Nuclear power material purchase demand prediction method and system based on time convolutional network

Abstract

The invention discloses a nuclear power material purchase demand prediction method and system based on a time convolutional network, and relates to the technical field of time series data analysis, and the method comprises the steps: collecting a data set, carrying out the preprocessing of the data set, classifying the data set, setting a representation mode of the input data set, and constructing a time convolutional neural network model. A data set is divided, model training is performed by using the divided data set, time sequence data is input into a trained model, and a material purchasing demand prediction value is output through multi-layer convolution, residual connection and full connection layer processing. According to the method, the time sequence data features are deeply learned through expansion convolution and residual connection, the ability of capturing the complex time dependency relationship by the model is improved, gradient flow in a deep network is promoted through introduction of the residual connection, the gradient disappearance problem is avoided, and the training efficiency and the learning ability of the complex features of the model are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of time series data analysis, and in particular to a method and system for predicting nuclear power material procurement demand based on a time convolutional network. Background Art

[0002] In modern supply chain management, material procurement demand forecasting is a key link to ensure supply chain efficiency and reduce operating costs. Traditional demand forecasting methods mainly include time series analysis, regression analysis, and machine learning models. Although these methods have achieved certain results in practical applications, they still have many limitations and problems.

[0003] First, traditional time series methods such as moving average, exponential smoothing, and autoregressive integrated moving average (ARIMA) perform well in dealing with simple seasonality and trends, but their prediction accuracy is low when facing complex nonlinear time series data. In addition, these methods have strict assumptions on data and are difficult to deal with outliers and missing values ​​in actual data. Second, regression analysis methods rely on the assumption of linear relationships. When there are nonlinear relationships in demand data, the predictive ability of regression models will be significantly reduced. At the same time, regression models rely heavily on feature engineering, and improper feature selection will lead to poor model performance. For high-dimensional data, the computational complexity of regression models will also increase significantly, affecting their application in big data environments. Third, machine learning methods such as support vector machines, random forests, and gradient boosting trees have achieved certain success in demand forecasting, but these methods usually need to split and reconstruct data when dealing with time series data, which increases the complexity of model design. In addition, traditional machine learning methods are not good at capturing long-term dependencies and are difficult to fully utilize the temporal characteristics of data.

[0004] When processing time series data, recurrent neural network (RNN) is a commonly used deep learning algorithm, especially long short-term memory network (LSTM) and gated recurrent unit (GRU), which solve the gradient vanishing and gradient exploding problems to a certain extent. However, RNN takes a long time to train and is difficult to parallelize, which limits its application on large-scale data. In addition, when processing ultra-long sequence data, the memory capacity and prediction performance of the RNN model are still challenged.

[0005] Based on the limitations of the above existing technologies, the temporal convolutional network (TCN), as an emerging time series prediction method, has shown great application potential. TCN can effectively capture long-term dependencies through convolution and dilated convolution operations, and has the advantage of parallel computing, which significantly improves the training speed and prediction performance of the model. TCN has shown higher stability and accuracy when processing large-scale complex time series data, and has gradually become an important tool in the field of time series prediction.

[0006] Although TCN has shown excellent performance in time series forecasting, its application research in material procurement demand forecasting is still relatively limited. Therefore, further exploration of material procurement demand forecasting methods based on TCN aims to solve the shortcomings of traditional methods in processing complex time series data, improve the accuracy and stability of demand forecasting, and provide more scientific and efficient decision support for supply chain management. Summary of the invention

[0007] In view of the above-mentioned existing problems, the present invention provides a method and system for predicting nuclear power material procurement demand based on a temporal convolutional network, so as to solve the problems of poor accuracy and stability in processing complex time series data in the prior art.

[0008] In order to solve the above technical problems, a nuclear power material procurement demand prediction method based on temporal convolutional network is proposed, including:

[0009] Collect data sets, preprocess them, classify them, and set the representation of the input data sets; build a temporal convolutional neural network model, divide the data sets, and use the divided data sets to train the model; input the time series data into the trained model, and after processing through multiple layers of convolution, residual connection, and fully connected layers, output the forecast value of material procurement demand.

[0010] As a preferred solution of the method for predicting nuclear power material procurement demand based on a temporal convolutional network described in the present invention, the preprocessing includes collecting a data set and preprocessing the data set.

[0011] The dataset includes timestamps, historical sales data, historical inventory levels, historical purchasing data, market prices, and economic indicators.

[0012] The preprocessing includes missing value filling, outlier processing and data normalization; the missing value filling includes filling the missing values ​​with the mean of the observed values ​​in the data set; the outlier processing includes identifying outliers in the data set using the triple standard method, and when an observed value deviates from the mean value by more than three times the standard deviation, the observed value is determined to be an outlier.

[0013] As a preferred solution of the nuclear power material procurement demand prediction method based on time convolutional network described in the present invention, the classification of the data set includes collecting the data set, preprocessing the data set, and dividing the collected data set into a training set, a validation set and a test set, and setting the representation method of the input data set.

[0014] The setting of the representation mode of the input data set includes setting the representation mode of the input data to a time window sliding method, and converting the time series data into an input matrix and a target vector.

[0015] As a preferred solution of the nuclear power material procurement demand forecasting method based on time convolutional network described in the present invention, the time convolutional neural network model includes designing a network architecture with multiple layers of time convolutional layers, using extended convolution for each convolutional layer, performing residual connection, setting network parameters, and performing network layer convolution.

[0016] The network architecture includes an input layer, a convolutional layer, an activation function, a residual connection and a full connection; the input layer includes creating a sliding window, extracting a fixed-length sequence from the original data, converting the data in each sliding window into a feature matrix, and adjusting the feature matrix to the input format specified by the model.

[0017] The convolution layer is included in the extended convolution layer, and the receptive field of the convolution kernel is increased layer by layer. The formula is expressed as:

[0018]

[0019] Among them, H l is the output of layer l, is the convolution kernel weight of the lth layer, H l-1 is the output of the l-1th layer, b l is the bias of the lth layer, d l is the expansion factor and σ is the activation function.

[0020] As a preferred solution of the nuclear power material procurement demand prediction method based on time convolutional network described in the present invention, the residual connection includes, after the output of each convolutional layer, adding the input features to the output features of the convolutional layer and setting the network parameters.

[0021] The residual connection formula is expressed as:

[0022]

[0023] in, is the output after residual connection, H l is the output of the lth layer, H l-1 is the output of the l-1th layer.

[0024] The network parameter setting includes setting the number of convolution layers, convolution kernel size, expansion factor and activation function; the expansion factor is expressed as:

[0025] d l =2 l-1

[0026] Among them, d l is the expansion factor, and l is the number of network architecture layers.

[0027] The activation function is expressed as:

[0028] σ=max(0,Conv(x))

[0029] Among them, σ is the activation function and Conv(x) is the output of the convolutional layer.

[0030] As a preferred solution of the nuclear power material procurement demand prediction method based on time convolutional network described in the present invention, the full connection includes flattening the feature map after multi-layer convolution into a one-dimensional vector, and mapping it through a fully connected layer, and finally outputting the predicted material demand.

[0031] The output material demand formula is expressed as:

[0032]

[0033] Among them, y is the material demand output by the fully connected layer, is the weight of the fully connected layer, is the bias of the fully connected layer, It is the output after three layers of convolution.

[0034] As a preferred solution of the nuclear power material procurement demand prediction method based on time convolutional network described in the present invention, the model training includes using training set data to perform model training through back propagation algorithm and Adam optimizer, and minimizing the mean square error loss function.

[0035] The back propagation algorithm formula is:

[0036]

[0037] Among them, L is the number of convolution layers, z (l) is the input of the lth convolutional layer, a (l) is the output of the lth convolutional layer after the activation function, z (l+1) is the input of the l+1th convolutional layer.

[0038] The Adam optimizer formula is expressed as:

[0039]

[0040] Among them, θ t+1 is the updated parameter after the t+1th round of iteration, θ t is the updated parameter after the tth round of iteration, η is the learning rate, is the first-order moment estimate of the gradient, is the second-order moment estimate of the gradient, and ∈ is a hyperparameter.

[0041] The loss function formula is:

[0042]

[0043] Among them, L is the loss function, N is the number of samples, and y i is the true value, is the predicted value, and i is the variable index.

[0044] Another object of the present invention is to provide a nuclear power material procurement demand forecasting system based on a temporal convolutional network. The present invention improves the accuracy of procurement plans, avoids additional expenses for excessive and emergency procurement, and improves decision-making efficiency and scientificity. The system of the present invention accurately predicts future material demand by analyzing time series data such as historical sales and inventory, and provides solid data support for the procurement plans of nuclear power companies, thereby optimizing inventory management, reducing inventory backlogs and out-of-stock risks, improving turnover rates, and achieving cost savings.

[0045] As a preferred solution of the nuclear power material procurement demand forecasting system based on the time convolutional network described in the present invention, it is characterized by including a data preprocessing module, a time convolutional neural network model building module, a residual connection module and a model training module.

[0046] The data preprocessing module is used to collect the original data set, clean, convert and format it, and divide the preprocessed data set into a training set, a validation set and a test set to support model training, tuning and evaluation.

[0047] The temporal convolutional neural network model building module is used to design and build a temporal convolutional neural network model to learn features from time series data and perform predictions.

[0048] The residual connection module is used to enhance the training stability and performance of the model, allowing the gradient to flow directly to deeper layers of the network.

[0049] The model training module is used to train the model using the training set data through a back propagation algorithm and an optimizer, and to optimize the model parameters by minimizing the loss function.

[0050] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the method described in the method for predicting nuclear power material procurement demand based on a temporal convolutional network are implemented.

[0051] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the method described in the method for predicting nuclear power material procurement demand based on a temporal convolutional network are implemented.

[0052] The beneficial effects of the present invention are as follows: the present invention ensures the quality and consistency of the model input data by collecting and preprocessing the data set, improves the training efficiency and prediction accuracy, and reduces errors; the missing value filling, outlier processing and data normalization in the preprocessing enhance the model's ability to interpret the data, avoid oversensitivity, and ensure the stability of the training process; the classification of the data set and the setting of the input representation method provide structured input for model training and improve the prediction accuracy; the constructed multi-layer time convolutional neural network model deeply learns the time series data features through extended convolution and residual connection, and improves the model's ability to capture complex time dependencies; the introduction of residual connection promotes the gradient flow in the deep network, avoids the gradient vanishing problem, and improves the training efficiency and the learning ability of the model's complex features; the setting of the fully connected layer realizes the direct conversion of features to prediction results, further improves the accuracy and efficiency of the prediction, and the back propagation algorithm and Adam optimizer in the model training process optimize the parameters and accelerate the convergence speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0054] Figure 1 An overall flow chart of a method for predicting nuclear power material procurement demand based on a temporal convolutional network provided for an embodiment of the present invention.

[0055] Figure 2 A system solution flow chart of a nuclear power material procurement demand forecasting system based on a temporal convolutional network provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0058] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an embodiment that is mutually exclusive with other embodiments, either individually or selectively.

[0059] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0060] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0061] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0062] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides a method for predicting nuclear power material procurement demand based on a temporal convolutional network, comprising:

[0063] S1: Collect data sets, preprocess and classify them, and set the representation method of input data sets.

[0064] The preprocessing includes collecting a data set and preprocessing the data set.

[0065] The dataset includes timestamps, historical sales data, historical inventory levels, historical purchasing data, market prices, and economic indicators.

[0066] The preprocessing includes missing value filling, outlier processing and data normalization; the missing value filling includes filling the missing values ​​with the mean of the observed values ​​in the data set; the outlier processing includes identifying outliers in the data set using the triple standard method, and when an observed value deviates from the mean value by more than three times the standard deviation, the observed value is determined to be an outlier; the data normalization includes normalizing the data to the interval [0,1], that is, the standardized mean is 0 and the standard deviation is 1.

[0067] The classification of the data set includes collecting the data set, preprocessing the data set, dividing the collected data set into a training set, a validation set and a test set, and setting a representation method of the input data set.

[0068] The setting of the representation mode of the input data set includes setting the representation mode of the input data to a time window sliding method, and converting the time series data into an input matrix and a target vector.

[0069] S2: Build a temporal convolutional neural network model, divide the data set, and use the divided data set for model training.

[0070] Furthermore, the model training includes using the training set data to perform model training through a back propagation algorithm and an Adam optimizer, and minimizing a mean square error loss function.

[0071] The back propagation algorithm formula is:

[0072]

[0073] Among them, L is the number of convolution layers, z (l) is the input of the lth convolutional layer, a (l) is the output of the lth convolutional layer after the activation function, z (l+1) is the input of the l+1th convolutional layer.

[0074] The Adam optimizer formula is expressed as:

[0075]

[0076] Among them, θ t+1 is the updated parameter after the t+1th round of iteration, θ t is the updated parameter after the tth round of iteration, η is the learning rate, is the first moment estimate of the gradient, is the second-order moment estimate of the gradient, and ∈ is a hyperparameter.

[0077] The loss function formula is:

[0078]

[0079] Among them, L is the loss function, N is the number of samples, and y i is the true value, is the predicted value, and i is the variable index.

[0080] S3: Input the time series data into the trained model, and after processing through multiple layers of convolution, residual connection and fully connected layers, output the forecast value of material procurement demand.

[0081] Furthermore, the temporal convolutional neural network model includes designing a network architecture with multiple layers of temporal convolutional layers, using extended convolution for each convolutional layer, performing residual connection, setting network parameters, and performing network layer convolution.

[0082] The network architecture includes an input layer, a convolutional layer, an activation function, a residual connection and a full connection; the input layer includes creating a sliding window, extracting a fixed-length sequence from the original data, converting the data in each sliding window into a feature matrix, and adjusting the feature matrix to the input format specified by the model.

[0083] The convolution layer is included in the extended convolution layer, and the receptive field of the convolution kernel is increased layer by layer. The formula is expressed as:

[0084]

[0085] Among them, H l is the output of layer l, is the convolution kernel weight of the lth layer, H l-1 is the output of the l-1th layer, b l is the bias of the lth layer, d l is the expansion factor and σ is the activation function.

[0086] Furthermore, the residual connection includes, after the output of each convolutional layer, adding the input features to the output features of the convolutional layer and setting network parameters.

[0087] The residual connection formula is expressed as:

[0088]

[0089] in, is the output after residual connection, H l is the output of the lth layer, H l-1 is the output of the l-1th layer.

[0090] The network parameter setting includes setting the number of convolution layers, convolution kernel size, expansion factor and activation function; the expansion factor is expressed as:

[0091] d l =2 l-1

[0092] Among them, d l is the expansion factor, and l is the number of network architecture layers.

[0093] The activation function is expressed as:

[0094] σ=max(0,Conv(x))

[0095] Among them, σ is the activation function and Conv(x) is the output of the convolutional layer.

[0096] The full connection includes flattening the feature map after multiple layers of convolution into a one-dimensional vector, mapping it through a fully connected layer, and finally outputting the predicted material demand.

[0097] The output material demand formula is expressed as:

[0098]

[0099] Among them, y is the material demand output by the fully connected layer, is the weight of the fully connected layer, is the bias of the fully connected layer, It is the output after three layers of convolution.

[0100] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0101] Example 2 is an embodiment of the present invention, which provides a method for predicting nuclear power material procurement demand based on a temporal convolutional network. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0102] The trained temporal convolutional network can predict the demand for material procurement by inputting recent time series data. Using the sliding window method to input these features, the temporal convolutional network captures the complex patterns and long-term dependencies in the data through multiple layers of convolution and extracts features of different time scales. After the convolutional layer outputs, the residual connection is used to ensure the stability of the model. Finally, the fully connected layer integrates the features and outputs the corresponding target variable as the demand prediction value, which represents the material demand at this time point.

[0103] The prediction results are shown in the following table:

[0104]

[0105]

[0106] In the experiment, the test set was predicted by using the TCN method and the LSTM method proposed in the present invention, and then the effects were compared. From the above results, we can see that the predicted value of TCN is closer to the true value than the predicted value of LSTM, and the predicted value of LSTM is far from the true value as a whole, and the effect is poor.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0108] Example 3, reference Figure 2 , which is the third embodiment of the present invention, provides a nuclear power material procurement demand forecasting system based on a temporal convolutional network, including a data preprocessing module 100, a temporal convolutional neural network model building module 200, a residual connection module 300 and a model training module 400.

[0109] The data preprocessing module 100 is used to collect the original data set, clean, convert and format it, and divide the preprocessed data set into a training set, a validation set and a test set to support model training, tuning and evaluation.

[0110] The temporal convolutional neural network model construction module 200 is used to design and construct a temporal convolutional neural network model to learn features from time series data and perform predictions.

[0111] The residual connection module 300 is used to enhance the training stability and performance of the model through residual connections, allowing the gradient to flow directly to deeper layers of the network.

[0112] The model training module 400 is used to train the model using the training set data through a back propagation algorithm and an optimizer, and to optimize the model parameters by minimizing the loss function.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0114] Embodiment 4, the fourth embodiment of the present invention, is different from the first three embodiments in that:

[0115] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0116] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0117] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0118] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A method for predicting nuclear power material procurement demand based on a temporal convolutional network, characterized by: include, Collect data sets, preprocess and classify them, and set the representation of input data sets; Build a temporal convolutional neural network model, divide the data set, and use the divided data set for model training; The time series data is input into the trained model, and after being processed by multiple layers of convolution, residual connection and full connection layers, the predicted value of material procurement demand is output.

2. The method for predicting nuclear power material procurement demand based on a temporal convolutional network according to claim 1, characterized in that: The preprocessing includes collecting a data set and preprocessing the data set; Said data set includes timestamps, historical sales data, historical inventory levels, historical purchasing data, market prices and economic indicators; The preprocessing includes missing value filling, outlier processing and data normalization; the missing value filling includes filling the missing values ​​with the mean of the observed values ​​in the data set; the outlier processing includes identifying outliers in the data set using the triple standard method, and when an observed value deviates from the mean value by more than three times the standard deviation, the observed value is determined to be an outlier.

3. The method for predicting nuclear power material procurement demand based on a temporal convolutional network as claimed in claim 2, characterized in that: The classification of the data set includes collecting the data set, preprocessing the data set, dividing the collected data set into a training set, a validation set and a test set, and setting a representation method of the input data set; The setting of the representation mode of the input data set includes setting the representation mode of the input data to a time window sliding method, and converting the time series data into an input matrix and a target vector.

4. The method for predicting nuclear power material procurement demand based on a temporal convolutional network as claimed in claim 3 is characterized by: The temporal convolutional neural network model includes designing a network architecture of multiple layers of temporal convolutional layers, using dilated convolution for each convolutional layer, performing residual connection, setting network parameters, and performing network layer convolution; The network architecture includes an input layer, a convolution layer, an activation function, a residual connection and a full connection; the input layer includes creating a sliding window, extracting a fixed-length sequence from the original data, converting the data in each sliding window into a feature matrix, and adjusting the feature matrix to an input format specified by the model; The convolution layer is included in the extended convolution layer, and the receptive field of the convolution kernel is increased layer by layer. The formula is expressed as: Among them, H l is the output of layer l, is the convolution kernel weight of the lth layer, H l-1 is the output of the l-1th layer, b l is the bias of the lth layer, d l is the expansion factor and σ is the activation function.

5. The method for predicting nuclear power material procurement demand based on a temporal convolutional network as claimed in claim 4, characterized in that: The residual connection includes, after the output of each convolutional layer, adding the input features to the output features of the convolutional layer and setting network parameters; The residual connection formula is expressed as: in, is the output after residual connection, H l is the output of the lth layer, H l-1 is the output of the l-1th layer; The network parameter setting includes setting the number of convolution layers, convolution kernel size, expansion factor and activation function; the expansion factor is expressed as: d l ＝2 l-1 Among them, d l is the expansion factor, l is the number of network architecture layers; The activation function is expressed as: σ=max(0,Conv(x)) Among them, σ is the activation function and Conv(x) is the output of the convolutional layer.

6. The method for predicting nuclear power material procurement demand based on a temporal convolutional network as claimed in claim 5, characterized in that: The full connection includes flattening the feature map after multiple layers of convolution into a one-dimensional vector, mapping it through a fully connected layer, and finally outputting the predicted material demand; The output material demand formula is expressed as: Among them, y is the material demand output by the fully connected layer, is the weight of the fully connected layer, is the bias of the fully connected layer, It is the output after three layers of convolution.

7. The method for predicting nuclear power material procurement demand based on a temporal convolutional network as claimed in claim 6, characterized in that: The model training includes using the training set data to perform model training through a back propagation algorithm and an Adam optimizer, and minimizing a mean square error loss function; The back propagation algorithm formula is: Among them, L is the loss function, z (l) is the input of the lth convolutional layer, a (l) is the output of the lth convolutional layer after the activation function, z (l+1) is the input of the l+1th convolutional layer; The Adam optimizer formula is expressed as: Among them, θ t+1 is the updated parameter after the t+1th round of iteration, θ t is the updated parameter after the tth round of iteration, η is the learning rate, is the first moment estimate of the gradient, is the second-order moment estimate of the gradient, ∈ is a hyperparameter; The loss function formula is: Among them, L is the loss function, N is the number of samples, and y i is the true value, is the predicted value, and i is the variable index.

8. A system using the method for predicting nuclear power material procurement demand based on a temporal convolutional network as claimed in any one of claims 1 to 7, characterized in that: It includes data preprocessing module, temporal convolutional neural network model building module, residual connection module and model training module; The data preprocessing module is used to collect the original data set, clean, convert and format it, and divide the preprocessed data set into a training set, a validation set and a test set to support model training, tuning and evaluation; The temporal convolutional neural network model building module is used to design and build a temporal convolutional neural network model to learn features from time series data and perform predictions; The residual connection module is used to enhance the training stability and performance of the model, allowing the gradient to flow directly to deeper layers of the network; The model training module is used to train the model using the training set data through a back propagation algorithm and an optimizer, and to optimize the model parameters by minimizing the loss function.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for predicting nuclear power material procurement demand based on a temporal convolutional network described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting nuclear power material procurement demand based on a temporal convolutional network described in any one of claims 1 to 7 are implemented.

Images