Thermal power plant multi-time scale generating capacity prediction method and related device

Abstract

The invention discloses a thermal power plant multi-time scale generating capacity prediction method and a related device, and the method comprises the steps: obtaining the operation data of a thermal power plant, carrying out the data missing processing of the operation data of the thermal power plant, and constructing an input matrix with a mask; based on the input matrix with the mask, utilizing a reconstruction error constraint model to output a latent variable vector; according to the latent variable vector, based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained, and a final effective auxiliary feature set is obtained through screening from the K optimal lag auxiliary sequences; according to the final effective auxiliary feature set, the multi-time-scale generating capacity of the thermal power plant is predicted through a prediction network, and the method and the related device give consideration to data missing processing, time delay related selection and multi-scale information extraction at the same time to conduct multi-time-scale generating capacity prediction.

Description

Technical Field

[0001] This invention belongs to the field of intelligent prediction and optimal scheduling technology of power systems, and relates to a method and related devices for predicting the power generation of thermal power plants at multiple time scales. Background Technology

[0002] The power generation of thermal power plants is affected by a variety of factors, including fuel quality, boiler operating status, equipment parameters, environmental conditions, load changes, market pricing, and dispatch strategies. Its time-series characteristics are complex and significantly nonlinear. Furthermore, the following problems commonly exist in actual production and operation: (1) Data loss and noise interference are common: communication interruption of acquisition equipment, abnormal location, instrument error, etc. can cause time series data to be discontinuous, jump or partially missing. If used directly for modeling, the prediction effect will be significantly reduced.

[0003] (2) There are multi-scale lag relationships between variables: for example, the impact of coal quality changes on boiler steam parameters has several time step lags, and the response of load commands to power generation also has dynamic changes. Traditional models cannot explicitly identify the most effective lag period.

[0004] (3) Lack of effective auxiliary variable selection mechanism: In a large amount of unit operation data, it is difficult to automate which variables have the greatest predictive contribution to power generation and which variables have high correlation within a specific lag window.

[0005] Therefore, there is an urgent need for a prediction method that can simultaneously handle data missingness, time-delay correlation selection, and multi-scale information extraction to adapt to complex thermal power operation scenarios. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and related apparatus for predicting power generation at multiple time scales in thermal power plants. This method and apparatus simultaneously address issues such as missing data, time-delay correlation selection, and multi-scale information extraction for predicting power generation at multiple time scales.

[0007] To achieve the above objectives, this invention discloses a method for predicting the power generation of thermal power plants across multiple time scales, comprising: Acquire operational data from thermal power plants, perform data loss handling on the operational data, and construct a masked input matrix. ; Based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. ; According to the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained.

[0008] The final set of effective auxiliary features is obtained by filtering from the K optimal hysteresis auxiliary sequences. ; Based on the final effective auxiliary feature set Using prediction networks, we can predict the power generation of thermal power plants at multiple time scales.

[0009] Furthermore, the masked input matrix for:

[0010] in, Represents the mask matrix This represents element-wise multiplication. For the original multivariate data matrix, Built based on the operating data of thermal power plants; The mask matrix for:

[0011] in, Indicates the first The first variable An indicator of whether a time is valid. A value of 1 indicates that the condition is valid. A value of 0 indicates that the value is missing.

[0012] Furthermore, the local-global consistency loss of the reconstruction error constraint model for:

[0013] in, Represents Euclidean distance. This is the global trend vector. , This represents the global trend encoding function.

[0014] Furthermore, the step of using the latent variable vector The process of obtaining K optimal lag auxiliary sequences based on time-delay correlation spectrum projection is as follows: Construct a set of time delay values ,in, , Indicates the first Each time delay length; Based on latent variable vectors Candidate variable sequence Using time delay value set Generate several lag sequences ; Calculate the latent variables corresponding to the main predictor variables With each lag sequence The lag distance between them; Select the K lag sequences with the smallest lag distances. As K optimal lag auxiliary sequences.

[0015] Furthermore, the final effective auxiliary feature set is obtained by filtering from the K optimal lag auxiliary sequences. The process is as follows: The K optimal lag auxiliary sequences are input into the auxiliary sequence encoder to obtain K embedding vectors. ; Based on contrastive loss, from K embedding vectors Several effective embedding vectors are selected from the set, and these are used to construct the final effective auxiliary feature set. .

[0016] Furthermore, the contrast loss for:

[0017] in, Represents the embedding vector of the main sequence; " represents the vector dot product.

[0018] Furthermore, the prediction network is trained based on multi-head attention layers, bidirectional temporal convolutions, Transformer, GRU, or LSTM.

[0019] This invention discloses a multi-timescale power generation prediction system for thermal power plants, comprising: The acquisition module is used to acquire the operating data of thermal power plants, perform data missing handling on the operating data, and construct an input matrix with a mask. ; Reconstruction module, used based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. ; Projection module, used to project based on the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained.

[0020] The filtering module is used to filter the K optimal lag auxiliary sequences to obtain the final effective auxiliary feature set. ; The prediction module is used to predict based on the final effective auxiliary feature set. Using prediction networks, we can predict the power generation of thermal power plants at multiple time scales.

[0021] The present invention discloses a computer device, including 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 multi-timescale power generation prediction method for thermal power plants.

[0022] The present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multi-timescale power generation prediction method for thermal power plants.

[0023] The present invention has the following beneficial effects: In practical operation, the multi-timescale power generation prediction method and related device for thermal power plants described in this invention acquire the operating data of the thermal power plant, perform data missing processing on the operating data, and construct a masked input matrix. Based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. This automatically suppresses mask noise and improves the stability of the representation. Furthermore, based on the latent variable vector... Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained, and the final effective auxiliary feature set is obtained from the K optimal lag auxiliary sequences. This allows us to identify the auxiliary variables that contribute most to the prediction, enabling the model to extract correlations across time scales, based on the final set of effective auxiliary features. It utilizes a forecasting network to predict the power generation of thermal power plants at multiple time scales, applicable to minute-level, hour-level, and daily-level forecasts, and can support spot pricing, output planning, and AGC regulation. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0028] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0029] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0030] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0031] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0033] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0034] Example 1 refer to Figure 1 The multi-timescale power generation prediction method for thermal power plants described in this invention includes the following steps: 1) Acquire the operating data of thermal power plants and perform time-series preprocessing; 11) Multivariate data acquisition for thermal power plants; The following categories of raw data are obtained from the power plant's DCS (Distributed Control System), PLC (Programmable Logic Controller), and dispatching interface: Main steam parameters: main steam pressure and temperature; Boiler parameters: furnace negative pressure, burner opening and flue gas temperature; Auxiliary machine parameters: feedwater pump flow rate and fan speed; Fuel indicators: lower calorific value, ash content and volatile matter of coal fed into the furnace; Dispatch signals: load commands and AGC (Automatic Generation Control) adjustment signals; Actual power generation (predicted target).

[0035] The original multivariate data matrix is ​​constructed from all the original data. for:

[0036] in, For the first The variable in the first... The observation values ​​at each time point; This indicates the total number of variables, such as boiler pressure, coal supply, and main steam temperature. Indicates the length of the time step; 12) Perform data alignment and outlier identification; Because different sampling frequencies are different, a unified sampling interval (e.g., 5 min, 10 min) is required, and intermittent data or data with inconsistent timestamps should be aligned. When the value jump exceeds the set threshold, exceeds the range allowed by the mechanism, or a continuous constant value phenomenon occurs, it is judged as an outlier. Outliers are marked as missing and are not directly filled.

[0037] 13) Construct a missing mask; Construct a mask matrix for:

[0038] in, Indicates the first The first variable An indicator of whether a time is valid. A value of 1 indicates that the condition is valid. A value of 0 indicates that the value is missing.

[0039] Construct a masked input matrix for:

[0040] in, Represents the mask matrix This indicates element-wise multiplication.

[0041] 2) Multi-granularity information compression and structural encoding; 21) Local structure encoding; Using multi-layer convolutional or sliding window encoders, short-term trends are captured, such as the impact of short-term increases in feedwater flow on pressure and load, and rapid disturbances in furnace temperature and steam temperature, and the mean of latent variables is output. With variance :

[0042] in, Represents a vector of latent variables; Represents a multidimensional normal distribution; Represents the vector of latent variable means; Represents the variance vector of latent variables; This means placing the vector elements on the diagonal to generate a diagonal matrix.

[0043] 22) Global structure encoding; Build a global projector The entire sequence is transformed into a global structural representation, capturing intraday power generation patterns, unit scheduling rules, and slow-changing characteristics of the thermal system. Alignment loss is used to ensure that local features align with the global trend, resulting in a global trend vector. for:

[0044] in, This represents the global trend encoding function.

[0045] 23) Local-to-global alignment loss; By utilizing the performance of the reconstructed error constraint model under missing conditions, it becomes easier to capture fluctuation trends, long-term changes, and boiler operation balance relationships, ultimately outputting latent variables. It is used for downstream time-delay correlation analysis.

[0046] Constructing Local-Global Consistency Loss for:

[0047] in, This represents the Euclidean distance.

[0048] This loss allows the model to simultaneously consider local disturbances and the overall structure for safe operation.

[0049] 3) Projection of the time-delay correlation spectrum; This module is used to automatically find the most suitable lagged auxiliary variable for prediction from multiple candidate variables. The latent variable is then input. Output the optimal lag distance for each candidate variable.

[0050] 31) Set up multiple time-delay windows; Let the set of time delay values ​​be defined. for:

[0051] in, Indicates the first Each time lag length (e.g., 5 min: combustion disturbance affects short-term output; 15 min: lag of coal quantity change on steam parameters; 30–60 min: stage changes in the load curve). Indicates the time delay quantity.

[0052] 32) Generate time-delay sequences; For candidate variable sequences Generate multi-delay versions, where, in the time delay The shift sequence below :

[0053] in, Represents a time shift function; Alignment between sequences occurs at the most probable causal location.

[0054] 33) Calculate the dynamic time-bending distance; Calculate the latent variables corresponding to the main predictor variables With each lag sequence Minimum hysteresis distance between for:

[0055] in, Indicates the main sequence (e.g., unit output); This represents the dynamic time warping algorithm.

[0056] 34) Select the optimal lag; Select the K most important variables in terms of distance from the ranking as auxiliary variables, and construct the K optimal lag auxiliary sequences based on these. for:

[0057] in, Indicates the number of selections (e.g., 5, 10).

[0058] 4) Auxiliary sequence embedding and filtering; To further improve the quality of the auxiliary sequences, low-contribution variables are filtered out through embedding encoding and contrastive learning. The K optimal lag auxiliary sequences obtained in step 3) are input, and the simplified auxiliary sequence embedding is output.

[0059] 41) Embedded encoding; Encoder using auxiliary sequences , obtain variables Embedded vector for:

[0060] in, Representing variables The optimal time delay.

[0061] 42) Comparison of loss structures; Let highly correlated variables be positive samples and low-correlation variables be negative samples, where the positive sample set P and the negative sample set N are:

[0062]

[0063] Where P is the set of positive samples and N is the set of negative samples.

[0064] Construct contrast loss for:

[0065] in, Represents the embedding vector of the main sequence; " represents the vector dot product.

[0066] 43) Filtering and output of auxiliary sequence embeddings; By optimizing the comparison loss, the auxiliary sequence, which is highly consistent with the main sequence in terms of time-delay correlation, is represented by its distance from the main sequence embedding vector. Sequences that are closer to the main sequence are selected, while weakly correlated or uncorrelated sequences are selected further away. After training, based on the distance metric in the embedding space, the K closest auxiliary sequences to the main sequence are selected as the final effective auxiliary feature set. for:

[0067] in, Indicates the first Embedded representation of auxiliary variables; Indicates the number of auxiliary variables.

[0068] 5) Fusion prediction of main sequence-auxiliary sequence; 51) Feature fusion; Constructing a unified fusion feature vector for:

[0069] in, This represents a vector concatenation function; Indicates the index of the selected auxiliary variable.

[0070] 52) Construct a prediction network; Optional deep models can be used, such as multi-head attention layers, bidirectional temporal convolutions, Transformer / GRU / LSTM, etc.

[0071] 53) Output short-term / intraday / multi-period forecast values; Using deep networks Output at time Predicted power generation value for:

[0072] in, Indicates the predicted time index; Output forecast results, such as power generation in the next 15 minutes / 30 minutes / 60 minutes, intraday operating curves, and total power generation forecasts for the day, for use in dispatching applications, unit load tracking, and optimized operation.

[0073] This invention has the following characteristics: Maintaining robust performance even under data-missing conditions: The encoder automatically suppresses mask noise and improves the stability of the representation through a multi-granularity information compression (Glocal) constraint model.

[0074] Explicitly utilizing time-lag correlations: It can automatically search multiple lag windows to find the auxiliary variables that contribute most to the prediction, enabling the model to extract correlations across time scales.

[0075] Automatic variable selection and projection: No manual variable selection is required. The model automatically generates auxiliary sequences based on lagged correlations, improving prediction performance and interpretability.

[0076] The model can be trained end-to-end: all modules can be parameterized, without the need for prior data padding or pre-construction of lag features, which is beneficial for production system deployment.

[0077] Adaptable to multiple time scale forecasting scenarios: Suitable for minute-level, hour-level, and daily-level forecasting, and can support spot price quotations, power output planning, and AGC adjustment.

[0078] Example 2 The multi-timescale power generation prediction system for thermal power plants of the present invention includes: The acquisition module is used to acquire the operating data of thermal power plants, perform data missing handling on the operating data, and construct an input matrix with a mask. ; Reconstruction module, used based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. ; Projection module, used to project based on the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained.

[0079] The filtering module is used to filter the K optimal lag auxiliary sequences to obtain the final effective auxiliary feature set. ; The prediction module is used to predict based on the final effective auxiliary feature set. Using prediction networks, we can predict the power generation of thermal power plants at multiple time scales.

[0080] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0081] Example 3 A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the multi-timescale power generation prediction method for thermal power plants, including, for example, acquiring operating data of the thermal power plant, performing data missing processing on the operating data, and constructing a masked input matrix. Based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. According to the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained. The final set of effective auxiliary features is obtained by filtering from the K optimal hysteresis auxiliary sequences. According to the final effective auxiliary feature set The system utilizes a predictive network to forecast the multi-timescale power generation of thermal power plants. The memory may include main memory, such as high-speed random access memory (RAM), or non-volatile memory, such as at least one disk storage device. The processor, network interface, and memory are interconnected via an internal bus, which can be an industry-standard architecture bus, a peripheral component interconnection standard bus, or an extended industry-standard architecture bus. The bus can be categorized as an address bus, data bus, and control bus. The memory stores programs; specifically, the programs may include program code, which includes computer operation instructions. The memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0082] Example 4 A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multi-timescale power generation prediction method for thermal power plants, including, for example,: acquiring operating data of the thermal power plant, performing data missing processing on the operating data of the thermal power plant, and constructing a masked input matrix. Based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. According to the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained. The final set of effective auxiliary features is obtained by filtering from the K optimal hysteresis auxiliary sequences. According to the final effective auxiliary feature set The system utilizes a predictive network to forecast the multi-timescale power generation of thermal power plants. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0083] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0085] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0088] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0089] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for predicting power generation at multiple time scales in thermal power plants, characterized in that, include: Acquire operational data from thermal power plants, perform data loss handling on the operational data, and construct a masked input matrix. ; Based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. ; According to the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained. The final set of effective auxiliary features is obtained by filtering from the K optimal hysteresis auxiliary sequences. ; Based on the final effective auxiliary feature set Using prediction networks, we can predict the power generation of thermal power plants at multiple time scales.

2. The method for predicting power generation at multiple time scales in thermal power plants according to claim 1, characterized in that, The masked input matrix for: in, Represents the mask matrix This represents element-wise multiplication. For the original multivariate data matrix, Built based on the operating data of thermal power plants; The mask matrix for: in, Indicates the first The first variable An indicator of whether a time is valid. A value of 1 indicates that the condition is valid. A value of 0 indicates that the value is missing.

3. The method for predicting power generation at multiple time scales in thermal power plants according to claim 1, characterized in that, The local-global consistency loss of the reconstruction error constraint model for: in, Represents Euclidean distance. This is the global trend vector. , This represents the global trend encoding function.

4. The method for predicting power generation at multiple time scales in thermal power plants according to claim 1, characterized in that, The according to the latent variable vector The process of obtaining K optimal lag auxiliary sequences based on time-delay correlation spectrum projection is as follows: Construct a set of time delay values ,in, , Indicates the first Each time delay length; Based on latent variable vectors Candidate variable sequence Using time delay value set Generate several lag sequences ; Calculate the latent variables corresponding to the main predictor variables With each lag sequence The lag distance between them; Select the K lag sequences with the smallest lag distances. As K optimal lag auxiliary sequences.

5. The method for predicting power generation at multiple time scales in thermal power plants according to claim 1, characterized in that, The final effective auxiliary feature set is obtained by filtering from the K optimal lag auxiliary sequences. The process is as follows: The K optimal lag auxiliary sequences are input into the auxiliary sequence encoder to obtain K embedding vectors. ; Based on contrastive loss, from K embedding vectors Several effective embedding vectors are selected from the set, and these are used to construct the final effective auxiliary feature set. .

6. The method for predicting the multi-timescale power generation of thermal power plants according to claim 1, characterized in that, The contrast loss for: in, Represents the embedding vector of the main sequence; "" represents the vector dot product.

7. The method for predicting power generation at multiple time scales in thermal power plants according to claim 1, characterized in that, The prediction network is trained based on multi-head attention layers, bidirectional temporal convolutions, Transformer, GRU, or LSTM.

8. A multi-timescale power generation prediction system for thermal power plants, characterized in that, include: The acquisition module is used to acquire the operating data of thermal power plants, perform data missing handling on the operating data, and construct an input matrix with a mask. ; Reconstruction module, used based on the masked input matrix By reconstructing the error constraint model, the latent variable vector is output. ; Projection module, used to project based on the latent variable vector Based on time-delay correlation spectrum projection, K optimal lag auxiliary sequences are obtained. The filtering module is used to filter the K optimal lag auxiliary sequences to obtain the final effective auxiliary feature set. ; The prediction module is used to predict based on the final effective auxiliary feature set. Using prediction networks, we can predict the power generation of thermal power plants at multiple time scales.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-timescale power generation prediction method for thermal power plants as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-timescale power generation prediction method for thermal power plants as described in any one of claims 1-7.

Images