A steam pipeline network time series prediction method, system and computer equipment

Abstract

The present application discloses a steam network time series prediction method, system and computer equipment, which relates to the field of time series prediction. The method includes: obtaining the operation data of a thermal power plant within a set time period to obtain the operation time series data; performing feature engineering processing on the operation time series data to obtain an operation feature set; the operation feature set includes multiple features and the time point corresponding to each feature; according to the operation feature set, a steam network prediction model is used to determine the steam flow at the transmission end point of the steam network in the thermal power plant in the future period to determine the heat transfer performance of the steam network; the steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer. The present application improves the accuracy of steam network time series prediction.

Description

Technical Field

[0001] The present application relates to the field of time series prediction, and in particular to a method, system and computer equipment for predicting time series of a steam pipeline network. Background Art

[0002] In order to improve the heat transfer efficiency in thermal power plants, steam pipe networks are usually used to transfer heat from the heat source to the heat user site hundreds or even thousands of meters away. The heat transfer efficiency of the steam pipe network is highly dependent on its own conditions such as the outer diameter of the pipe, the transmission distance, the thermal conductivity, density, specific heat capacity, thickness of the insulation layer, as well as the steam pressure, temperature, external environment, etc., and has problems such as large system inertia, large time lag, and large disturbances.

[0003] In a thermal power plant, there is a strong temporal relationship between external user demands. For example, there is high demand during the day (8:00-17:00), very low demand in the early morning (23:00-6:00), and medium demand in other time periods. However, the current technical solution is to directly input the original data into a machine learning model (such as a tree model, a fully connected neural network, etc.), without considering the temporal relationship between the data. In addition, the model structure is simple, and the generalization under complex working conditions is insufficient. The long-distance prediction performance is poor, and it cannot meet the current thermal power plant's accuracy requirements for steam pipeline prediction systems under large-scale, multi-user demands.

[0004] Furthermore, the prior art generally predicts the steam network based on a Bayesian fully connected neural network model, learns the historical data of the steam network through the fully connected neural network, and introduces the Bayesian theorem to convert each neural network parameter into a set of probability distributions to improve the generalization ability of the model. However, the fully connected neural network structure is relatively simple and usually has a small number of parameters. The learning ability and generalization ability of the model are average, and the effect on long-distance time series prediction problems is poor. After the introduction of Bayesian, the generalization ability of the model can be improved to a certain extent, but the effect on time series prediction problems and the learning ability of the model have no obvious improvement. In addition, since the Bayesian theorem P(A|B)=(P(B|A)*P(A)) / P(B) introduces a division operation, it is not conducive to the accelerated parallel operation of the model on the graphics processing unit (GPU), resulting in low model operation efficiency. Summary of the invention

[0005] The purpose of this application is to provide a steam pipeline network timing prediction method, system and computer equipment, which can improve the accuracy of steam pipeline network timing prediction.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a steam network time series prediction method, comprising:

[0008] Obtain the operation data of the thermal power plant within a set time period and obtain the operation sequence data;

[0009] Performing feature engineering processing on the running time series data to obtain a running feature set; the running feature set includes multiple features and a time point corresponding to each feature;

[0010] According to the operation feature set, a steam network prediction model is used to determine the transmission terminal steam flow of the steam network in the thermal power plant in the future period to determine the heat transfer performance of the steam network; the steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer; the steam network prediction model is pre-trained using a training sample set, and each training sample in the training sample set includes a sample operation feature set and the corresponding actual transmission terminal steam flow of the steam network.

[0011] In a second aspect, the present application provides a steam network time series prediction system, comprising:

[0012] The data preparation module is used to obtain the operation data of the thermal power plant within a set time period and obtain the operation sequence data;

[0013] A feature engineering module, used to perform feature engineering processing on the running time series data to obtain a running feature set; the running feature set includes multiple features and a time point corresponding to each feature;

[0014] A model prediction module is used to determine the transmission terminal steam flow of the steam network in the thermal power plant in a future period of time using a steam network prediction model based on the operation feature set, so as to determine the heat transfer performance of the steam network; the steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer; the steam network prediction model is pre-trained using a training sample set, and each training sample in the training sample set includes a sample operation feature set and the corresponding actual transmission terminal steam flow of the steam network.

[0015] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the steam network timing prediction methods described above.

[0016] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0017] The present application provides a steam pipeline network time series prediction method, system and computer equipment. Aiming at the complex working conditions of the steam pipeline network in the thermal power plant, by collecting the operating data of the thermal power plant within a set time period, a complete feature engineering method for time series characteristics is proposed, and the prediction problem of the steam pipeline network is converted into a time series prediction problem. It can focus on the performance of the operating data in different time periods and the relationship between different operating data. The law contained in the data is mined through the feature engineering method, and then the time series relationship of the user's needs in different time periods is considered. A steam pipeline network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer, which can significantly improve the accuracy of time series prediction, thereby effectively improving the prediction effect of the steam pipeline network performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is an application environment diagram of a steam network time series prediction method in one embodiment of the present application;

[0020] Figure 2 A schematic flow chart of a steam network timing prediction method provided in one embodiment of the present application;

[0021] Figure 3 An overall architecture diagram of a steam network time series prediction method provided in one embodiment of the present application;

[0022] Figure 4 A schematic diagram of functional modules of a steam network timing prediction system provided in one embodiment of the present application;

[0023] Figure 5 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0025] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0026] The steam network time series prediction method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the runtime sequence data to be processed to the server 104. After the server 104 receives the runtime sequence data to be processed, for the runtime sequence data to be processed, the server 104 performs feature engineering processing on the runtime sequence data to obtain an operation feature set. According to the operation feature set, the steam network prediction model is used to determine the transmission terminal steam flow of the steam network in the thermal power plant in the future period to determine the heat transfer performance of the steam network. The server 104 can feedback the transmission terminal steam flow of the steam network in the thermal power plant in the future period to the terminal 102. In addition, in some embodiments, the steam pipeline network timing prediction method can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly perform timing prediction on the running timing data to be processed, or the server 104 can obtain the running timing data to be processed from the data storage system and perform timing prediction on the running timing data to be processed.

[0027] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0028] In an exemplary embodiment, Figure 2 and Figure 3 As shown, a steam network timing prediction method is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps 201 to 203. Among them:

[0029] Step 201, obtaining the operation data of the thermal power plant within a set time period to obtain the operation sequence data.

[0030] In the present application, the operating time series data includes the outer diameter of the steam pipeline network, the transmission distance of the steam pipeline network, the thermal conductivity of the insulation layer, the density of the insulation layer, the specific heat capacity of the insulation layer, the thickness of the insulation layer, the steam pressure, the steam temperature, the external ambient temperature, the steam flow rate at the transmission starting point, and the steam flow rate at the transmission end point at each time point within the set time period.

[0031] Specifically, based on the operating data in the thermal power plant, 11 point data are collected, including the outer diameter of the steam pipeline network, the transmission distance of the steam pipeline network, the thermal conductivity of the insulation layer, the density of the insulation layer, the specific heat capacity of the insulation layer, the thickness of the insulation layer, the steam pressure, the steam temperature, the external ambient temperature, the steam flow rate at the transmission starting point, and the steam flow rate at the transmission end point. The data are stored in the database as initial data. After the missing values ​​of the initial data are processed, they are stored in the database again as model data, i.e., runtime series data.

[0032] The above 11 point data are all numerical features. The runtime data obtained in step 201 of the present application is two-dimensional data, with the horizontal axis being the 11 point data and the vertical axis being the time point. If the predicted object is a steam pipe network, the vertical axis is each time point within the set time period. If the predicted object is multiple steam pipe networks, the vertical axis is the number of steam pipe networks multiplied by the number of time points within the set time period.

[0033] Step 202: Perform feature engineering processing on the runtime data to obtain a runtime feature set, wherein the runtime feature set includes multiple features and a time point corresponding to each feature.

[0034] Specifically, step 202 includes two parts: feature generation and timing characteristics analysis:

[0035] (1) Performing feature generation processing on the runtime time series data to obtain preliminary time series features, wherein the feature generation processing includes date information extraction, distribution change, feature derivation and hysteresis sliding processing.

[0036] 1) For the time index in the runtime data, extract the holiday information, hour information, and season information in the date data with one click. For example, if the current time index is 2024-04-0312:10:00, you can retrieve whether 2024-04-03 is a holiday, whether the current time is 12 o'clock, and the season it belongs to.

[0037] 2) Change the distribution of numerical features. For example, log transformation can effectively handle outliers and make the data closer to normal distribution.

[0038] 3) Feature derivation: a priori information can be added by performing arithmetic operations between different features. For example, feature derivation of pipeline outer diameter and transmission distance can generate four new features: transmission distance + steam pipeline outer diameter, transmission distance - steam pipeline outer diameter, transmission distance * steam pipeline outer diameter, and transmission distance / steam pipeline outer diameter.

[0039] 4) Lag sliding, for example, taking yesterday’s data at the same time as a new feature.

[0040] This application uses the above method to perform feature distribution change, feature derivation and hysteresis sliding on 11 point data to generate richer features.

[0041] (2) Performing time series characteristic analysis on the preliminary time series characteristics to determine the operating feature set, wherein the time series characteristic analysis includes autocorrelation detection and seasonal decomposition processing.

[0042] 1) Perform autocorrelation detection on the preliminary time series features to generate autocorrelation coefficients: Where N is the length of the preliminary time series feature, k∈(0,N), μ is the mean of the preliminary time series feature, and x t is the value of the point feature x at time t.

[0043] 2) Use seasonal decomposition method to decompose the preliminary time series characteristics into trend and seasonal data: Y t =T t +S t +R t , where Y t Represents the preliminary time series characteristics, T t It represents the trend term, which is usually obtained by local smoothing of the sequence. The local period is usually 1.5 times that of the seasonal term. t It represents the seasonal term, which is usually obtained by the local weighted average of the sequence minus the trend term. The local period is usually an odd number such as 3, 5, etc. t It is the residual term, which is the remaining value after removing the trend term and seasonal term from the preliminary time series characteristics, reflecting the random volatility of the preliminary time series characteristics.

[0044] The operating feature set obtained in step 202 of the present application is still two-dimensional data, the horizontal axis is the features generated by the 11 point data, and the vertical axis is the time point.

[0045] The 11 point data at each time point obtained in this application generated new features (including festival features, hourly features, seasonal features, conversion features, derivative features, lag features, trend items, seasonal items, etc.) after feature generation and time series characteristic analysis, which increased the richness of the data and played a beneficial role in mining the laws contained in the data and improving the model effect.

[0046] In this application, feature engineering is divided into two major functional points. The feature generation part can generate richer feature information, and the time series characteristic analysis can decompose the numerical features, which helps to improve the steam network prediction model's understanding of the data.

[0047] Step 203, based on the operation feature set, using a steam network prediction model, determine the transmission terminal steam flow of the steam network in the thermal power plant in a future period to determine the heat transfer performance of the steam network.

[0048] The steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer. The steam network prediction model is pre-trained using a training sample set, and each training sample in the training sample set includes a sample operation feature set and the actual transmission terminal steam flow of the corresponding steam network.

[0049] The steam network prediction model constructed in this application adopts the transformer structure that is currently widely used in time series prediction and natural language processing. A neural network containing two layers of transformer encoding layers is selected for modeling and improved for steam network prediction. The goal of the model is to minimize Among them, L is the target value, y i represents the actual steam flow at the end point of transmission at the i-th time point, Represents the predicted steam flow at the transmission end point at the i-th time point output by the steam network prediction model. The steam network prediction model is trained and solved using the stochastic gradient descent method, and the steam network prediction model is saved. The steam network prediction model is updated every hour and can capture changes in the on-site environment in a timely manner.

[0050] The steam network prediction model includes a vectorization module, a multi-head attention module, a first-layer normalization module, a hybrid expert module, and a second-layer normalization module connected in sequence. The first-layer normalization module is also connected to the vectorization module. The second-layer normalization module is also connected to the first-layer normalization module.

[0051] Step 203 specifically includes:

[0052] (1) The operation feature set is vectorized by the vectorization module to obtain a three-dimensional vector.

[0053] Specifically, the operation feature set is first sampled in time series to obtain preliminary three-dimensional data. The preliminary three-dimensional data is (m-2l+1)*n*l; wherein m is the number of time points in the operation feature set, n is the number of features in the operation feature set, and l is the sampling length, which is determined according to actual needs. That is, firstly, the m*n two-dimensional data is sampled in time series according to the time sequence, and the two-dimensional data after time series sampling becomes (m-2l+1)*n*l three-dimensional data.

[0054] Then the preliminary three-dimensional data is sliced ​​to obtain a three-dimensional vector. The three-dimensional vector is Among them, b is the number of samples input each time the steam network prediction model is trained, that is, the vector dimension of the input steam network prediction model is b*n*l, s is the size of the slice, and p is the interval step of each slice.

[0055] Since the goal of time series prediction is to understand the correlation of data between different time steps, but a single time step does not have semantic meaning, extracting local semantic information is crucial to analyzing the connection between features.

[0056] This application enhances locality and captures comprehensive semantic information that cannot be obtained at the point level by aggregating time steps into subsequence-level slices. The specific operation is: converting the vector from b*n*l to Convert to Models are in a three-dimensional format for easy manipulation.

[0057] (2) After the three-dimensional vector is divided into multiple parts by the multi-head attention module, the attention mechanism operation is performed on each of them to obtain multiple attention vectors, and the multiple attention vectors are spliced ​​to obtain a multi-head attention vector.

[0058] Specifically, the dimension of the three-dimensional vector is expressed as bn*p*lps, where lps represents Let t be the number of multi-head attention, the attention mechanism can be expressed as Among them, Q, K, and V are all vectorized matrices, which are obtained by linearly transforming the three-dimensional vector through three different fully connected layers. Their data dimensions are all bn*p*lps, d=lps, K T is the transpose of K. The multi-head attention mechanism is to split Q, K, and V into t parts in the lps dimension and then perform the attention mechanism operation separately. The final output of the multi-head attention can be expressed as concat(Attenion1…Attenion t), whose dimension is also bn*p*lps, and Attenion1 is the output of the first multi-head attention.

[0059] (3) The multi-head attention vector is normalized by the first-layer normalization module and added to the three-dimensional vector to obtain a preliminary normalized vector.

[0060] Specifically, the first-layer normalization module is used to perform normalization processing on the p*lps dimension of the multi-head attention vector bn*p*lps, and the original three-dimensional vector is added at the same time.

[0061] (4) Determine preliminary prediction data according to the preliminary normalized vector through the hybrid expert module.

[0062] This application uses a hybrid expert module to replace the original fully connected layer to effectively improve the model performance. The hybrid expert module consists of two parts: ① Expert list [expert 1, ..., expert z], each expert in the expert list is obtained by changing the output of the first layer normalization module through the fully connected layer. Among them, the fully connected layer of each expert is determined by its own random initialization vector value. These initialized vector values ​​generate the most suitable vector for each expert as the model's learning iterations. ② Expert probability selection, the output of the first layer normalization module is linearly changed and then Generate bn*p*z probability value vector [p1,...,p z ]. Among them, e j It is the result obtained by the j-th expert after linearly changing the output of the first-layer normalization module.

[0063] The expert list generated by ① is weightedly summed by probability. The specific process can be expressed in torch pseudo code as torch.sum(torch.multiply([bn*p*z*1],[bn*p*z*n]), dim=2) to achieve better results.

[0064] (5) The preliminary prediction data is normalized by the second-layer normalization module and added to the preliminary normalized vector, and finally output through a fully connected layer to obtain the transmission terminal steam flow of the steam pipeline network in the thermal power plant in the future period.

[0065] If the predicted object is a steam pipeline network, the output result of the second normalization module after the simple fully connected layer is the steam flow rate at the transmission end of this steam pipeline network at each time point in the future period. If the predicted object is multiple steam pipeline networks, the output result of the second normalization module after the simple fully connected layer is the steam flow rate at the transmission end of each steam pipeline network at each time point in the future period.

[0066] In addition, the steam network prediction model can use other transformer structures, such as using only the decoding layer, using the encoding layer and the decoding layer at the same time, and improving other parts of the transformer structure.

[0067] Aiming at the complex working conditions of the steam network in the thermal power plant, this application proposes a complete feature engineering method for time series characteristics by collecting the steam network and its related operating data, converting the prediction problem of the steam network into a time series prediction problem, focusing on the performance of the operating data in different time periods and the relationship between different operating data, mining the rules contained in the data through feature engineering methods, and then considering the time series relationship of user needs in different time periods. Based on vector slicing, hybrid expert strategy and transformer coding layer, a steam network prediction model is constructed, which can significantly improve the accuracy of time series prediction, is more suitable for the current time series prediction process of thermal power plant needs, and can effectively improve the prediction effect of steam network performance.

[0068] The present application also provides an application scenario, which applies the above-mentioned steam pipeline timing prediction method. Specifically: The steam pipeline timing prediction method provided in this embodiment can be applied in the scenario of improving the heat transfer efficiency of thermal power plants. In the process of improving the heat transfer efficiency of thermal power plants, first, the heat transfer efficiency of each steam pipeline in the thermal power plant, that is, the above-mentioned steam pipeline timing prediction method is used to predict the steam flow rate at the transmission end of the steam pipeline in the thermal power plant in the future period, and the heat transfer performance of each steam pipeline in the thermal power plant is determined according to the prediction results, and then the distribution of the steam pipeline in the thermal power plant is adjusted or the working parameters of the steam pipeline are adjusted to improve the heat transfer efficiency of the thermal power plant.

[0069] Based on the same inventive concept, the embodiment of the present application also provides a steam pipe network timing prediction system for implementing the steam pipe network timing prediction method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more steam pipe network timing prediction system embodiments provided below can refer to the limitations of the steam pipe network timing prediction method above, and will not be repeated here.

[0070] In an exemplary embodiment, Figure 4 As shown, a steam network time series prediction system is provided, including: a data preparation module 301, a feature engineering module 302 and a model prediction module 303.

[0071] The data preparation module 301 is used to obtain the operation data of the thermal power plant within a set time period to obtain the operation sequence data.

[0072] The feature engineering module 302 is used to perform feature engineering processing on the runtime data to obtain an operation feature set, which includes multiple features and a time point corresponding to each feature.

[0073] The model prediction module 303 is used to determine the steam flow rate of the transmission terminal of the steam network in the thermal power plant in the future period according to the operation feature set and using the steam network prediction model to determine the heat transfer performance of the steam network. The steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer. The steam network prediction model is pre-trained using a training sample set, and each training sample in the training sample set includes a sample operation feature set and the corresponding actual steam flow rate of the transmission terminal of the steam network.

[0074] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store video tag processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a steam pipeline timing prediction method is implemented.

[0075] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0076] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0077] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0078] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0079] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0080] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0081] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0082] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0083] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A steam network time series prediction method, characterized in that: The steam network time series prediction method comprises: Obtain the operation data of the thermal power plant within a set time period and obtain the operation sequence data; Performing feature engineering processing on the running time series data to obtain a running feature set; the running feature set includes multiple features and a time point corresponding to each feature; According to the operation feature set, a steam network prediction model is used to determine the steam flow rate at the transmission terminal of the steam network in the thermal power plant in the future period to determine the heat transfer performance of the steam network; the steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer; the steam network prediction model is pre-trained using a training sample set, and each training sample in the training sample set includes a sample operation feature set and the corresponding actual steam flow rate at the transmission terminal of the steam network; The steam network prediction model includes a vectorization module, a multi-head attention module, a first-layer normalization module, a hybrid expert module, and a second-layer normalization module connected in sequence; the first-layer normalization module is also connected to the vectorization module; the second-layer normalization module is also connected to the first-layer normalization module; According to the operation characteristic set, a steam network prediction model is used to determine the steam flow rate at the transmission end point of the steam network in the thermal power plant in the future period, specifically including: The operation feature set is vectorized by the vectorization module to obtain a three-dimensional vector, which specifically includes: performing time-series sampling on the operation feature set to obtain preliminary three-dimensional data; the preliminary three-dimensional data is (m-2l+1)*n*l; wherein m is the number of time points in the operation feature set, n is the number of features corresponding to each time point in the operation feature set, and l is the sampling length; slicing the preliminary three-dimensional data to obtain a three-dimensional vector; the three-dimensional vector is Wherein, b is the number of samples input into the steam network prediction model each time it is trained, s is the size of the slice, and p is the interval step length of each slice; After the multi-head attention module divides the three-dimensional vector into multiple parts, the attention mechanism operation is performed on each of them to obtain multiple attention vectors, and the multiple attention vectors are spliced ​​to obtain a multi-head attention vector; The multi-head attention vector is normalized by the first-layer normalization module, and added to the three-dimensional vector to obtain a preliminary normalized vector; Determining preliminary prediction data according to the preliminary normalized vector by the hybrid expert module; The preliminary prediction data is normalized by the second-layer normalization module and added to the preliminary normalized vector to obtain the transmission terminal steam flow of the steam pipeline network in the thermal power plant in the future period.

2. The steam network time series prediction method according to claim 1, characterized in that: The operating time series data includes the pipe outer diameter of the steam pipeline network, the transmission distance of the steam pipeline network, the thermal conductivity of the insulation layer, the density of the insulation layer, the specific heat capacity of the insulation layer, the thickness of the insulation layer, the steam pressure, the steam temperature, the external ambient temperature, the steam flow rate at the transmission starting point, and the steam flow rate at the transmission end point at each time point within the set time period.

3. The steam network time series prediction method according to claim 1, characterized in that: Perform feature engineering processing on the running series data to obtain a running feature set, which specifically includes: Performing feature generation processing on the running time series data to obtain preliminary time series features; A timing characteristic analysis is performed on the preliminary timing characteristics to determine an operation characteristic set.

4. The steam network time series prediction method according to claim 3, characterized in that: The feature generation process includes date information extraction, distribution change, feature derivation and hysteresis sliding processing.

5. The steam network time series prediction method according to claim 3, characterized in that: The time series characteristic analysis includes autocorrelation detection and seasonal decomposition processing.

6. The steam network time series prediction method according to claim 1, characterized in that: The multi-head attention vector is normalized by the first-layer normalization module, specifically including: Through the first layer normalization module in the multi-head attention vector Standardize the dimensions.

7. A steam network time series prediction system, applied to the steam network time series prediction method according to any one of claims 1 to 6, characterized in that: The steam network time series prediction system comprises: The data preparation module is used to obtain the operation data of the thermal power plant within a set time period and obtain the operation sequence data; A feature engineering module, used to perform feature engineering processing on the running time series data to obtain a running feature set; the running feature set includes multiple features and a time point corresponding to each feature; A model prediction module is used to determine the transmission terminal steam flow of the steam network in the thermal power plant in a future period of time using a steam network prediction model based on the operation feature set, so as to determine the heat transfer performance of the steam network; the steam network prediction model is constructed based on vector slicing, hybrid expert strategy and transformer coding layer; the steam network prediction model is pre-trained using a training sample set, and each training sample in the training sample set includes a sample operation feature set and the corresponding actual transmission terminal steam flow of the steam network.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the steam network timing prediction method described in any one of claims 1-6.

Images