored in a computer-readable storage medium if it is implemented in the form of a software functional unit and sold or used as an independent product. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0112] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A flow prediction method for a power distribution communication network, characterized in that: include: Obtain historical time series flow data of the power distribution communication network to be predicted; The historical time series flow data is input into a preset flow prediction model, so that the flow prediction model extracts features from the historical time series flow data through a built-in time convolutional neural network to obtain time series flow features, and the time series flow features are input into a built-in long short-term memory network, so that the long short-term memory network determines the predicted flow of the distribution communication network to be predicted in a future time period based on the time series flow features.

2. The flow prediction method of the power distribution communication network according to claim 1, characterized in that: After determining the predicted flow of the power distribution communication network to be predicted in a future period, it includes: Determine the low flow period based on the predicted flow in the future period; When the traffic trough period arrives, the maintenance plan of the power distribution communication network to be predicted is executed, and the power distribution communication network to be predicted is subjected to off-peak maintenance.

3. The flow prediction method of the power distribution communication network according to claim 1, characterized in that: The training of the traffic prediction model includes: Obtain the outbound and inbound traffic of each device port in a historical period of the power distribution communication network to obtain a time series traffic sample; The time series traffic samples are supplemented with missing values, subjected to logarithmic processing, and normalized and converted into supervised data to obtain training samples; The training samples are input into a traffic prediction model provided with a time convolutional neural network and a long short-term memory network for training until the loss function converges.

4. The flow prediction method of the power distribution communication network according to claim 3, characterized in that: The time series traffic samples are supplemented with missing values, logarithmically processed, and normalized and converted into supervised data to obtain training samples, including: For the time series traffic sample, supplement the missing values ​​to obtain a filled time series traffic sample; Perform logarithmic processing on the filled time series traffic samples to obtain logarithmically converted time series traffic samples; The time series traffic samples after logarithmic transformation are normalized to obtain normalized time series traffic samples; For the normalized time series traffic samples, based on the sliding window strategy, starting from the left side of the normalized time series traffic samples, the window is gradually slid to the right according to the time step to generate several input sequences and the target value corresponding to each input sequence. The target value is used as the label of the corresponding input sequence to generate training samples; among which, the target value of each input sequence is the traffic value at the next time point of the last time point in the input sequence.

5. The flow prediction method of the power distribution communication network according to claim 4, characterized in that: The missing values ​​of the time series traffic samples are supplemented by the following formula: in, It represents the missing value of the flow data of the distribution communication network at day i, hour j, and minute k; It represents the flow data of the power distribution communication network at the i-1 day, the jth hour, and the kth minute; It represents the traffic data of the power distribution communication network at the i+1th day, the jth hour, and the kth minute.

6. The flow prediction method of the power distribution communication network according to claim 5, characterized in that: The padded time series traffic samples are logarithmically processed using the following formula: x′=log(1+x); Among them, x is the traffic data in the filled time series traffic sample, and x' is the traffic data obtained after logarithmic transformation.

7. The flow prediction method of the power distribution communication network according to claim 6, characterized in that: The logarithmically transformed time series traffic samples are normalized using the following formula: Among them, x * The traffic data in the time series traffic sample after normalization, x min is the minimum value of the flow data in the time series flow sample after logarithmic transformation, x man It is the maximum value of the flow data in the time series flow sample after logarithmic transformation.

8. A flow prediction device for a power distribution communication network, characterized in that: include: Data acquisition module and model prediction module; The data acquisition module is used to acquire the historical time series flow data of the power distribution communication network to be predicted; The model prediction module is used to input the historical time series flow data into a preset flow prediction model, so that the flow prediction model can extract features from the historical time series flow data through a built-in time convolutional neural network to obtain time series flow features, and input the time series flow features into a built-in long short-term memory network, so that the long short-term memory network can determine the predicted flow of the distribution communication network to be predicted in a future time period based on the time series flow features.

9. A terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the flow prediction method for the distribution communication network as described in any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute the flow prediction method for the power distribution communication network according to any one of claims 1 to 7.

Images