Combustion furnace control parameter tuning method, device, equipment, storage medium and product

Abstract

The invention discloses a combustion furnace control parameter adjusting and optimizing method, device and equipment, a storage medium and a product, and relates to the technical field of computers.The combustion furnace control parameter adjusting and optimizing method comprises the steps that multi-modal data information is collected; through a preset deep contrast learning model, converting the multi-modal data information to obtain a multi-modal feature vector; and inputting the multi-modal feature vector into a preset conversion encoder to obtain a combustion furnace tuning control parameter corresponding to the multi-modal feature vector. According to the method, the collected multi-modal data information is processed into the multi-modal feature vectors, and the multi-modal feature vectors are matched to the corresponding combustion furnace tuning control parameters through the conversion encoder, so that the combustion efficiency is improved, and energy waste is reduced.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, equipment, storage medium and product for optimizing control parameters of a combustion furnace. Background Art

[0002] With the rapid development of coal-fired furnaces and increasingly stringent air pollutant emission standards, the demand for efficient combustion in furnaces is becoming increasingly urgent.

[0003] Existing combustion furnaces often control the thermal state of the combustion furnace based on the experience of on-site operators in judging the furnace temperature trend, and the control parameters are adjusted manually. However, the thermal state of the combustion furnace is affected by many factors. Manual control alone will lead to low combustion efficiency, thereby causing energy waste.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of this application is to provide a method, device, equipment, storage medium and product for optimizing control parameters of a combustion furnace, aiming to solve the technical problem of energy waste.

[0006] To achieve the above objectives, the present application proposes a method for optimizing combustion furnace control parameters, the method comprising:

[0007] Collect multimodal data information;

[0008] The multimodal data information is transformed through a preset deep contrast learning model to obtain a multimodal feature vector;

[0009] The multimodal feature vector is input into a preset conversion encoder to match and obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

[0010] In one embodiment, the step of converting the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector includes:

[0011] When the multimodal data information is picture data, encoding the picture data by a preset picture encoder to obtain a picture feature vector;

[0012] When the multimodal data information is text data, encoding the text data by a preset text encoder to obtain a text feature vector;

[0013] When the multimodal data information is voice data, the voice data is converted into voice image data through Fourier transform, and the voice image data is encoded through a preset voice encoder to obtain a voice feature vector.

[0014] In one embodiment, the step of converting the multimodal data information by a preset deep contrastive learning model to obtain a multimodal feature vector includes:

[0015] Obtain multimodal data samples and multimodal feature vector labels;

[0016] Inputting the multimodal data samples into a preset model to be trained;

[0017] According to the model to be trained, converting the multimodal data sample into a multimodal training feature vector;

[0018] Analyzing the difference between the multimodal training feature vector and the multimodal feature vector label to obtain a conversion loss value;

[0019] If the conversion loss value does not meet the preset loss standard, return to the step of converting the multimodal data sample into a multimodal training feature vector according to the model to be trained, and stop training until the conversion loss value meets the loss standard to obtain a deep contrast learning model that meets the loss standard.

[0020] In one embodiment, the step of inputting the multimodal feature vector into a preset conversion encoder comprises:

[0021] Fusing the multimodal feature vectors to obtain target fused data;

[0022] Blocking the target fusion data to obtain target blocks;

[0023] Performing a linear transformation on the target block to obtain a target block vector;

[0024] Based on a preset position identifier, adding position information to each block in the target block vector;

[0025] The target block vector with added position information is input to a preset conversion encoder.

[0026] In one embodiment, the step of inputting the multimodal feature vector into a preset conversion encoder to match and obtain the combustion furnace tuning control parameter corresponding to the multimodal feature vector comprises:

[0027] Inputting the multimodal feature vector into the conversion encoder;

[0028] Performing layer normalization on the multimodal feature vector by the conversion encoder to obtain a normalized vector;

[0029] By using the preset multi-head self-attention and cross-attention mechanisms, the normalized vector is matched with the corresponding control scheme in the preset tuning control scheme library to obtain the tuning control scheme;

[0030] Determine whether the tuning control scheme meets the preset tuning conditions. If not, return to the step of performing layer normalization on the multimodal feature vector through the conversion encoder to obtain a normalized vector;

[0031] When the optimization control scheme satisfies the output condition, the control parameters in the optimization control scheme are determined as the combustion furnace optimization control parameters.

[0032] In one embodiment, the step of inputting the multimodal feature vector into a preset conversion encoder and matching to obtain the combustion furnace tuning control parameter corresponding to the multimodal feature vector comprises:

[0033] When the similarity of the multimodal feature vectors collected from different combustion furnaces reaches a preset range, the combustion furnace tuning control parameters are sent to the combustion furnaces whose similarity of the multimodal feature vectors reaches the preset range.

[0034] In addition, to achieve the above-mentioned purpose, the present application also proposes a combustion furnace control parameter tuning device, the combustion furnace control parameter tuning device comprising:

[0035] An acquisition module, used for acquiring multimodal data information;

[0036] A conversion module, used to convert the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector;

[0037] The tuning module is used to input the multimodal feature vector into a preset conversion encoder to match and obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

[0038] In addition, to achieve the above-mentioned purpose, the present application also proposes a combustion furnace control parameter tuning device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the combustion furnace control parameter tuning method as described above.

[0039] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the combustion furnace control parameter tuning method described above are implemented.

[0040] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the combustion furnace control parameter tuning method as described above are implemented.

[0041] One or more technical solutions proposed in this application have at least the following technical effects:

[0042] Compared with the related art, the existing combustion furnaces often control the thermal state of the combustion furnace based on the experience of the on-site operator's judgment of the furnace temperature trend, and the control parameters are manually adjusted. However, the thermal state of the combustion furnace is affected by many factors, and manual control alone will lead to low combustion efficiency, thereby causing energy waste. In comparison, the present application collects multimodal data information; transforms the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector; inputs the multimodal feature vector into a preset conversion encoder, and matches the combustion furnace tuning control parameters corresponding to the multimodal feature vector. It is understandable that the present application uses the method of processing the collected multimodal data information into a multimodal feature vector, matching the multimodal feature vector to the corresponding combustion furnace tuning control parameters through a conversion encoder, thereby improving combustion efficiency and reducing energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0044] 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 or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0045] Figure 1 A schematic diagram of a flow chart provided for the first embodiment of the combustion furnace control parameter tuning method of the present application;

[0046] Figure 2 A schematic diagram of a flow chart provided for the second embodiment of the combustion furnace control parameter tuning method of the present application;

[0047] Figure 3 This is a schematic diagram of the module structure of the combustion furnace control parameter tuning device according to an embodiment of the present application;

[0048] Figure 4 Schematic diagram of the equipment structure of the hardware operating environment involved in the combustion furnace control parameter tuning method in the embodiment of the present application.

[0049] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0050] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0051] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0052] The main solutions of the embodiments of this application are:

[0053] Collect multimodal data information;

[0054] The multimodal data information is transformed through a preset deep contrast learning model to obtain a multimodal feature vector;

[0055] The multimodal feature vector is input into a preset conversion encoder to obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

[0056] In this embodiment, the present application uses the combustion furnace control parameter tuning device as the execution body. For the convenience of description, it is specifically described below as "device".

[0057] Since existing combustion furnaces often control the thermal state of the combustion furnace based on the experience of on-site operators in judging the furnace temperature trend, and the control parameters are adjusted manually, the thermal state of the combustion furnace is affected by many factors. Manual control alone will lead to low combustion efficiency, thereby causing energy waste.

[0058] The present application provides a solution, by collecting multimodal data information; transforming the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector; inputting the multimodal feature vector into a preset conversion encoder to obtain a combustion furnace tuning control parameter corresponding to the multimodal feature vector. It can be understood that the present application adopts the method of processing the collected multimodal data information into a multimodal feature vector, matching the multimodal feature vector to the corresponding combustion furnace tuning control parameter through a conversion encoder, thereby improving combustion efficiency and reducing energy waste.

[0059] Based on this, the embodiment of the present application provides a method for optimizing the control parameters of a combustion furnace, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the combustion furnace control parameter tuning method of the present application.

[0060] In this embodiment, the combustion furnace control parameter tuning method includes steps S10 to S30:

[0061] Step S10, collecting multimodal data information;

[0062] It should be noted that the multimodal data information is an information set containing multiple types of data (such as images, texts, voices, etc.) These different types of data together describe or record multiple aspects of an object or event.

[0063] Step S20, transforming the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector;

[0064] It should be noted that the deep contrastive learning model is a deep learning model used to learn the similarities and differences between different data samples. Through contrastive learning, the model can capture the intrinsic characteristics and relationships of the data, thereby improving the ability to understand and represent the data.

[0065] In a feasible implementation, step S20 may include:

[0066] When the multimodal data information is picture data, encoding the picture data by a preset picture encoder to obtain a picture feature vector;

[0067] When the multimodal data information is text data, encoding the text data by a preset text encoder to obtain a text feature vector;

[0068] When the multimodal data information is voice data, the voice data is converted into voice image data through Fourier transform, and the voice image data is encoded through a preset voice encoder to obtain a voice feature vector.

[0069] It should be noted that the multimodal data information is a collection of information containing multiple types of data (such as pictures, text, voice, etc.). These different types of data together describe or record multiple aspects of an object or event. The picture encoder is an encoder specially designed to process picture data. It converts picture data into feature vectors through a specific algorithm, and these vectors can capture the key visual information of the picture. The picture feature vector is a set of numerical values ​​obtained after the picture encoder encodes the picture data. These numerical vectors can represent the key features of the original picture and can be used for further analysis and processing. The text encoder is an encoder specially designed to process text data. It converts text data into feature vectors through a specific algorithm, and these vectors can capture the semantic and structural information of the text. The text feature vector is a set of numerical values ​​obtained after the text encoder encodes the text data. These numerical vectors can represent the key features of the original text and can be used for further analysis and processing. The Fourier transform is a mathematical transform used to convert a signal from the time domain to the frequency domain. In speech processing, the Fourier transform can convert a speech signal into a spectrum image, thereby converting the speech data into a form that can be processed by an image processing algorithm. The speech image data is a data form obtained by converting speech data through Fourier transform. This data form usually displays the spectral characteristics of speech in the form of an image. The speech encoder is an encoder specially designed to process speech data. In this context, it processes the speech image data obtained through Fourier transform and converts it into a feature vector. The speech feature vector is a set of numerical values ​​obtained after the speech encoder encodes the speech image data. These numerical vectors can represent the key features of the original speech and can be used for further analysis and processing.

[0070] In another feasible implementation manner, before step S20, the following may be included:

[0071] Obtain multimodal data samples and multimodal feature vector labels;

[0072] Inputting the multimodal data samples into a preset model to be trained;

[0073] According to the model to be trained, converting the multimodal data sample into a multimodal training feature vector;

[0074] Analyzing the difference between the multimodal training feature vector and the multimodal feature vector label to obtain a conversion loss value;

[0075] If the conversion loss value does not meet the preset loss standard, return to the step of converting the multimodal data sample into a multimodal training feature vector according to the model to be trained, and stop training until the conversion loss value meets the loss standard to obtain a deep contrast learning model that meets the loss standard.

[0076] For example, refer to Figure 2 After data cleaning and other preprocessing, the collected text, pictures, and sounds are used to build the training set, validation set, and test set of the multimodal model and input them into the combustion furnace control parameter tuning model. The Image Encoder, Text Encoder, and Voice Encoder in the deep contrast learning module convert the input data into corresponding feature vector Embeddings. The Image Encoder uses the CNN-based ResNet model, the Text Encoder uses the NLP text transformer model, and the Voice Encoder first converts the sound signal into an image signal through Fourier transform, and then uses the vision transformer model.

[0077] When training the deep contrastive learning module, images, text and voice data are randomly taken from the training set. The main task of this module is to use cosine similarity to compare the similarity of the three embedding vectors to determine whether the randomly extracted images, text and voice data match, and perform gradient back propagation to continuously optimize the training.

[0078] Step S30: input the multimodal feature vector into a preset conversion encoder to obtain a combustion furnace tuning control parameter corresponding to the multimodal feature vector.

[0079] It should be noted that the multimodal feature vector is a set of numerical values ​​obtained after the deep contrast learning model transforms the multimodal data information. These numerical vectors can represent the key features of the original multimodal data and can be used for further analysis and processing. The conversion encoder is an encoder for converting a multimodal feature vector into another form of data representation. In this context, the purpose of the conversion encoder is to map the feature vector to a parameter space related to the combustion furnace tuning control. The combustion furnace tuning control parameters are a series of parameters used to adjust and optimize the performance of the combustion furnace. These parameters may include temperature, pressure, fuel flow, etc. By precisely controlling these parameters, the combustion efficiency can be improved, emissions can be reduced, and the stability and safety of the combustion process can be ensured.

[0080] In a feasible implementation manner, the step of inputting the multimodal feature vector into a preset conversion encoder includes:

[0081] Fusing the multimodal feature vectors to obtain target fused data;

[0082] Blocking the target fusion data to obtain target blocks;

[0083] Performing a linear transformation on the target block to obtain a target block vector;

[0084] Based on a preset position identifier, adding position information to each block in the target block vector;

[0085] The target block vector with added position information is input to a preset conversion encoder.

[0086] It is understood that the target fusion data refers to the data obtained by merging the multimodal feature vectors through a fusion method (such as weighted averaging, splicing, etc.). This fusion is intended to integrate information from different modalities for subsequent processing and analysis. Target chunking refers to the process of dividing the target fusion data into multiple small blocks or fragments. Chunking can be to meet specific processing requirements, such as parallel processing, local feature extraction, etc. The target chunk vector refers to the vector representation obtained after linear transformation of the target chunk. The linear transformation may include operations such as matrix multiplication to convert the chunk data into a format suitable for neural network processing. Position information In scenarios where sequence data is processed or the spatial structure of the data needs to be considered, position information refers to the position identifier of each chunk in the original data. Adding position information can help the model understand the structure and order relationship of the data. A transformation encoder is an encoder that converts a target chunk vector into another form of data representation. The transformation encoder may include a deep learning model, such as Transformer, for further feature extraction and task-related processing.

[0087] For example, the network input is 1024*1024 resolution, 3-channel furnace status related pictures plus text and voice data, aligned and fused into four-channel data, and the input data is evenly divided into blocks from top to bottom and from left to right, each block size is P*P*3, and the insufficient blocks are filled with (255, 255, 255) white. The total number of blocks is B = 1024*1024 / (P*P), and the input after block into patches becomes I = B*(P*P*3), so that each patch passes through a fully connected layer and is linearly transformed into a D-dimensional vector. B D-dimensional vectors are connected end to end, which is a B*D-dimensional Flattened Patch Embedding.

[0088] In order to identify the location of each block in Patch Embedding, a trainable variable Position embedding is introduced, which adds position information to the input sequence of the network.

[0089] In another feasible implementation, step S30 may include:

[0090] Inputting the multimodal feature vector into the conversion encoder;

[0091] Performing layer normalization on the multimodal feature vector by the conversion encoder to obtain a normalized vector;

[0092] By using the preset multi-head self-attention and cross-attention mechanisms, the normalized vector is matched with the corresponding control scheme in the preset tuning control scheme library to obtain the tuning control scheme;

[0093] Determine whether the tuning control scheme meets the preset tuning conditions. If not, return to the step of performing layer normalization on the multimodal feature vector through the conversion encoder to obtain a normalized vector;

[0094] When the optimization control scheme satisfies the output condition, the control parameters in the optimization control scheme are determined as the combustion furnace optimization control parameters.

[0095] For example, the role of layer normalization (Norm) is to normalize the hidden layer in the neural network to a standard normal distribution, that is, independent and identically distributed, so as to speed up the training speed and accelerate the convergence:

[0096]

[0097] In the above formula, x is the element of the input matrix, and the mean is calculated in units of the rows of the matrix;

[0098]

[0099] In the above formula, the variance is calculated in units of the rows of the matrix;

[0100]

[0101] Then, we subtract the mean of each row from the element of each row and divide it by the standard deviation of the row to get the normalized value, ∈ is equal to 10e-8, α and β are two trainable parameters introduced to make up for the information lost in the normalization process, ⊙ represents element multiplication, we generally initialize α to all 1 and β to all 0.

[0102] After the Embedding calculated by the previous module is concatenated, it enters the Transformer encoder that repeats N times. At the beginning, the encoder performs layer normalization on the input, and then uses the multi-head self-attention and cross-attention mechanisms. On the keys and values, the Feature Attention pattern defined by us is used for the attention. The attention pattern is not restricted to a local window, which reduces the computational amount while retaining more global information, as shown in the following formula:

[0103]

[0104] Similar to the Attention of the Transformer, A in Feature Attention mqk is not obtained by taking the inner product of the Query and Key matrices, but is obtained from the input feature z q through a fully connected layer. At the same time, there are p q and Δp mqk . Among them, p q ∈R 2 represents the coordinates of the reference point, which is a 2D vector, while Δp mqk ∈R 2 represents the position offset (offsets) of the k-th sampling point in the m-th attention head relative to the reference point. K is the total number of sampled keys (K << HM). It can be seen that each query samples K positions in each head and only needs to interact with the features at these positions. x(p q +Δp mqk ) represents the value interpolated based on the sampling point position, and there is no need to start learning from the global position like Transformer Attention at the beginning and then gradually transition to focusing on local, sparse, and truly meaningful positions.

[0105] The position offset Δp mqk is learnable and is obtained from the query feature z q ∈R C through a fully connected layer. And, the scalar attention weight A mqk is the same and is directly obtained from the query feature z q ∈R C through a fully connected layer. Therefore, under the deformable attention mechanism, there is actually no real key to interact and calculate with the query. At the same time, it is normalized among the K sampling points, rather than being obtained by the interaction calculation between the query and the key like Transformer Attention. Therefore, A mqk is in the range of [0, 1], and

[0106] In another feasible implementation manner, after the step of inputting the multimodal feature vector into a preset conversion encoder to obtain the combustion furnace tuning control parameter corresponding to the multimodal feature vector, the step includes:

[0107] When the similarity of the multimodal feature vectors collected from different combustion furnaces reaches a preset range, the combustion furnace tuning control parameters are sent to the combustion furnaces whose similarity of the multimodal feature vectors reaches the preset range.

[0108] It is understandable that different heating process parameter controls are performed for steel billets with different physical and chemical indicators and sizes, and the combustion furnace is optimized from single-machine control to cluster control.

[0109] The present embodiment provides a method for tuning the control parameters of a combustion furnace. The present application processes the collected multimodal data information into a multimodal feature vector, and matches the multimodal feature vector to the corresponding combustion furnace tuning control parameter through a conversion encoder, thereby improving combustion efficiency and reducing energy waste.

[0110] For example, in order to help understand the implementation process of the combustion furnace control parameter tuning method obtained by combining this embodiment with the above-mentioned embodiment 1, please refer to Figure 2 , Figure 2 A brief flow chart of a combustion furnace control parameter tuning method is provided, specifically:

[0111] The flowchart describes a multimodal data processing method, which involves the embedding representation and fusion processing of three data types: image, text, and speech. The following is a detailed description of the process:

[0112] Data Input: The system receives three types of data input: image, text, and voice.

[0113] Embedding representation: For each type of data, the system first performs embedding processing to convert the original data into a vector representation. Image data is converted through the Image Embedding module. Text data is converted through the Text Embedding module. Voice data is converted through the Voice Embedding module.

[0114] Fusion processing: The embedded image, text and voice data are sent to the fusion module for preliminary integration processing.

[0115] Position Embedding: In order to preserve the spatial or sequence information of the data, the system adds position embedding to the fused data.

[0116] Block processing: The data that is fused and embedded with position is processed in blocks (Patch) to facilitate subsequent parallel processing.

[0117] Normalization: The divided data is processed by normalization to adjust the distribution of the data to make it more suitable for subsequent neural network processing.

[0118] Multi-Head Self-Attention Mechanism: The normalized data is input into the Multi-Head Self-Attention module, which can capture the long-distance dependencies within the data.

[0119] CFFN module: The output of multi-head self-attention passes through the CFFN (Convolutional Feed-Forward Network) module for further feature extraction and nonlinear transformation.

[0120] Cross-Attention Mechanism: The output of CFFN passes through a Cross-Attention module, which allows information interaction between different modalities and enhances the expressiveness of features.

[0121] Query Selection: The output of the cross-attention passes through the query selection module, which may be used to select or generate a specific query vector to guide the subsequent information retrieval or decision-making process.

[0122] Multilayer Perceptron Head: Finally, the processed data is passed through the Multilayer Perceptron Head (MLP Head) for classification or prediction of other tasks.

[0123] Similarity judgment: The system also includes a similarity judgment module, which is used to compare the similarities between different data and output similar (Similar) or dissimilar (Not Similar) results.

[0124] This application realizes the comprehensive analysis and processing of image, text and voice data through the embedded representation, fusion processing and attention mechanism of multimodal data, which is applicable to a variety of machine learning and artificial intelligence application scenarios. When writing patent documents, the functions and operation details of each module can be further described in detail according to the specific technical implementation and application scenarios.

[0125] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the combustion furnace control parameter tuning method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0126] This application also provides a combustion furnace control parameter tuning device, please refer to Figure 3, the combustion furnace control parameter tuning device comprises:

[0127] The acquisition module 10 is used to acquire multimodal data information;

[0128] A conversion module 20, configured to convert the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector;

[0129] The tuning module 30 is used to input the multimodal feature vector into a preset conversion encoder to obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

[0130] And / or, the conversion module 20 includes:

[0131] A first encoding module, configured to, when the multimodal data information is picture data, encode the picture data by using a preset picture encoder to obtain a picture feature vector;

[0132] A second encoding module, for encoding the text data by a preset text encoder to obtain a text feature vector when the multimodal data information is text data;

[0133] The third encoding module is used to convert the voice data into voice image data through Fourier transform when the multimodal data information is voice data, and encode the voice image data through a preset voice encoder to obtain a voice feature vector.

[0134] And / or, the combustion furnace control parameter tuning device comprises:

[0135] A first acquisition module is used to acquire multimodal data samples and multimodal feature vector labels;

[0136] A first input module, used for inputting the multimodal data sample into a preset model to be trained;

[0137] A first prediction module, configured to convert the multimodal data sample into a multimodal training feature vector according to the model to be trained;

[0138] A first judgment module is used to analyze the difference between the multimodal training feature vector and the multimodal feature vector label to obtain a conversion loss value;

[0139] The second prediction module is used to return to the step of converting the multimodal data sample into a multimodal training feature vector according to the model to be trained if the conversion loss value does not meet the preset loss standard, and stop training until the conversion loss value meets the loss standard to obtain a deep contrast learning model that meets the loss standard.

[0140] And / or, the tuning module 30 includes:

[0141] A first fusion module, used for fusing the multimodal feature vectors to obtain target fusion data;

[0142] A first block division module is used to divide the target fusion data into blocks to obtain target blocks;

[0143] A first transformation module, used for performing a linear transformation on the target block to obtain a target block vector;

[0144] A first adding module, configured to add position information to each block in the target block vector based on a preset position identifier;

[0145] The second input module is used to input the target block vector with added position information to a preset conversion encoder.

[0146] And / or, the tuning module 30 includes:

[0147] A third input module, configured to input the multimodal feature vector into the conversion encoder;

[0148] A first normalization module, configured to perform layer normalization on the multimodal feature vector through the conversion encoder to obtain a normalized vector;

[0149] A first matching module is used to match the corresponding control scheme in a preset tuning control scheme library according to the normalized vector through a preset multi-head self-attention and cross-attention mechanism to obtain a tuning control scheme;

[0150] A second judgment module is used to judge whether the tuning control scheme meets the preset tuning conditions. If not, it returns to the step of performing layer normalization on the multimodal feature vector through the conversion encoder to obtain a normalized vector;

[0151] The first determination module is used to determine the control parameters in the optimization control scheme as the combustion furnace optimization control parameters when the optimization control scheme meets the output condition.

[0152] And / or, the combustion furnace control parameter tuning device comprises:

[0153] The first sending module is used to send the combustion furnace tuning control parameters to the combustion furnace whose similarity of the multimodal feature vector reaches the preset range when the similarity of the multimodal feature vector collected by different combustion furnaces reaches the preset range.

[0154] The combustion furnace control parameter tuning device provided by the present application adopts the combustion furnace control parameter tuning method in the above embodiment, which can solve the technical problem of energy waste. Compared with the prior art, the beneficial effects of the combustion furnace control parameter tuning device provided by the present application are the same as the beneficial effects of the combustion furnace control parameter tuning method provided by the above embodiment, and other technical features in the combustion furnace control parameter tuning device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0155] The present application provides a combustion furnace control parameter tuning device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the combustion furnace control parameter tuning method in the above-mentioned embodiment one.

[0156] Reference below Figure 4 , which shows a schematic diagram of the structure of a combustion furnace control parameter tuning device suitable for implementing the embodiment of the present application. The combustion furnace control parameter tuning device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, tablet computers, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The combustion furnace control parameter tuning device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0157] like Figure 4As shown, the combustion furnace control parameter tuning device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the combustion furnace control parameter tuning device are also stored in the RAM 1004. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 may allow the combustion furnace control parameter tuning device to communicate wirelessly or by wire with other devices to exchange data. Although the combustion furnace control parameter tuning device having various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0158] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0159] The combustion furnace control parameter tuning device provided by the present application adopts the combustion furnace control parameter tuning method in the above embodiment, which can solve the technical problem of energy waste. Compared with the prior art, the beneficial effects of the combustion furnace control parameter tuning device provided by the present application are the same as the beneficial effects of the combustion furnace control parameter tuning method provided by the above embodiment, and other technical features in the combustion furnace control parameter tuning device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0160] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0161] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0162] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the combustion furnace control parameter tuning method in the above-mentioned embodiment.

[0163] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0164] The computer-readable storage medium may be included in the combustion furnace control parameter tuning device; or may exist independently without being assembled into the combustion furnace control parameter tuning device.

[0165] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the combustion furnace control parameter tuning device, the combustion furnace control parameter tuning device: collects multi-modal data information;

[0166] The multimodal data information is transformed through a preset deep contrast learning model to obtain a multimodal feature vector;

[0167] The multimodal feature vector is input into a preset conversion encoder to obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

[0168] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0169] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0170] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0171] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned combustion furnace control parameter tuning method, and can solve the technical problem of energy waste. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the combustion furnace control parameter tuning method provided in the above-mentioned embodiment, and will not be described in detail here.

[0172] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned combustion furnace control parameter tuning method when executed by a processor.

[0173] The computer program product provided in this application can solve the technical problem of energy waste. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the combustion furnace control parameter tuning method provided in the above embodiment, which will not be repeated here.

[0174] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for optimizing combustion furnace control parameters, characterized in that: The method includes: Collect multimodal data information; The multimodal data information is transformed through a preset deep contrast learning model to obtain a multimodal feature vector; The multimodal feature vector is input into a preset conversion encoder to match and obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

2. The method according to claim 1, characterized in that The step of converting the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector includes: When the multimodal data information is picture data, encoding the picture data by a preset picture encoder to obtain a picture feature vector; When the multimodal data information is text data, encoding the text data by a preset text encoder to obtain a text feature vector; When the multimodal data information is voice data, the voice data is converted into voice image data through Fourier transform, and the voice image data is encoded through a preset voice encoder to obtain a voice feature vector.

3. The method according to claim 1, characterized in that The step of converting the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector includes: Obtain multimodal data samples and multimodal feature vector labels; Inputting the multimodal data samples into a preset model to be trained; According to the model to be trained, converting the multimodal data sample into a multimodal training feature vector; Analyzing the difference between the multimodal training feature vector and the multimodal feature vector label to obtain a conversion loss value; If the conversion loss value does not meet the preset loss standard, return to the step of converting the multimodal data sample into a multimodal training feature vector according to the model to be trained, and stop training until the conversion loss value meets the loss standard to obtain a deep contrast learning model that meets the loss standard.

4. The method according to claim 1, characterized in that The step of inputting the multimodal feature vector into a preset conversion encoder comprises: Fusing the multimodal feature vectors to obtain target fused data; Blocking the target fusion data to obtain target blocks; Performing a linear transformation on the target block to obtain a target block vector; Based on a preset position identifier, adding position information to each block in the target block vector; The target block vector with added position information is input to a preset conversion encoder.

5. The method according to claim 1, characterized in that The step of inputting the multimodal feature vector into a preset conversion encoder to match and obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector comprises: Inputting the multimodal feature vector into the conversion encoder; Performing layer normalization on the multimodal feature vector by the conversion encoder to obtain a normalized vector; By using the preset multi-head self-attention and cross-attention mechanisms, the normalized vector is matched with the corresponding control scheme in the preset tuning control scheme library to obtain the tuning control scheme; Determine whether the tuning control scheme meets the preset tuning conditions. If not, return to the step of performing layer normalization on the multimodal feature vector through the conversion encoder to obtain a normalized vector; When the optimization control scheme satisfies the output condition, the control parameters in the optimization control scheme are determined as the combustion furnace optimization control parameters.

6. The method according to claim 1, characterized in that The step of inputting the multimodal feature vector into a preset conversion encoder to match and obtain the combustion furnace tuning control parameter corresponding to the multimodal feature vector comprises: When the similarity of the multimodal feature vectors collected from different combustion furnaces reaches a preset range, the combustion furnace tuning control parameters are sent to the combustion furnaces whose similarity of the multimodal feature vectors reaches the preset range.

7. A combustion furnace control parameter tuning device, characterized in that: The device comprises: An acquisition module, used for acquiring multimodal data information; A conversion module, used to convert the multimodal data information through a preset deep contrast learning model to obtain a multimodal feature vector; The tuning module is used to input the multimodal feature vector into a preset conversion encoder to match and obtain the combustion furnace tuning control parameters corresponding to the multimodal feature vector.

8. A combustion furnace control parameter tuning device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the combustion furnace control parameter tuning method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the combustion furnace control parameter tuning method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the combustion furnace control parameter tuning method according to any one of claims 1 to 6 are implemented.

Images