Process for producing synthesis gas through clean conversion of raw gas

Abstract

The invention relates to the field of intelligent production, and provides a process for producing synthesis gas by clean conversion of raw gas, which comprises the following steps of: performing multi-scale feature extraction on a catalyst real-time state image acquired by an industrial camera by adopting an image recognition and analysis technology based on machine vision; and a reaction temperature value and a reaction pressure value are automatically recommended based on a remarkable interactive fusion feature between the catalyst state shallow feature and the catalyst state semantic feature. Therefore, the working state of the catalyst can be monitored in real time, and the change of the activity of the catalyst can be found in time, so that the reaction condition is quickly adjusted, and the problem caused by the change of the reaction condition is avoided. Moreover, the method can realize intelligent management of the reaction process, so that the self-adaptive reaction temperature and pressure value can be recommended more accurately, and the stable proceeding of the reaction process is ensured.

Description

Technical Field

[0001] The present application relates to the field of intelligent production, and more specifically, to a process for producing synthesis gas by clean conversion of waste gas. Background Art

[0002] Raw coal gas usually refers to the gas produced during the coal mining process, which contains a large amount of combustible components such as carbon monoxide and hydrogen. With the adjustment of energy structure and the improvement of environmental protection requirements, how to effectively utilize these by-products has become an important research topic. Clean conversion technology can reduce environmental pollution and improve resource utilization. By converting raw coal gas into synthesis gas (mainly hydrogen and carbon monoxide), it can be used as a chemical raw material to produce various chemicals, such as dimethyl oxalate (DMO) and ethylene glycol.

[0003] In the process of clean conversion of raw coal gas to form ethylene glycol, precise control of reaction conditions is crucial. Reasonable reaction condition control combined with efficient catalyst use is the key factor to achieve efficient and low-cost production of ethylene glycol. However, traditional reaction process control usually relies on manual adjustment based on predetermined procedures or operator experience. In this mode, when reaction conditions change, the reactor control system may not respond quickly, resulting in decreased production efficiency or unstable product quality. In addition, manual control methods are difficult to achieve extremely high accuracy, especially when reaction conditions need to be adjusted quickly, and manual intervention often cannot adjust them to the optimal working point in a timely and accurate manner. This control method not only increases the complexity and uncertainty of the operation, but may also cause errors due to human factors, affecting the quality and consistency of the final product.

[0004] Therefore, an optimized process for clean conversion of raw coal gas to produce synthesis gas is desired. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present application provides a process for producing synthesis gas by clean conversion of raw coal gas.

[0006] According to one aspect of the present application, a process for producing synthesis gas by clean conversion of raw coal gas is provided, which comprises: S1: compressing the raw raw gas to obtain compressed raw gas; S2: extracting hydrogen and carbon monoxide from the compressed raw coal gas; S3: After pressurizing the carbon monoxide, the pressurized carbon monoxide is introduced into a dimethyl oxalate synthesis device to obtain dimethyl oxalate; S4: inputting the dimethyl oxalate and the hydrogen into an ethylene glycol synthesis unit to obtain ethylene glycol.

[0007] This application has significant technical effects due to the adoption of the above technical solutions: The process for producing synthesis gas by clean conversion of waste gas provided in the present application adopts image recognition and analysis technology based on machine vision to perform multi-scale feature extraction on the real-time state image of the catalyst collected by the industrial camera, so as to automatically recommend the reaction temperature value and the reaction pressure value based on the significant interactive fusion features between the shallow features of the catalyst state and the semantic features of the catalyst state. In this way, the working state of the catalyst can be monitored in real time, and the changes in the activity of the catalyst can be discovered in time, so as to quickly adjust the reaction conditions and avoid problems caused by changes in the reaction conditions. In addition, this method can realize the intelligent management of the reaction process, which is conducive to more accurately recommending adaptive reaction temperature and pressure values ​​to ensure the stable progress of the reaction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 The present invention is a flow chart of a process for producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application.

[0010] Figure 2 In the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application, the carbon monoxide is pressurized and then introduced into a dimethyl oxalate synthesis unit to obtain dimethyl oxalate.

[0011] Figure 3 The present invention is a flow chart of inputting the dimethyl oxalate and the hydrogen into an ethylene glycol synthesis unit to obtain ethylene glycol in the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application.

[0012] Figure 4 The present invention is a data flow diagram of inputting the dimethyl oxalate and the hydrogen into an ethylene glycol synthesis unit to obtain ethylene glycol in the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application.

[0013] Figure 5 In the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly interactively encoded with catalyst state significant features to obtain a flow chart of catalyst state shallow-semantically significant sparse interactive fusion features.

[0014] Figure 6In the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application, kernel sparsification and global association are performed on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain a flow chart of the global coding matrix between the catalyst state kernel features.

[0015] Figure 7 In the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application, based on the global coding matrix between the catalyst state kernel features, the catalyst state shallow feature map and the catalyst state semantic feature map are subjected to masked significance joint interaction to obtain a flowchart of the catalyst state shallow-semantic significant sparse interactive fusion features. DETAILED DESCRIPTION

[0016] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0017] It should be understood that the various steps described in the method implementation of the present application can be performed in different orders and / or in parallel. In addition, the method implementation may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0018] In the description of the embodiments of the present application, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0019] It should be noted that the modifications of "one" and "plurality" mentioned in this application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0020] It should also be noted that in this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0021] Raw coal gas is generally produced during coal mining, and it contains many combustible components, such as carbon monoxide and hydrogen. In view of the adjustment of energy structure and the improvement of environmental protection requirements, the effective use of these by-products has become an important research topic. Clean conversion technology can reduce environmental pollution and improve resource utilization. By converting raw coal gas into synthesis gas (mainly hydrogen and carbon monoxide), it can be used as a chemical raw material to produce chemicals such as dimethyl oxalate (DMO) and ethylene glycol.

[0022] In the process of converting raw coal gas into ethylene glycol, precise control of reaction conditions is extremely critical. Reasonable control of reaction conditions and the use of efficient catalysts are important factors in achieving efficient and low-cost production of ethylene glycol. However, traditional reaction process control often relies on manual adjustments based on predetermined procedures or operator experience. In this mode, once the reaction conditions change, the reactor control system may not be able to respond in time, resulting in reduced production efficiency or unstable product quality. Moreover, manual control is difficult to achieve extremely high precision, especially when reaction conditions need to be adjusted quickly, and manual intervention usually cannot adjust them to the optimal working point in a timely and accurate manner. This control method not only increases the complexity and uncertainty of the operation, but may also cause errors due to human factors, affecting the quality and consistency of the final product.

[0023] Based on this, the present application proposes a process for producing synthesis gas by clean conversion of raw coal gas. Figure 1 The flowchart of the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application is as follows. Figure 1 As shown, the process for producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application includes: S1, compressing the raw raw coal gas to obtain compressed raw coal gas; S2, extracting hydrogen and carbon monoxide from the compressed raw coal gas; S3, pressurizing the carbon monoxide, and passing the pressurized carbon monoxide into a dimethyl oxalate synthesis unit to obtain dimethyl oxalate; S4, inputting the dimethyl oxalate and the hydrogen into an ethylene glycol synthesis unit to obtain ethylene glycol.

[0024] In step S1, the raw raw coal gas introduced is compressed to obtain compressed raw coal gas. It should be understood that in order to increase the pressure of the raw raw coal gas and provide suitable operating conditions for the subsequent gas separation and synthesis reactions, the raw raw coal gas introduced needs to be compressed first. In an embodiment of the present application, a specific implementation method for compressing the raw raw coal gas introduced to obtain compressed raw coal gas is: the raw raw coal gas (0.04MPaG) introduced is removed from the original gas through the inlet separator to remove saturated water and some oil dust and other impurities, and then input into the inlet of the raw coal gas compressor, compressed to 3.0MPaG in five stages and sent to the next process.

[0025] In step S2, hydrogen and carbon monoxide are extracted from the compressed raw coal gas. It should be understood that the raw coal gas contains CH4, which can be converted into effective synthesis gas (mainly hydrogen and carbon monoxide) by non-catalytic oxidation. At the same time, some impurities in the raw coal gas can be decomposed into non-toxic or easy-to-handle components by high-temperature conversion reaction, which will not affect the subsequent process and product quality, and will also reduce the harm to the environment in the clean conversion process. Considering that the actual process requirements for extracting hydrogen and carbon monoxide are different, it is necessary to divert the gas obtained by the conversion process, one part is used to extract hydrogen and the other part is used to extract carbon monoxide. In the process flow of extracting hydrogen, although H2 and CO are obtained through the conversion process, the content of these two gases cannot meet the demand for them in the subsequent ethylene glycol production, so it is necessary to perform a conversion operation on the converted gas. Conversion refers to the conversion reaction of CO and H2O (g) in the process gas to generate H2 and CO under certain pressure and temperature conditions under the action of a catalyst. Then, the converted gas is subjected to low-temperature methanol washing to remove acidic impurities in the gas. Finally, pressure swing adsorption is used to purify H2. In the process of extracting carbon monoxide, the converted gas needs to be heat-recovered first. The heat recovery operation can make full use of the thermal energy in the high-temperature gas after conversion, improve the energy efficiency of the entire clean conversion production process, and make the gas temperature meet the requirements of the subsequent acid gas removal process. Similarly, in order to remove acidic impurities in the converted gas, it is washed with low-temperature methanol to remove acidic impurities such as H2S and CO2, thereby achieving the purpose of purifying the raw gas. Finally, the gas after acid gas removal is subjected to pressure swing adsorption to purify CO.

[0026] In step S3, the carbon monoxide is pressurized and then introduced into a dimethyl oxalate synthesis device to obtain dimethyl oxalate. Specifically, Figure 2 In the process of producing synthesis gas by clean conversion of raw coal gas according to the embodiment of the present application, the carbon monoxide is pressurized and then introduced into a dimethyl oxalate synthesis device to obtain dimethyl oxalate. Figure 2 As shown, after the carbon monoxide is pressurized, the pressurized carbon monoxide is introduced into a dimethyl oxalate synthesis device to obtain dimethyl oxalate, including: S31, performing a synthesis treatment on the pressurized carbon monoxide to obtain synthetic dimethyl oxalate; S32, performing a distillation, a decompression and an impurity removal treatment on the synthetic dimethyl oxalate to obtain the dimethyl oxalate.

[0027] In step S31, the pressurized carbon monoxide is subjected to a synthesis treatment to obtain synthetic dimethyl oxalate. It should be understood that pressurizing carbon monoxide can make it more active in the synthesis reaction, which helps to speed up the reaction rate, and pressurization increases the concentration of carbon monoxide, which further increases the product generation rate. In the present application, the pressurized carbon monoxide is subjected to a synthesis treatment specifically referring to the reaction of carbon monoxide with methyl nitrite in a synthesis reactor, and dimethyl oxalate (DMO) is synthesized under the action of a synthesis catalyst with Pd / Al2O3 as the main component, while NO is generated, and dimethyl carbonate is also generated as a by-product. The main reaction chemical equations of the synthesis treatment are as follows: 2CO+2CH3ONO→(CH3COO)2+2NO In step S32, the synthetic dimethyl oxalate is subjected to rectification, decompression and impurity removal to obtain the synthetic dimethyl oxalate. It should be understood that in order to remove impurities in the synthetic dimethyl oxalate, such as unreacted carbon monoxide, methyl nitrite and other byproducts, it is necessary to carry out corresponding treatment on the synthetic dimethyl oxalate and then purify the product. Specifically, the boiling point difference of different components is first utilized to separate the synthetic dimethyl oxalate from other components, and then decompression operation is carried out to increase the relative volatility between the components, and the impurities of the synthetic dimethyl oxalate are further separated, and finally, adsorption method, extraction method, etc. can be used to remove the residual impurities. In the present application embodiment, the synthetic dimethyl oxalate is subjected to rectification, decompression and impurity removal to obtain a method for implementing the synthetic dimethyl oxalate: the synthetic dimethyl oxalate solution is stripped by the dimethyl oxalate stripping tower, the dimethyl oxalate pre-light removal tower is rectified, and the dimethyl oxalate decompression light removal tower is flashed to remove impurities, and finally dimethyl oxalate is obtained.

[0028] In step S4, the dimethyl oxalate and the hydrogen are input into an ethylene glycol synthesis unit to obtain ethylene glycol. It should be understood that after the dimethyl oxalate raw material from the dimethyl oxalate synthesis unit and the hydrogen from the pressure swing adsorption hydrogen extraction process are mixed, a catalytic hydrogenation reaction occurs under a certain temperature, pressure and catalyst to generate hydrogenated crude ethylene glycol, and the reaction equation is as follows: (COOCH3)2+4H2 catalysis (CH2OH)2+2CH3OH It is worth mentioning that in this process, the hydrogenation cycle gas separated from the ethylene glycol synthesis process can be compressed and returned to the ethylene glycol synthesis process for recycling, and the purge gas can be sent to the hydrogen recovery process. In the hydrogen recovery process, pressure swing adsorption technology can be used to recover and purify hydrogen from the ethylene glycol synthesis purge gas. Considering that the hydrogenation reaction product contains a variety of by-products such as ethanol and butanediol, it is necessary to set up an ethylene glycol separation and refining process to separate and obtain ethylene glycol products that meet the standards. In this application, ethylene glycol is cooled to 95°C through a liquid phase hydrogenation feed cooler and then mixed with hydrogen and enters a liquid phase hydrogenation reactor. The mixture of ethylene glycol and hydrogen enters from the top of the reactor, passes through the catalyst bed, and completes gas-liquid separation at the bottom of the reactor. The impurity adsorption, hydrogenation reaction and other processes are completed in the catalyst bed. The liquid phase returns to the light removal tower for secondary distillation to obtain a refined ethylene glycol product.

[0029] Accordingly, in the case where the dimethyl oxalate and the hydrogen are input into an ethylene glycol synthesis unit to obtain ethylene glycol, the technical concept of the present application is to collect a real-time state image of the catalyst by an industrial camera, and to use image recognition and analysis technology based on machine vision to extract multi-scale features of the real-time state image of the catalyst, so as to automatically recommend reaction temperature values ​​and reaction pressure values ​​based on the significant interactive fusion features between the shallow features of the catalyst state and the semantic features of the catalyst state. In this way, the working state of the catalyst can be monitored in real time, and changes in catalyst activity can be discovered in time, so as to quickly adjust the reaction conditions and avoid problems caused by changes in reaction conditions. Moreover, this method can realize intelligent management of the reaction process, and thus is conducive to more accurately recommending adaptive reaction temperature and pressure values, ensuring the stable progress of the reaction process.

[0030] Specifically, Figure 3 The present invention is a flow chart of inputting the dimethyl oxalate and the hydrogen into an ethylene glycol synthesis unit to obtain ethylene glycol in the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application. Figure 4 This is a data flow diagram of the process of producing synthesis gas by clean conversion of raw coal gas according to an embodiment of the present application, in which the dimethyl oxalate and the hydrogen are input into an ethylene glycol synthesis unit to obtain ethylene glycol. Figure 3 and Figure 4As shown, the dimethyl oxalate and the hydrogen are input into an ethylene glycol synthesis device to obtain ethylene glycol, including: S41, obtaining a real-time state image of the catalyst collected by an industrial camera; S42, performing multi-scale feature extraction of the catalyst state on the real-time state image of the catalyst to obtain a shallow feature map of the catalyst state and a semantic feature map of the catalyst state; S43, performing joint interactive encoding of catalyst state significant features on the shallow feature map of the catalyst state and the semantic feature map of the catalyst state to obtain shallow-semantically significant sparse interactive fusion features of the catalyst state; S44, obtaining an optimization result based on the shallow-semantically significant sparse interactive fusion features of the catalyst state, and the optimization result includes a decoding value of a recommended reaction temperature and a decoding value of a recommended reaction pressure.

[0031] In step S41, a real-time state image of the catalyst collected by the industrial camera is obtained. It should be understood that the real-time state image of the catalyst contains color information, morphology information, surface state information, etc. of the catalyst, which can provide important clues about the activity of the catalyst and the progress of the reaction. By collecting the real-time state image of the catalyst and analyzing the content in the image, it can help to understand the state of the catalyst in real time, and then achieve more accurate reaction condition regulation, thereby ensuring the stable progress of the reaction.

[0032] In step S42, the catalyst real-time state image is subjected to catalyst state multi-scale feature extraction to obtain a catalyst state shallow feature map and a catalyst state semantic feature map. Specifically, in an embodiment of the present application, the catalyst real-time state image is subjected to catalyst state multi-scale feature extraction to obtain a catalyst state shallow feature map and a catalyst state semantic feature map, including: inputting the catalyst real-time state image into a catalyst state multi-scale feature extractor based on a hollow pyramid network to obtain the catalyst state shallow feature map and the catalyst state semantic feature map. It should be understood that the catalyst real-time state image contains features of different scales, such as shallow features reflecting the texture and color of the catalyst surface and higher-level semantic information. Considering that the hollow convolution in the pyramid network can expand the receptive field without increasing the number of parameters, this enables the network to integrate information over a larger range, thereby better understanding the overall state of the catalyst. At the same time, by setting different hollow rates, the pyramid network can capture contextual information of different scales, thereby better understanding the content in the catalyst real-time state image. Based on this, in order to more effectively capture the various details and deep semantic information of the catalyst surface, in the technical solution of the present application, the real-time state image of the catalyst is input into a catalyst state multi-scale feature extractor based on a hollow pyramid network to obtain a catalyst state shallow feature map and a catalyst state semantic feature map.

[0033] In step S43, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly interactively encoded with catalyst state salient features to obtain catalyst state shallow-semantically salient interactive fusion features. Specifically, Figure 5 In the process of producing synthesis gas by clean conversion of raw coal gas according to the embodiment of the present application, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly interactively encoded with catalyst state significant features to obtain a flow chart of catalyst state shallow-semantically significant sparse interactive fusion features. Figure 5 As shown, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly interactively encoded with catalyst state significant features to obtain catalyst state shallow-semantically significant sparse interactive fusion features, including: S431, kernel sparsification and global association are performed on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain a global coding matrix between catalyst state kernel features; S432, based on the global coding matrix between catalyst state kernel features, mask saliency joint interaction is performed on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain the catalyst state shallow-semantically significant sparse interactive fusion features.

[0034] It should be understood that the catalyst state feature contains multiple levels of information. The catalyst state shallow feature map reflects the basic structure of the catalyst surface, such as edges and textures; while the catalyst state semantic feature map provides higher-level information, such as active sites and overall structure. Therefore, in order to fully understand the state of the catalyst, achieve effective fusion of catalyst state features to better capture key information of the catalyst state, and improve the accuracy and reliability of reaction condition recommendations, in the technical solution of the present application, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly interactively encoded with catalyst state significant features to obtain a catalyst state shallow-semantically significant sparse interactive fusion feature map as a catalyst state shallow-semantically significant sparse interactive fusion feature.

[0035] In detail, first, the catalyst state shallow feature map and the catalyst state semantic feature map are subjected to kernel sparsification and global association to obtain a global coding matrix between the catalyst state kernel features. That is, kernel sparsification refers to retaining the most critical part of the features and removing redundant information through sparsification technology during feature extraction. Global association refers to establishing the relationship between different features in the entire feature map to help the model understand the interdependence between features, so as to better capture the complex changes of the catalyst state. Specifically, first, the catalyst state shallow feature map and the catalyst state semantic feature map are processed by using a kernel feature distillation network to capture and extract the most critical and significant features of the data, and retain the core data in the original features, reduce the interference of redundant information, and obtain the catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector. Then, in order to better characterize and depict the correlation strength and intrinsic pattern information between the two, the catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector are subjected to global association encoding to obtain a global coding matrix between the catalyst state kernel features. In particular, the element values ​​in the global encoding matrix reflect the degree of correlation between each feature dimension, thereby reflecting the potential data patterns and correlations, and providing support for subsequent mask sparsification.

[0036] Then, based on the global coding matrix between the catalyst state kernel features, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly interacted by mask significance to obtain the catalyst state shallow-semantic significant sparse interactive fusion features. In this way, the significant part of the catalyst state feature can be enhanced to fully understand the state of the catalyst and improve the expressive power of the feature. Specifically, by inputting the global coding matrix into the sparse mask unit for mask sparse processing, the most important features for catalyst state evaluation are focused on and strengthened, and unimportant information is removed to obtain the correlation sparse weight matrix between the catalyst state kernel features. At the same time, in order to pay more attention to and focus on the interaction between the two at the local granularity, the catalyst state shallow feature map and the catalyst state semantic feature map are input into the position-by-position correlation interaction network to obtain the catalyst state shallow-semantic position-by-position interactive fusion feature map. Finally, the interactive fusion feature map is sparsely modulated based on the sparse weight matrix to weight the features with different contributions and influences based on the different importance of each weight to obtain the final catalyst state shallow-semantically significant sparse interactive fusion feature map, which can reflect and better express the most relevant and significant features in the original catalytic state data.

[0037] Specifically, Figure 6In the process of producing synthesis gas by clean conversion of raw coal gas according to the embodiment of the present application, kernel sparsification and global association are performed on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain a global encoding matrix between the catalyst state kernel features. Figure 6 As shown, kernel sparsification and global association are performed on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain a global coding matrix between catalyst state kernel features, including: S4311, inputting the catalyst state shallow feature map and the catalyst state semantic feature map into a kernel feature distillation network to obtain a catalyst state shallow kernel feature vector and a catalyst state semantic kernel feature vector; S4312, performing global association encoding on the catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector to obtain a global coding matrix between the catalyst state kernel features.

[0038] More specifically, in an embodiment of the present application, the catalyst state shallow feature map and the catalyst state semantic feature map are input into a kernel feature distillation network to obtain a catalyst state shallow kernel feature vector and a catalyst state semantic kernel feature vector, including: performing feature fine-grained decomposition on the catalyst state shallow feature map to obtain a set of catalyst state shallow feature vectors; calculating the feature difference coefficient of each catalyst state shallow feature vector in the set of catalyst state shallow feature vectors relative to other catalyst state shallow feature vectors to obtain a set of catalyst state shallow feature difference coefficients; normalizing the set of catalyst state shallow feature difference coefficients to obtain a set of normalized catalyst state shallow feature difference coefficients; using each normalized catalyst state shallow feature difference coefficient in the set of normalized catalyst state shallow feature difference coefficients as a weight, calculating the catalyst state shallow feature difference coefficient. The catalyst state shallow kernel feature vector is obtained by weighting and summing a set of feature vectors; wherein, the feature difference coefficient of each catalyst state shallow feature vector in the set of catalyst state shallow feature vectors relative to other catalyst state shallow feature vectors is calculated to obtain a set of catalyst state shallow feature difference coefficients, including: extracting a predetermined catalyst state shallow feature vector from the set of catalyst state shallow feature vectors; subtracting the predetermined catalyst state shallow feature vector from other catalyst state shallow feature vectors in the set of catalyst state shallow feature vectors by position, and calculating a norm of each catalyst state shallow feature difference vector in the obtained set of catalyst state shallow feature difference vectors to obtain a set of catalyst state shallow feature norm values; calculating the global mean of the set of catalyst state shallow feature norm values ​​to obtain the catalyst state shallow feature difference coefficient.

[0039] More specifically, in an embodiment of the present application, the catalyst state shallow feature map is subjected to a fine-grained feature decomposition to obtain a set of catalyst state shallow feature vectors, including: expanding each feature matrix of the catalyst state shallow feature map along the channel dimension into feature vectors to obtain a set of catalyst state shallow feature vectors.

[0040] More specifically, in an embodiment of the present application, the catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector are globally associated and encoded to obtain a global coding matrix between the catalyst state kernel features, including: multiplying the catalyst state shallow kernel feature vector with the transposed vector of the catalyst state semantic kernel feature vector to obtain a matrix as the global coding matrix between the catalyst state kernel features.

[0041] Specifically, Figure 7 In the process of producing synthesis gas by clean conversion of raw coal gas according to the embodiment of the present application, based on the global coding matrix between the catalyst state kernel features, the catalyst state shallow feature map and the catalyst state semantic feature map are subjected to masked significant joint interaction to obtain the catalyst state shallow-semantic significant sparse interactive fusion feature flow chart. Figure 7 As shown, based on the global coding matrix between the catalyst state kernel features, the catalyst state shallow feature map and the catalyst state semantic feature map are subjected to masked saliency joint interaction to obtain the catalyst state shallow-semantic saliency sparse interactive fusion feature, including: S4321, inputting the global coding matrix between the catalyst state kernel features into a sparse mask unit to obtain a catalyst state kernel feature correlation sparse weight matrix; S4322, inputting the catalyst state shallow feature map and the catalyst state semantic feature map into a position-by-position correlation interaction network to obtain a catalyst state shallow-semantic position-by-position interactive fusion feature map; S4323, based on the catalyst state kernel feature correlation sparse weight matrix, The catalyst state shallow-semantic position-by-position interactive fusion feature map is subjected to feature-significant sparse modulation to obtain a catalyst state shallow-semantic significant sparse interactive fusion feature map as the catalyst state shallow-semantic significant sparse interactive fusion feature; wherein, the catalyst state shallow feature map and the catalyst state semantic feature map are input into a position-by-position associated interactive network to obtain a catalyst state shallow-semantic position-by-position interactive fusion feature map, including: performing position-wise multiplication of each feature matrix along the channel dimension in the catalyst state shallow feature map with the feature values ​​of the corresponding positions in each feature matrix along the channel dimension in the catalyst state semantic feature map to obtain the catalyst state shallow-semantic position-by-position interactive fusion feature map.

[0042] More specifically, in an embodiment of the present application, the global coding matrix between the catalyst state kernel features is input into a sparse mask unit to obtain a sparse weight matrix associated with the catalyst state kernel features, including: comparing the eigenvalues ​​of each position in the global coding matrix between the catalyst state kernel features with a preset threshold to obtain a sparse weight matrix associated with the catalyst state kernel features composed of multiple updated eigenvalues; wherein, in response to the eigenvalue being greater than a second preset threshold, the product of the eigenvalue and a constant three is used as the updated eigenvalue; in response to the eigenvalue being greater than a first preset threshold and less than or equal to the second preset threshold, the original value of the eigenvalue is used as the updated eigenvalue; in response to the eigenvalue being less than or equal to the first preset threshold, the value obtained by dividing the eigenvalue by the constant three is used as the updated eigenvalue.

[0043] More specifically, in an embodiment of the present application, based on the sparse weight matrix associated with the catalyst state kernel features, the catalyst state shallow-semantic position-by-position interactive fusion feature map is subjected to feature-significant sparse modulation to obtain a catalyst state shallow-semantic significant sparse interactive fusion feature map, including: matrix multiplying each feature matrix along the channel dimension in the catalyst state shallow-semantic position-by-position interactive fusion feature map with the sparse weight matrix associated with the catalyst state kernel features to obtain the catalyst state shallow-semantic significant sparse interactive fusion feature map.

[0044] In the embodiment of the present application, specifically, the catalyst state shallow feature map and the catalyst state semantic feature map are jointly and interactively encoded with catalyst state significant features, which can be expressed by the formula: ; in, and The catalyst state shallow feature map and the catalyst state semantic feature map are respectively For the feature map Perform kernel feature distillation. and are the catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector, respectively. For the feature map Perform feature fine-grained decomposition operations. is the first, second, ..., first eigenvector in the set of decomposed eigenvectors ,..., ,..., After decomposition, the eigenvectors represents the norm of the eigenvector, is the set of decomposed eigenvectors, represents the number of vectors in the set of decomposed eigenvectors minus one, yes The corresponding characteristic difference coefficient, Expressed as a natural constant The exponential function with base , is the number of vectors in the set of decomposed eigenvectors is the kernel eigenvector, yes The transposed vector of is matrix multiplication, is the global encoding matrix between the catalyst state kernel features, is a sparse masking operation, yes The eigenvalues ​​at each position in , is the preset threshold, is the correlation sparse weight matrix between the catalyst state kernel features, For point-by-position multiplication, is the catalyst state shallow-semantic position-by-position interactive fusion feature map, It is the catalyst state shallow-semantically significant sparse interactive fusion feature map.

[0045] In step S44, based on the shallow-semantic significant sparse interactive fusion feature of the catalyst state, an optimization result is obtained, and the optimization result includes a decoded value of a recommended reaction temperature and a decoded value of a recommended reaction pressure. Specifically, in an embodiment of the present application, based on the shallow-semantic significant sparse interactive fusion feature of the catalyst state, an optimization result is obtained, including: inputting the shallow-semantic significant sparse interactive fusion feature map of the catalyst state into a reaction parameter optimizer based on a decoder to obtain the optimization result. That is, the shallow-semantic significant sparse interactive fusion feature of the catalyst state obtained by significantly interactively fusion of the shallow feature map of the catalyst state and the semantic feature map of the catalyst state is used for decoding processing, so as to automatically recommend the reaction temperature value and the reaction pressure value. In this way, the working state of the catalyst can be monitored in real time, and the change of the catalyst activity can be discovered in time, so as to quickly adjust the reaction conditions and avoid problems caused by changes in the reaction conditions. And the reaction temperature and pressure values ​​can be recommended more accurately to ensure that the reaction conditions are always in the optimal state, thereby improving the efficiency of the reaction and realizing the intelligent management of the reaction process.

[0046] In particular, when the catalyst state shallow feature map and the catalyst state semantic feature map respectively represent the image semantic features of the catalyst real-time state image at different scales and different depths, when the catalyst state salient features are jointly interactively encoded, the catalyst state shallow-semantic salient sparse interactive fusion feature map will also cause joint interactive diversity based on association significance due to the differences in image semantic cross-scale and cross-depth kernel feature patterns, affecting the iterative consistency of the decoding mapping and reducing the accuracy of the decoding results.

[0047] Preferably, the catalyst state shallow-semantically significant sparse interaction fusion feature map is input into a decoder-based reaction parameter optimizer to obtain an optimization result, including: respectively calculating a first feature sum value of all feature values ​​of the catalyst state shallow-semantically significant sparse interaction fusion feature map, a second feature sum value of all feature values ​​of the catalyst state shallow feature map, and a third feature sum value of all feature values ​​of the catalyst state semantic feature map; dividing the first feature sum value by the second feature sum value and the third feature sum value, respectively, and calculating the square root of the sum of the squares of the two quotients to obtain the catalyst state shallow-semantically significant sparse interaction The catalyst state shallow-semantically significant sparse interactive fusion feature map is expanded into a catalyst state shallow-semantically significant sparse interactive fusion feature vector; the catalyst state shallow-semantically significant sparse interactive fusion feature value of the catalyst state shallow-semantically significant sparse interactive fusion feature vector is calculated by multiplying the first catalyst state shallow-semantically significant sparse interactive fusion feature value and the catalyst state shallow-semantically significant sparse interactive fusion distribution probability value to obtain the first catalyst state shallow-semantically significant sparse interactive fusion feature distribution probability product; the catalyst state shallow-semantically significant sparse interactive fusion feature vector is calculated by multiplying the second catalyst state shallow-semantically significant sparse interactive fusion feature value of the catalyst state shallow-semantically significant sparse interactive fusion feature vector by the product of the catalyst state shallow-semantically significant sparse interactive fusion feature distribution probability product. The product of the interactive fusion feature value and the absolute value of the difference between the shallow-semantically significant sparse interactive fusion distribution probability value of the catalyst state is obtained to obtain the second catalyst state shallow-semantically significant sparse interactive fusion feature inverse distribution probability product; the quotient of the first catalyst state shallow-semantically significant sparse interactive fusion feature value divided by the absolute value of the difference between the shallow-semantically significant sparse interactive fusion distribution probability value of the catalyst state is calculated to obtain the first catalyst state shallow-semantically significant sparse interactive fusion feature inverse distribution probability quotient; the second catalyst state shallow-semantically significant sparse interactive fusion feature value and the catalyst state are calculated. The quotient of the shallow-semantic significant sparse interactive fusion distribution probability value of the catalyst state is obtained to obtain the second catalyst state shallow-semantic significant sparse interactive fusion feature distribution probability quotient; after adding the first catalyst state shallow-semantic significant sparse interactive fusion feature distribution probability product and the second catalyst state shallow-semantic significant sparse interactive fusion feature inverse distribution probability product, the difference between the first catalyst state shallow-semantic significant sparse interactive fusion feature inverse distribution probability quotient and the second catalyst state shallow-semantic significant sparse interactive fusion feature distribution probability quotient is subtracted to obtain the catalyst state shallow-semantic significant sparse interactive fusion correction feature value;For each first catalyst state shallow-semantically significant sparse interactive fusion feature value and second catalyst state shallow-semantically significant sparse interactive fusion feature value of the catalyst state shallow-semantically significant sparse interactive fusion feature vector, matrix multiply the catalyst state shallow-semantically significant sparse interactive fusion correction feature matrix composed of the catalyst state shallow-semantically significant sparse interactive fusion correction feature value with the catalyst state shallow-semantically significant sparse interactive fusion feature vector to obtain a corrected catalyst state shallow-semantically significant sparse interactive fusion feature vector; and input the corrected catalyst state shallow-semantically significant sparse interactive fusion feature vector into the decoder-based reaction parameter optimizer to obtain the optimization result. ;

[0048] The correction process of the catalyst state shallow-semantically significant sparse interactive fusion feature vector is expressed as: ;and, ;in, is the catalyst state shallow-semantically significant sparse interactive fusion feature map, is the shallow characteristic diagram of the catalyst state, is the catalyst state semantic feature graph, is the feature value of each position in the catalyst state shallow-semantically significant sparse interactive fusion feature map, represents the characteristic value of each position in the shallow characteristic map of the catalyst state, represents the feature value of each position in the catalyst state semantic feature map, is the probability value of the shallow-semantically significant sparse interactive fusion distribution of the catalyst state, is the catalyst state shallow-semantically significant sparse interactive fusion feature vector, The shallow-semantically significant sparse interactive fusion feature vector of the catalyst state is represented by The first catalyst state shallow-semantically significant sparse interactive fusion feature value at each position, The shallow-semantically significant sparse interactive fusion feature vector of the catalyst state is represented by The second catalyst state shallow-semantically significant sparse interactive fusion feature value at each position, is the catalyst state shallow-semantically significant sparse interactive fusion correction feature matrix The catalyst state of each position is shallow-semantically significant and sparsely interactively fused and corrected feature values. is the catalyst state shallow-semantically significant sparse interactive fusion correction feature matrix, represents matrix multiplication, A shallow-semantically significant sparse interactive fusion feature vector representing the corrected catalyst state.

[0049] That is, the present application is directed to the region-boundary integral relationship of the high-dimensional feature manifold of the shallow-semantic significant sparse interactive fusion feature map of the catalyst state, and by defining the feature set of the shallow-semantic significant sparse interactive fusion feature map of the catalyst state as a whole relative to the spatial distribution probability of the shallow feature map of the catalyst state and the semantic feature map of the catalyst state as a manifold constraint boundary, to approximate the high-dimensional manifold of the feature set of the shallow-semantic significant sparse interactive fusion feature map of the catalyst state, which is composed of a single connected region representation of eigenvalue pairs, so as to avoid the mapping ambiguity of the diversified feature representation of the shallow-semantic significant sparse interactive fusion feature map of the catalyst state to the local manifold representation in the high-dimensional feature decoding regression convergence space, and improve the execution iteration consistency of each local feature distribution in the mapping task, so as to improve the feature decoding regression convergence effect, and improve the accuracy of the optimization result obtained by the decoder-based reaction parameter optimizer. In this way, the working state of the catalyst can be monitored in real time, and the change of catalyst activity can be discovered in time, so as to quickly adjust the reaction conditions and avoid problems caused by changes in reaction conditions. It can also recommend reaction temperature and pressure values ​​more accurately to ensure that the reaction conditions are always in the optimal state, thereby improving the efficiency of the reaction and realizing intelligent management of the reaction process.

[0050] In summary, the process of producing synthesis gas by clean conversion of waste gas based on the embodiment of the present application is explained, which uses image recognition and analysis technology based on machine vision to perform multi-scale feature extraction on the real-time state image of the catalyst collected by the industrial camera, so as to automatically recommend the reaction temperature value and the reaction pressure value based on the significant interactive fusion features between the shallow features of the catalyst state and the semantic features of the catalyst state. In this way, the working state of the catalyst can be monitored in real time, and the changes in the activity of the catalyst can be discovered in time, so as to quickly adjust the reaction conditions and avoid problems caused by changes in the reaction conditions. In addition, this method can realize the intelligent management of the reaction process, which is conducive to more accurately recommending adaptive reaction temperature and pressure values ​​to ensure the stable progress of the reaction process.

Claims

1. A process for producing synthesis gas by clean conversion of waste gas, characterized in that: include: S1: compressing the raw raw coal gas to obtain compressed raw coal gas; S2: extracting hydrogen and carbon monoxide from the compressed raw coal gas; S3: After pressurizing the carbon monoxide, the pressurized carbon monoxide is introduced into a dimethyl oxalate synthesis device to obtain dimethyl oxalate; S4: inputting the dimethyl oxalate and the hydrogen into an ethylene glycol synthesis unit to obtain ethylene glycol.

2. The process for producing synthesis gas by clean conversion of waste gas according to claim 1, characterized in that: After the carbon monoxide is pressurized, the pressurized carbon monoxide is introduced into a dimethyl oxalate synthesis device to obtain dimethyl oxalate, comprising: subjecting the pressurized carbon monoxide to a synthetic treatment to obtain synthetic dimethyl oxalate; The synthetic dimethyl oxalate is subjected to rectification, decompression and impurity removal treatments to obtain the dimethyl oxalate.

3. The process for producing synthesis gas by clean conversion of waste gas according to claim 2 is characterized in that: The dimethyl oxalate and the hydrogen are introduced into an ethylene glycol synthesis unit to obtain ethylene glycol, comprising: Obtain real-time status images of the catalyst collected by industrial cameras; Performing catalyst state multi-scale feature extraction on the catalyst real-time state image to obtain a catalyst state shallow feature map and a catalyst state semantic feature map; The catalyst state shallow feature map and the catalyst state semantic feature map are jointly interactively encoded with catalyst state significant features to obtain catalyst state shallow-semantically significant sparse interactive fusion features, including: performing kernel sparsification and global association on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain a global encoding matrix between catalyst state kernel features; based on the global encoding matrix between catalyst state kernel features, performing masked saliency joint interaction on the catalyst state shallow feature map and the catalyst state semantic feature map to obtain the catalyst state shallow-semantically significant sparse interactive fusion features; Based on the shallow-semantically significant sparse interactive fusion features of the catalyst state, an optimization result is obtained, and the optimization result includes a decoded value of a recommended reaction temperature and a decoded value of a recommended reaction pressure.

4. The process for producing synthesis gas by clean conversion of waste gas according to claim 3 is characterized in that: Performing catalyst state multi-scale feature extraction on the catalyst real-time state image to obtain a catalyst state shallow feature map and a catalyst state semantic feature map, including: inputting the catalyst real-time state image into a catalyst state multi-scale feature extractor based on a hollow pyramid network to obtain the catalyst state shallow feature map and the catalyst state semantic feature map.

5. The process for producing synthesis gas by clean conversion of waste gas according to claim 4 is characterized in that: The catalyst state shallow feature map and the catalyst state semantic feature map are subjected to kernel sparsification and global association to obtain a global encoding matrix between catalyst state kernel features, including: Inputting the catalyst state shallow feature map and the catalyst state semantic feature map into a kernel feature distillation network to obtain a catalyst state shallow kernel feature vector and a catalyst state semantic kernel feature vector; The catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector are globally associated and encoded to obtain a global encoding matrix between the catalyst state kernel features.

6. The process for producing synthesis gas by clean conversion of waste gas according to claim 5, characterized in that: Inputting the catalyst state shallow feature map and the catalyst state semantic feature map into a kernel feature distillation network to obtain a catalyst state shallow kernel feature vector and a catalyst state semantic kernel feature vector, including: Performing feature fine-grained decomposition on the catalyst state shallow feature map to obtain a set of catalyst state shallow feature vectors; Calculating the characteristic difference coefficient of each catalyst state shallow feature vector relative to other catalyst state shallow feature vectors in the set of catalyst state shallow feature vectors to obtain a set of catalyst state shallow feature difference coefficients; Normalizing the set of catalyst state shallow characteristic difference coefficients to obtain a set of normalized catalyst state shallow characteristic difference coefficients; Using each normalized catalyst state shallow feature difference coefficient in the set of normalized catalyst state shallow feature difference coefficients as a weight, calculating the weighted sum of the set of catalyst state shallow feature vectors to obtain the catalyst state shallow kernel feature vector; The step of calculating the characteristic difference coefficient of each catalyst state shallow feature vector relative to other catalyst state shallow feature vectors in the set of catalyst state shallow feature vectors to obtain a set of catalyst state shallow feature difference coefficients includes: extracting a predetermined catalyst state shallow feature vector from the set of catalyst state shallow feature vectors; Subtracting the predetermined catalyst state shallow feature vector from other catalyst state shallow feature vectors in the set of catalyst state shallow feature vectors by position, and calculating a norm of each catalyst state shallow feature difference vector in the obtained set of catalyst state shallow feature difference vectors to obtain a set of catalyst state shallow feature norm values; The global mean of the set of catalyst state shallow feature norm values ​​is calculated to obtain the catalyst state shallow feature difference coefficient.

7. The process for producing synthesis gas by clean conversion of waste gas according to claim 6, characterized in that: The catalyst state shallow kernel feature vector and the catalyst state semantic kernel feature vector are globally associated with each other to obtain a global coding matrix between the catalyst state kernel features, including: multiplying the catalyst state shallow kernel feature vector with the transposed vector of the catalyst state semantic kernel feature vector to obtain a matrix as the global coding matrix between the catalyst state kernel features.

8. The process for producing synthesis gas by clean conversion of waste gas according to claim 7 is characterized in that: Based on the global encoding matrix between the catalyst state kernel features, the catalyst state shallow feature map and the catalyst state semantic feature map are subjected to mask saliency joint interaction to obtain the catalyst state shallow-semantic saliency sparse interaction fusion feature, including: Inputting the catalyst state kernel feature global encoding matrix into a sparse mask unit to obtain a catalyst state kernel feature correlation sparse weight matrix; Inputting the catalyst state shallow feature map and the catalyst state semantic feature map into a position-by-position association interaction network to obtain a catalyst state shallow-semantic position-by-position interaction fusion feature map; Based on the correlation sparse weight matrix between the catalyst state kernel features, the catalyst state shallow-semantic position-by-position interactive fusion feature map is subjected to feature significant sparse modulation to obtain a catalyst state shallow-semantic significant sparse interactive fusion feature map as the catalyst state shallow-semantic significant sparse interactive fusion feature; Among them, the catalyst state shallow feature map and the catalyst state semantic feature map are input into a position-by-position association interaction network to obtain a catalyst state shallow-semantic position-by-position interactive fusion feature map, including: performing positional point multiplication on the feature values ​​of corresponding positions in each feature matrix along the channel dimension in the catalyst state shallow feature map and each feature matrix along the channel dimension in the catalyst state semantic feature map to obtain the catalyst state shallow-semantic position-by-position interactive fusion feature map.

9. The process for producing synthesis gas by clean conversion of waste gas according to claim 8, characterized in that: Inputting the catalyst state kernel feature global encoding matrix into the sparse mask unit to obtain the catalyst state kernel feature correlation sparse weight matrix, including: Comparing the eigenvalues ​​of each position in the global encoding matrix between the catalyst state kernel features with a preset threshold value to obtain a sparse weight matrix of correlation between the catalyst state kernel features composed of a plurality of updated eigenvalues; wherein, in response to the eigenvalue being greater than a second preset threshold, the product of the eigenvalue and a constant three is used as the updated eigenvalue; In response to the feature value being greater than a first preset threshold and less than or equal to the second preset threshold, taking the original value of the feature value as the updated feature value; In response to the eigenvalue being less than or equal to the first preset threshold, a value obtained by dividing the eigenvalue by constant three is used as the updated eigenvalue.

10. The process for producing synthesis gas by clean conversion of waste gas according to claim 9, characterized in that: Based on the catalyst state shallow-semantically significant sparse interactive fusion features, an optimization result is obtained, including: inputting the catalyst state shallow-semantically significant sparse interactive fusion feature map into a decoder-based reaction parameter optimizer to obtain the optimization result.

Images