Multi-purpose modular operation optimization system for municipal solid waste incineration process

Abstract

Provided in the present invention is a multi-purpose modular operation optimization system for a municipal solid waste incineration process. The system comprises a multi-modal historical data synchronous driving module, an MSWI-process virtual control object module, an MSWI-process loop control module, an MSWI-process monitoring module, an operation parameter assisted decision-making module, a data-acquisition forward isolation module, an operation parameter reverse transmission module, an MSWI-process single-objective / multi-objective operation optimization module, a difficult-to-measure procedure parameter soft-measurement module, a multi-modal data driven procedure parameter prediction module, a vision-driven combustion state identification module, a flame combustion line quantification module and a multi-modal data acquisition module. The multi-purpose modular operation optimization system for a municipal solid waste incineration process provided in the present invention can achieve functions such as laboratory operation-optimization algorithm verification, laboratory procedure-parameter-modeling algorithm simulation and real-time verification, industrial field data acquisition and procedure parameter modeling, and industrial field assisted decision-making and operation optimization, and the system is convenient to use.

Description

[Corrected 11.05.2024 according to Regulation 26] Multi-purpose modular operation optimization system for municipal solid waste incineration process Technical Field

[0001] The present invention relates to the technical field of municipal solid waste incineration, and in particular to a multi-purpose modular operation optimization system for a municipal solid waste incineration process. Background Art

[0002] With a global annual growth rate of 8% to 10% for municipal solid waste (MSW), MSW incineration (MSWI), characterized by its harmlessness, volume reduction, and resource utilization, has become one of the primary technologies for addressing my country's "garbage siege" problem. The MSWI process is a typical process industry. While self-sufficient in energy, it can also provide various forms of energy, such as electricity and heat, while also minimizing the risk of environmental pollution emissions. Studies have shown that the mass reduction, volume reduction, and energy recovery rates of MSWI can reach 70%, 90%, and 19%, respectively. Its potential economic and environmental benefits have been recognized by developing countries.

[0003] The MSWI process plays a key role in low-carbon development, environmental protection, and sustainable energy, and has become a foundational industry for my country's ecological civilization development and circular economy in the new era. In comparison, my country's MSW classification policies and corresponding management systems are still under development and being promoted. The composition of collected MSW is characterized by high uncertainty, low calorific value, and high volatility. Therefore, foreign ACC systems are difficult to directly apply to the operational control of my country's MSWI process. Currently, domestic incineration plants primarily utilize a model in which domain experts (i.e., knowledge workers) manually implement operational rules based on mechanistic and empirical understanding of operating condition fluctuations to meet multiple scenario requirements. This manual control model, characterized by intelligent autonomous behavior, is plagued by issues such as limited expert attention, experience differences, and subjective control. This makes it difficult to ensure continuous operational stability, hindering the pollution and carbon reduction efforts of MSWI power plants. According to the "Automatic Monitoring Data Publicity System for Municipal Waste Incineration Power Plants," 21 MSWI power plants have closed in my country since 2020, involving over 50 incinerators. Therefore, it is imperative to independently develop intelligent operational optimization technologies tailored to the characteristics of my country's MSW. Safety concerns for MSWI power plants and the closed nature of their distributed control systems (DCSs) make it difficult to interact with external algorithms. Specifically, non-enterprise systems cannot directly collect data or write parameters to the MSWI control system. These limitations hinder online verification of modeling, control, and optimization algorithms developed in the laboratory for the MSWI process. Therefore, an operational optimization algorithm verification system is essential for enabling the practical application of laboratory-based theoretical and technological research. Furthermore, a prerequisite for algorithm development in the laboratory is the availability of real-time multimodal data from the industrial site, including video and process data. Furthermore, consideration must be given to ensuring that the operational optimization algorithms, after offline verification, are gradually recognized by field experts in the industrial field for practical application. Furthermore, the applicability and portability of the laboratory hardware and software used to implement the operational optimization algorithms in the industrial field are crucial. Therefore, the design of a multipurpose, modular operational optimization system for municipal solid waste incineration processes is highly desirable.

[0004] Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-purpose modular operation optimization system for the municipal solid waste incineration process, which can realize laboratory operation optimization algorithm verification, laboratory process parameter modeling algorithm simulation real-time verification, industrial field data collection and process parameter modeling, industrial field auxiliary decision-making and operation optimization and other functions, and is easy to use.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A multi-purpose modular operation optimization system for a municipal solid waste incineration process comprises: a multi-modal historical data synchronization drive module, an MSWI process virtual control object module, an MSWI process loop control module, an MSWI process monitoring module, an operation parameter auxiliary decision module, a data acquisition forward isolation module, an operation parameter reverse transmission module, an MSWI process single-objective / multi-objective operation optimization module, a difficult-to-measure process parameter soft measurement module, a multi-modal data-driven process parameter prediction module, a visually driven combustion state recognition module, a flame combustion line quantification module and a multi-modal data acquisition module. The multi-modal historical data synchronization drive module is connected to the multi-modal data acquisition module, the multi-modal data acquisition module is connected to the flame combustion line quantification module, the visually driven combustion state recognition module, the multi-modal data-driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the MSWI process single-objective / multi-objective operation optimization module. The flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the MSWI process single-objective / multi-objective operation optimization module are connected to the operation parameter auxiliary decision module and the operation parameter reverse transmission module, the operation parameter reverse transmission module is connected to the operation parameter auxiliary decision module, the operation parameter auxiliary decision module is connected to the MSWI process monitoring module and the field process monitoring system, the field process monitoring system is connected to the data acquisition forward isolation module, the data acquisition forward isolation module is connected to the multimodal data acquisition module, the MSWI process monitoring module is connected to the MSWI process loop control module, the MSWI process loop control module is connected to the MSWI process virtual control object module, the field process monitoring system is connected to the field loop control system, and the field loop control system is connected to the actuator and instrumentation device;

[0008] The multimodal historical data synchronization driving module is used to provide a multimodal data source for the MSWI process;

[0009] The MSWI process virtual control object module is used to simulate the MSWI process built in the laboratory;

[0010] The MSWI process loop control module is used to implement loop control of the virtual MSWI process;

[0011] The MSWI process monitoring module is used to monitor the virtual MSWI process;

[0012] The operating parameter auxiliary decision module is used to obtain operating parameters and conduct comparative analysis and decision making on them;

[0013] The data acquisition forward isolation module is used to realize data acquisition of all process variables in the monitoring module of the virtual MSWI process through physical isolation;

[0014] The operating parameter reverse transmission module is used to reversely transmit the operating parameter optimization values ​​obtained from the MSWI process single-objective / multi-objective operation optimization module, the operating parameters of the flame combustion line quantification module, the visual-driven combustion state recognition module, the multimodal data-driven process parameter prediction module, and the operating parameter detection values ​​of the difficult-to-measure process parameter soft measurement module in a physically isolated manner;

[0015] The MSWI process single-objective / multi-objective operation optimization module is used to optimize the MSWI process operation parameters based on multimodal data and hard-to-measure parameter soft measurement models;

[0016] The difficult-to-measure process parameter soft measurement module is used to implement soft measurement modeling of difficult-to-measure parameters based on multimodal data and production reports;

[0017] The multimodal data driven process parameter prediction module is used to realize single-step / multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, etc. based on multimodal data;

[0018] The visually driven combustion state recognition module is used to realize the recognition of the combustion state in the furnace based on the domain expert recognition mechanism for the MSWI process;

[0019] The flame burning line quantification module is used to realize the quantification of the flame burning line based on the domain expert recognition mechanism for the MSWI process;

[0020] The multimodal data acquisition module is used to realize the collection of simulated and actual left grate flame videos, right grate flame videos and historical process data of the MSWI process with multimodal data, as well as the processing and entry of various production reports related to product quality, environmental protection indicators and economic indicators generated by the actual process.

[0021] Optionally, the multimodal historical data synchronization drive module, the multimodal data acquisition module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module, the MSWI process single-objective / multi-objective operation optimization module, the operation parameter auxiliary decision module, the MSWI process monitoring module, the MSWI process loop control module and the MSWI process virtual control object module together constitute a laboratory operation optimization algorithm verification subsystem, which obtains the synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module through the multimodal data acquisition module, and obtains the synchronized left grate and right grate flame image data and historical process data through the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, The difficult-to-measure process parameter soft measurement module and the processing of the difficult-to-measure process parameter soft measurement module obtain various categories of operating parameters, and transmit the operating parameters to the MSWI process monitoring module through OPC, and download them to the MSWI process loop control module based on real PLC / DCS equipment, and then transmit the control quantity to the virtual actuator of the MSWI process virtual control object module in the form of analog output. The output of the virtual actuator then acts on the virtual object to generate a controlled variable output, which is transmitted to the MSWI process loop control module in the form of analog input through the virtual instrument device, and then transmitted to the MSWI process monitoring module, and then transmitted to the multimodal data acquisition module through OPC, and then fed back to the difficult-to-measure process parameter soft measurement module to complete the laboratory-oriented operation optimization algorithm verification.

[0022] Optionally, the multimodal historical data synchronization drive module, the multimodal data acquisition module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module and the difficult-to-measure process parameter soft measurement module together constitute a laboratory process parameter modeling algorithm simulation real-time verification subsystem, which obtains the synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module through the multimodal data acquisition module, and realizes the process parameter prediction, combustion state recognition, combustion line quantification and process parameter soft measurement results of the simultaneous release of real-time multimodal data from the industrial site through the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module and the difficult-to-measure process parameter soft measurement module.

[0023] Optionally, the data forward acquisition isolation module, multimodal data acquisition module, flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module and difficult-to-measure process parameter soft measurement module together constitute an industrial field data acquisition and process parameter modeling subsystem, which transmits the industrial field data to the forward server via OPC, and then transmits it to the multimodal data acquisition module after physical isolation forward acquisition. Through the flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module and difficult-to-measure process parameter soft measurement module, process parameter prediction, combustion state recognition, combustion line quantification and process parameter soft measurement results based on the synchronous release of real-time multimodal data from the industrial field are realized.

[0024] Optionally, the data forward acquisition isolation module, multimodal data acquisition module, flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module, difficult-to-measure process parameter soft measurement module, MSWI process single-objective / multi-objective operation optimization module, operation parameter reverse transmission module and operation parameter auxiliary decision module together constitute an industrial field auxiliary decision and operation optimization subsystem. The process data is transmitted to the forward server of the data acquisition forward isolation module via OPC, and is transmitted to the multimodal data acquisition module after physical isolation and forward acquisition. After processing by the flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module, difficult-to-measure process parameter soft measurement module and MSWI process single-objective / multi-objective operation optimization module, various categories of key operation parameter prediction values ​​and soft measurement values ​​as well as optimized operation parameter values ​​are obtained. After being transmitted to the reverse server of the operating parameter reverse transmission module, it is reversely transmitted to the operating parameter reverse receiving server through physical isolation. After the operating parameter auxiliary decision-making module performs auxiliary decision-making analysis, the optimized operating parameters are transmitted to the on-site monitoring system and PLC / DCS system using the OPC protocol or OCR recognition method according to the safety requirements of the MSWI plant. The control quantity is then transmitted to the actual MSWI process actuator in the form of analog output, acting on the actual object consisting of solid waste storage and transportation, solid waste combustion, waste heat exchange, flue gas purification and flue gas emission stages. The data is then collected through analog input by the instrument device to the PLC / DCS system and on-site monitoring system, and then transmitted to the data acquisition forward isolation module and multimodal data acquisition module through OPC, and then fed back to the MSWI process single-objective / multi-objective operation optimization module to complete the implementation of the operation optimization algorithm for the actual industrial site.

[0025] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the multi-purpose modular operation optimization system for the urban solid waste incineration process provided by the present invention includes a multi-modal historical data synchronization drive module, an MSWI process virtual control object module, an MSWI process loop control module, an MSWI process monitoring module, an operation parameter auxiliary decision module, a data acquisition forward isolation module, an operation parameter reverse transmission module, an MSWI process single-objective / multi-objective operation optimization module, a difficult-to-measure process parameter soft measurement module, a multi-modal data-driven process parameter prediction module, a visually driven combustion state recognition module, a flame combustion line quantification module and a multi-modal data acquisition module. The system can be simultaneously applied to actual Laboratories and industrial sites, avoiding the shortcomings of traditional simulation experiment systems that are too complex or too simple, can be modularly built according to specific needs, and can effectively meet the isolated data collection and software transplantation needs of industrial sites. The system can effectively integrate multi-modal data synchronous prediction and operation optimization control. The system takes into account the security requirements of different MSWI companies and can choose OPC server or OCR recognition method to transmit operation parameters. It has strong practicality and adaptability. The system can realize laboratory operation optimization algorithm verification, laboratory process parameter modeling algorithm simulation real-time verification, industrial site data collection and process parameter modeling, industrial site decision-making assistance and operation optimization and other functions, and is easy to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 is a process flow chart of the municipal solid waste incineration process;

[0028] FIG2 is a schematic diagram of the structure of a multi-purpose modular operation optimization system for a municipal solid waste incineration process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] The purpose of the present invention is to provide a multi-purpose modular operation optimization system for the municipal solid waste incineration process, which can realize laboratory operation optimization algorithm verification, laboratory process parameter modeling algorithm simulation real-time verification, industrial field data collection and process parameter modeling, industrial field auxiliary decision-making and operation optimization and other functions, and is easy to use.

[0031] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] The process flow of municipal solid waste incineration is shown in Figure 1.

[0033] As shown in Figure 2, the multi-purpose modular operation optimization system for the municipal solid waste incineration process provided by an embodiment of the present invention includes: a multi-modal historical data synchronization drive module, an MSWI process virtual control object module, an MSWI process loop control module, an MSWI process monitoring module, an operation parameter auxiliary decision module, a data acquisition forward isolation module, an operation parameter reverse transmission module, an MSWI process single-objective / multi-objective operation optimization module, a difficult-to-measure process parameter soft measurement module, a multi-modal data-driven process parameter prediction module, a visually driven combustion state recognition module, a flame combustion line quantification module and a multi-modal data acquisition module. The multi-modal historical data synchronization drive module is connected to the multi-modal data acquisition module, and the multi-modal data acquisition module is connected to the flame combustion line quantification module, the visually driven combustion state recognition module, the multi-modal data-driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the MSWI process single-objective / multi-objective operation optimization module. Module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the MSWI process single-objective / multi-objective operation optimization module are connected to the operation parameter auxiliary decision module and the operation parameter reverse transmission module, the operation parameter reverse transmission module is connected to the operation parameter auxiliary decision module, the operation parameter auxiliary decision module is connected to the MSWI process monitoring module and the field process monitoring system, the field process monitoring system is connected to the data acquisition forward isolation module, the data acquisition forward isolation module is connected to the multimodal data acquisition module, the MSWI process monitoring module is connected to the MSWI process loop control module, the MSWI process loop control module is connected to the MSWI process virtual control object module, the field process monitoring system is connected to the field loop control system, and the field loop control system is connected to the actuator and instrumentation device;

[0034] In Figure 2 , the dotted lines connecting the modules represent the connection method unique to the laboratory operation optimization algorithm verification subsystem, the solid lines represent the connection method common to all four subsystems, and the bold boxes represent the components of the actual MSWI process and related systems. The operating parameter decision-making support module differs between laboratory and industrial field applications. Furthermore, the subsystems that can be implemented based on the 13 modules in Figure 2 include, but are not limited to, the four subsystems mentioned above. The functions of each of the 13 modules are described below:

[0035] The multimodal historical data synchronization drive module is designed for a simulated MSWI process with multimodal data built in a laboratory. It enables the simultaneous release of historical left grate flame videos, right grate flame videos, and historical process data, providing a multimodal data source for the MSWI process in a laboratory setting.

[0036] The MSWI process virtual control object module: This module is designed for simulating MSWI processes with multimodal data built in the laboratory. It uses models built in the virtual actuator computer, virtual object computer, and virtual instrument device computer to simulate MSWI processes that are difficult to build in the laboratory.

[0037] The MSWI process loop control module: This module implements loop control of the virtual MSWI process;

[0038] The MSWI process monitoring module: This module implements monitoring of the virtual MSWI process;

[0039] The operating parameter auxiliary decision module: After obtaining relevant operating parameters from the operating parameter reverse receiving server, this module directly transmits these operating parameters to the M9 module or the on-site process monitoring system through comparative analysis and decision-making of the operating parameters, or transmits the operating parameters to the on-site process monitoring system through operating parameter OCR recognition;

[0040] The data acquisition forward isolation module: This module collects data of all process variables in the MSWI process monitoring module through physical isolation, avoiding any impact on the original control system of the MSWI process;

[0041] The operating parameter reverse transmission module: This module realizes the reverse transmission of the operating parameter optimization values ​​obtained from the MSWI process single-objective / multi-objective operation optimization module, as well as the operating parameters of the flame combustion line quantification module, the visual-driven combustion state recognition module, the multimodal data-driven process parameter prediction module, and the operating parameter detection values ​​of the M6-difficult-to-measure process parameter soft measurement module in a physically isolated manner, to avoid affecting the original control system of the MSWI process;

[0042] The MSWI process single-objective / multi-objective operation optimization module: This module optimizes the MSWI process operating parameters based on multimodal data and hard-to-measure parameter soft sensor models. It mainly optimizes the quality indicators and environmental indicators of the MSWI process to obtain the optimized operating parameter values ​​required by the M10 module and the on-site process monitoring system.

[0043] The difficult-to-measure process parameter soft measurement module: This module implements soft measurement modeling of difficult-to-measure parameters such as product quality parameter slag thermal loss rate and environmental indicator parameter dioxin based on multimodal data and production reports, providing support for single-objective / multi-objective operation optimization of the MSWI process;

[0044] The multimodal data-driven process parameter prediction module: This module implements single-step / multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, and steam flow based on multimodal data, providing support for loop control of the MSWI process;

[0045] The visually driven combustion state recognition module: This module implements the recognition of the combustion state in the furnace based on the domain expert recognition mechanism for the MSWI process, providing support for the loop control of the MSWI process;

[0046] The flame combustion line quantification module: This module implements the quantification of the flame combustion line based on the domain expert recognition mechanism for the MSWI process, providing support for the loop control of the MSWI process;

[0047] The multimodal data acquisition module: This module realizes the collection of simulated and actual left grate flame videos, right grate flame videos and historical process data of the MSWI process with multimodal data, as well as the processing and entry of various production reports related to product quality, environmental protection indicators and economic indicators generated in the actual process, providing data support for the M3-M7 modules.

[0048] The present invention is based on 13 modules and has 4 subsystems, which are introduced separately.

[0049] The multimodal historical data synchronization drive module, the multimodal data acquisition module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module, the MSWI process single-objective / multi-objective operation optimization module, the operation parameter auxiliary decision module, the MSWI process monitoring module, the MSWI process loop control module and the MSWI process virtual control object module together constitute a laboratory operation optimization algorithm verification subsystem, which obtains the synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module through the multimodal data acquisition module, and obtains the synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module through the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the difficult-to-measure process parameter The soft measurement module processes and obtains various types of operating parameters. These operating parameters are transmitted to the MSWI process monitoring module via OPC and downloaded to the MSWI process loop control module based on real PLC / DCS equipment. The controlled quantity is then transmitted to the virtual actuator of the MSWI process virtual control object module as an analog output. The output of the virtual actuator then acts on the virtual object to generate a controlled variable output. This is transmitted to the MSWI process loop control module as an analog input through the virtual instrument device, and then to the MSWI process monitoring module. It is then transmitted to the multimodal data acquisition module via OPC, and then fed back to the difficult-to-measure process parameter soft measurement module, completing the laboratory-oriented operation optimization algorithm verification. The functions of each module in the laboratory operation optimization algorithm verification subsystem are introduced respectively:

[0050] Multimodal historical data synchronization driver module: To reflect the dynamic characteristics of the MSWI process and the actual industrial conditions, the collected historical process data and flame video data are published synchronously in real time. It includes four components: network time server, historical right grate flame video publishing, historical left grate flame video publishing, and historical process data publishing. The specific implementation steps are as follows:

[0051] (1) The historical process data of the actual MSWI plant is stored in the form of files in the historical process data OPC server, and the data is published within the local area network through the OPC protocol;

[0052] (2) Based on the actual MSWI plant, the incinerator flame monitoring is divided into the left and right sides, and real-time playback is performed in two image simulation computers respectively;

[0053] (3) Connect the historical data OPC server, the left grate burning image simulator, the right grate burning image simulator and the network time server to the local area network through the switch;

[0054] (4) Through the network time server, the time of process data and image video is precisely controlled to be at the same moment, thereby realizing the simultaneous display of multimodal information.

[0055] Multimodal data acquisition module: Incinerator flame and process data are important bases for field experts to judge the stability of the MSWI process. Real-time acquisition and preprocessing of these multimodal data are particularly important. The specific steps for multimodal data acquisition include:

[0056] (1) Using two identical sets of video cameras, the flame videos of the left and right grates are captured online in real time;

[0057] (2) Transmitting the camera collected information to the video capture card via coaxial cable;

[0058] (3) Install the video acquisition card into the multimodal data acquisition computer and realize the preliminary display of flame information through video decoding;

[0059] (4) Preprocessing the collected flame video by denoising and enhancing it through image processing algorithms;

[0060] (5) Collecting process data from the OPC server to a multimodal data acquisition computer via industrial Ethernet in the form of an OPC client;

[0061] (6) The production reports are collected into the multimodal data collection computer by manual input by domain experts, and the sampling period is reasonably set to realize the synchronous storage of multimodal data after collection.

[0062] Flame combustion line quantification module: The flame combustion line is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve quantification include:

[0063] (1) Using image processing technology, conditional generative adversarial networks, and cycle-consistent generative adversarial networks, a complete image library is constructed, which includes a real normal combustion line sub-library, a real / generated abnormal combustion line sub-library, and a generated extremely abnormal combustion line flame image sub-library;

[0064] (2) Select typical images from the complete image library to form a typical template library and train the twin convolutional neural network;

[0065] (3) Extract the burning line features from the new flame image and use the similarity metric of the twin convolutional neural network to achieve adaptation with the flame image of the “typical template library”;

[0066] (4) Non-adaptive new flame image, using the nearest neighbor criterion to achieve combustion line quantification;

[0067] (5) Based on the redundant discrimination mechanism and combined with the experience of domain experts, the “typical template library” is adaptively updated using non-adaptive images.

[0068] Visually driven combustion state recognition module: The flame combustion state is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve visually driven combustion state recognition include:

[0069] (1) The flame image is dehazed and denoised using a dehazing algorithm based on artificial multi-exposure image fusion, feature normalization, notch filtering, median filtering and other preprocessing methods to obtain a clear image;

[0070] (2) Extract multiple features with physical meanings such as brightness, texture, and color of flame images to represent the image from multiple viewpoints, and simplify these features based on mutual information;

[0071] (3) The above simplified features are used as inputs to image classifiers such as support vector machines, deep forests, and convolutional neural networks to establish visually driven combustion state recognition models for the left and right grates respectively;

[0072] (4) For each new flame image, the combustion state represented by it is recognized.

[0073] Multimodal data-driven process parameter prediction module: Achieving one-step or even multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration is crucial for domain experts. The specific steps for achieving multimodal data-driven process parameter prediction include:

[0074] (1) Using the multimodal data stored in the M2 module, the flame image features are extracted according to the M4 module;

[0075] (2) Combining flame characteristics and process data serially into new features for training key process parameter prediction models;

[0076] (3) The multimodal process data newly collected by the M2 module is input into the above-mentioned key process parameter prediction model to obtain one-step or even multi-step prediction outputs of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration.

[0077] Soft measurement module for difficult-to-measure process parameters: This module implements soft measurement modeling for difficult-to-measure parameters such as product quality parameter slag thermal loss rate and environmental indicator parameter dioxin. This is crucial for single-objective / multi-objective operation optimization of the MSWI process. The specific steps for achieving prediction of difficult-to-measure process parameters include:

[0078] (1) Organize the difficult-to-measure process parameter data recorded in various production reports as the output true value of the soft sensor model;

[0079] (2) Select the multimodal data stored in the M2 module based on the true value of the soft measurement output of the difficult-to-measure process parameter, that is, obtain the multimodal data time period corresponding to the true value of the soft measurement output;

[0080] (3) For the flame data within the time period corresponding to the true value of the soft measurement output, the flame image features are extracted according to the M4 module, and the features are serially combined with the process data to form new features for the soft measurement model input;

[0081] (4) Using the above simplified features as input to regression models such as support vector machines, deep forests, and deep neural networks to establish soft measurement models for difficult-to-detect process parameters;

[0082] (5) The multimodal process data newly collected by the M2 module is input into the above-mentioned soft measurement model of difficult-to-measure process parameters to obtain soft measurement values ​​of difficult-to-measure process parameters such as slag thermal loss rate and dioxin concentration.

[0083] The MSWI process single-objective / multi-objective operation optimization module: In different scenarios, considering single or multiple objectives such as flue gas emission indicators, economic benefits, and slag heat reduction rate, the set values ​​of key controlled variables such as furnace temperature, flue gas oxygen content, and steam flow rate are adjusted in a timely manner as the MSWI process changes dynamically. This is very important for achieving optimal control of the MSWI process. The specific steps for achieving single-objective / multi-objective operation optimization of the MSWI process include:

[0084] (1) Based on the MSWI process analysis for single-objective / multi-objective optimization, a single-objective / multi-objective optimization model was established with key controlled variables such as furnace temperature, boiler steam flow rate, and flue gas oxygen content as decision variables to minimize pollutant emission indicators and heat reduction rate and maximize combustion efficiency and economic indicators;

[0085] (2) Based on the above-mentioned single / multi-objective optimization model, the multimodal data collected in real time by the M2 module and the soft measurement model of difficult-to-measure process parameters of the M6 ​​module are used to obtain the optimized set values ​​of key controlled variables using intelligent optimization algorithms such as genetic algorithm, particle swarm optimization algorithm, and differential evolution algorithm.

[0086] Operating parameter decision-making support module: The single-step / multi-step prediction values ​​and soft measurement values ​​of key operating parameter models are crucial for domain experts to execute control strategies and also determine whether to apply optimized operating parameters to the control of the MSWI process. The steps to implement operating parameter decision-making support are as follows:

[0087] (1) Based on the M3-M6 modules, the single-step / multi-step prediction values ​​and soft measurement values ​​of the key operating parameter model are collected to the operating parameter reverse receiving server;

[0088] (2) Based on the single-step / multi-step prediction values ​​and soft measurement values ​​of the key operating parameter models in different time periods, a statistically significant comparative analysis is conducted with the actual values ​​of these key operating parameters. Active learning by domain experts and automatic discrimination by setting thresholds are used to assist in the decision-making of the credibility of the single-step / multi-step prediction values ​​and soft measurement values ​​of these key operating parameter models.

[0089] (3) Based on the credibility of the single-step / multi-step prediction values ​​and soft measurement values ​​of the above key operating parameter models and the optimized operating parameters from M7, the credible operating parameters of domain expert decisions or automatic decisions are transmitted to the M10 module.

[0090] MSWI process monitoring module: includes functions such as decision-making on whether to use optimized operating parameters and real-time monitoring of the operating process. The latter includes interfaces such as the combustion process, grate operating status, boiler status, flue gas treatment, variable trend chart, and parameter setting. The specific steps are as follows:

[0091] (1) Develop a configuration monitoring system that includes interfaces for combustion process, grate operation status, boiler status, flue gas treatment, variable trend charts, and parameter settings to monitor the MSWI process in real time and display data;

[0092] (2) The process variable values ​​sent to the OPC Server in real time from the loop control module are received by the OPC Client and displayed graphically on the interfaces such as the incineration process and grate operation status to realize the monitoring function of the entire process, and are transmitted to the M2 module at the same time;

[0093] (3) Make decisions on whether to use optimized operating parameters based on production requirements, production indicators, and expert experience, and then set and modify control loop parameters;

[0094] (4) Download the determined control loop parameters to the M11-MSWI process loop control module.

[0095] MSWI process loop control module: This module builds a logical loop control system for the MSWI process based on the manufacturer's PLC / DCS equipment. The specific steps for implementing MSWI process loop control include:

[0096] (1) Power on and start the hardware based on the CPU, input, output, and communication modules of the PLC / DCS manufacturer;

[0097] (2) Connect to the M10 module via industrial Ethernet to achieve control network communication;

[0098] (3) Use ladder diagram language to write the start and stop, PID, alarm and interlock control programs of the MSWI process to realize the logic loop control function;

[0099] (4) The connection with the M12 module virtual actuator computer is realized through the AO / DO module, and the connection with the M12 module virtual instrument device computer is realized through the AI / DI module to realize data interaction with the M12 module.

[0100] MSWI process virtual control object module: The virtual actuators and detection instruments in this module need to effectively exchange data with the M11 module to support the latter's operation. The implementation steps of the MSWI process virtual control object are as follows:

[0101] (1) Based on the actual operating data of actuators such as dampers, fans, and hydraulic drives, a data-driven virtual actuator model is constructed; based on the actual operating data of sensor devices such as temperature, flow, and pressure, a data-driven dotted line sensor model is constructed; similarly, a data-driven approach is used to construct virtual object models;

[0102] (2) Based on the data acquisition board and terminal board and other equipment, the I / O module in the PLC / DCS control system is connected to the O / I terminal on the terminal board through a twisted pair cable, and then connected to the MSWI process virtual incineration object computer through Ethernet;

[0103] (3) The DI / DO (or AI / AO) in the actuator model computer and the DO / DI (or AO / AI) modules in the PLC / DCS control system respectively perform real-time data transmission.

[0104] (4) The AI / AO in the instrument model computer and the AO / AI modules in the PLC / DCS control system respectively perform real-time data transmission.

[0105] The multimodal historical data synchronization drive module, the multimodal data acquisition module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module and the difficult-to-measure process parameter soft measurement module together constitute a laboratory process parameter modeling algorithm simulation real-time verification subsystem. The multimodal data acquisition module obtains the synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module. The flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module and the difficult-to-measure process parameter soft measurement module realize the process parameter prediction, combustion state recognition, combustion line quantification and process parameter soft measurement results of the simultaneous release of real-time multimodal data from the industrial site. Each module of the laboratory process parameter modeling algorithm simulation real-time verification subsystem is introduced respectively:

[0106] Multimodal historical data synchronization drive module: Incinerator flame and process data are important bases for field experts to judge the stability of the MSWI process. Real-time collection and preprocessing of the above multimodal data are particularly important. The specific steps of multimodal data collection include:

[0107] (1) Using two identical sets of video cameras, the flame videos of the left and right grates are captured online in real time;

[0108] (2) Transmitting the camera collected information to the video capture card via coaxial cable;

[0109] (3) Install the video acquisition card into the multimodal data acquisition computer and realize the preliminary display of flame information through video decoding;

[0110] (4) Preprocessing the collected flame video by denoising and enhancing it through image processing algorithms;

[0111] (5) Collecting process data from the OPC server to the multimodal data acquisition computer via industrial Ethernet in the form of an OPC client;

[0112] (6) The production reports are collected into the multimodal data collection computer by manual input by domain experts, and the sampling period is reasonably set to realize the synchronous storage of multimodal data after collection.

[0113] Multimodal data acquisition module: Incinerator flame and process data are important bases for field experts to judge the stability of the MSWI process. Real-time acquisition and preprocessing of the above multimodal data are particularly important. The specific steps of multimodal data acquisition include:

[0114] (1) Using two identical sets of camera equipment, the flame videos of the left and right grates are collected online in real time;

[0115] (2) Transmitting the camera collected information to the video capture card via coaxial cable;

[0116] (3) Install the video acquisition card into the multimodal data acquisition computer and realize the preliminary display of flame information through video decoding;

[0117] (4) Using image recognition software in a computer, the collected flame video is preprocessed by denoising and enhancing;

[0118] (5) Collecting process data from the OPC server to a multimodal data acquisition computer via industrial Ethernet in the form of an OPC client;

[0119] (6) The production reports are collected into the multimodal data collection computer by manual input by domain experts, and the sampling period is reasonably set to realize the synchronous storage of multimodal data after collection.

[0120] Flame combustion line quantification module: The flame combustion line is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve quantification include:

[0121] (1) Using image processing technology, conditional generative adversarial networks, and cycle-consistent generative adversarial networks, a complete image library is constructed, which includes a real normal combustion line sub-library, a real / generated abnormal combustion line sub-library, and a generated extremely abnormal combustion line flame image sub-library;

[0122] (2) Select typical images from the complete image library to form a typical template library and train the twin convolutional neural network;

[0123] (3) Extract the burning line features from the new flame image and use the similarity metric of the twin convolutional neural network to achieve adaptation with the flame image of the “typical template library”;

[0124] (4) Non-adaptive new flame image, using the nearest neighbor criterion to achieve combustion line quantification;

[0125] (5) Based on the redundant discrimination mechanism and combined with the experience of domain experts, the “typical template library” is adaptively updated using non-adaptive images.

[0126] Visually driven combustion state recognition module: The flame combustion state is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve visually driven combustion state recognition include:

[0127] (1) The flame image is dehazed and denoised using a dehazing algorithm based on artificial multi-exposure image fusion, feature normalization, notch filtering, median filtering and other preprocessing methods to obtain a clear image;

[0128] (2) Extract multiple features with physical meanings such as brightness, texture, and color of flame images to represent the image from multiple viewpoints, and simplify these features based on mutual information;

[0129] (3) The above simplified features are used as inputs to image classifiers such as support vector machines, deep forests, and convolutional neural networks to establish visually driven combustion state recognition models for the left and right grates respectively;

[0130] (4) For each new flame image, the combustion state represented by it is recognized.

[0131] Multimodal data-driven process parameter prediction module: Achieving one-step or even multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration is crucial for domain experts. The specific steps for achieving multimodal data-driven process parameter prediction include:

[0132] (1) Using the multimodal data stored in the M2 module, the flame image features are extracted according to the M4 module;

[0133] (2) Combining flame characteristics and process data serially into new features for training key process parameter prediction models;

[0134] (3) The multimodal process data newly collected by the M2 module is input into the above-mentioned key process parameter prediction model to obtain one-step or even multi-step prediction outputs of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration.

[0135] Soft measurement module for difficult-to-measure process parameters: This module implements soft measurement modeling for difficult-to-measure parameters such as product quality parameter slag thermal loss rate and environmental indicator parameter dioxin. This is crucial for single-objective / multi-objective operation optimization of the MSWI process. The specific steps for achieving prediction of difficult-to-measure process parameters include:

[0136] (1) Organize the difficult-to-measure process parameter data recorded in various production reports as the output true value of the soft sensor model;

[0137] (2) Select the multimodal data stored in the M2 module based on the true value of the soft measurement output of the difficult-to-measure process parameter, that is, obtain the multimodal data time period corresponding to the true value of the soft measurement output;

[0138] (3) For the flame data within the time period corresponding to the true value of the soft measurement output, the flame image features are extracted according to the M4 module, and the features are serially combined with the process data to form new features for the soft measurement model input;

[0139] (4) Using the above simplified features as input to regression models such as support vector machines, deep forests, and deep neural networks to establish soft measurement models for difficult-to-detect process parameters;

[0140] (5) The multimodal process data newly collected by the M2 module is input into the above-mentioned soft measurement model of difficult-to-measure process parameters to obtain soft measurement values ​​of difficult-to-measure process parameters such as slag thermal loss rate and dioxin concentration.

[0141] The data forward acquisition isolation module, multimodal data acquisition module, flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module and difficult-to-measure process parameter soft measurement module together constitute the industrial field data acquisition and process parameter modeling subsystem, which transmits the industrial field data to the forward server via OPC, and then transmits it to the multimodal data acquisition module after physical isolation and forward acquisition. The flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module and difficult-to-measure process parameter soft measurement module realize the process parameter prediction, combustion state recognition, combustion line quantification and process parameter soft measurement results based on the synchronous release of real-time multimodal data of the industrial field. Each module of the industrial field data acquisition and process parameter modeling subsystem is introduced respectively:

[0142] Data forward acquisition isolation module: This module collects process data from the PLC / DCS control system network during the MSWI process through the standard OPC client provided by the industrial field DCS system manufacturer, and publishes the collected real-time data through the OPC communication protocol through physical isolation. The steps for implementing data forward acquisition isolation are as follows:

[0143] (1) Using the OPC server protocol of the PLC / DCS manufacturer, process data is collected from the on-site process monitoring system to the forward server and published externally in the form of an OPC server;

[0144] (2) Connect the forward server and the physically isolated forward acquisition machine to the same local area network through a switch;

[0145] (3) Using unidirectional optical fiber transmission to transmit the collected process data to the forward data analysis server in a physically isolated forward transmission mode, including process variable grouping, naming, and sampling time setting;

[0146] (4) The physically isolated forward transmission and forward data analysis servers are connected to the same local area network through a switch, and the forward data analysis server provides data services to the M2 module in the form of OPC services.

[0147] Multimodal data acquisition module: Incinerator flame and process data are important bases for field experts to judge the stability of the MSWI process. Real-time acquisition and preprocessing of the above multimodal data are particularly important. The specific steps of multimodal data acquisition include:

[0148] (1) Using two identical sets of camera equipment, the flame videos of the left and right grates are collected online in real time;

[0149] (2) Transmitting the camera collected information to the video capture card via coaxial cable;

[0150] (3) Install the video acquisition card into the multimodal data acquisition computer and realize the preliminary display of flame information through video decoding;

[0151] (4) Using image recognition software in a computer, the collected flame video is preprocessed by denoising and enhancing;

[0152] (5) Collecting process data from the OPC server to a multimodal data acquisition computer via industrial Ethernet in the form of an OPC client;

[0153] (6) The production reports are collected into the multimodal data collection computer by manual input by domain experts, and the sampling period is reasonably set to realize the synchronous storage of multimodal data after collection.

[0154] Flame combustion line quantification module: The flame combustion line is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve quantification include:

[0155] (1) Using image processing technology, conditional generative adversarial networks, and cycle-consistent generative adversarial networks, a complete image library is constructed, which includes a real normal combustion line sub-library, a real / generated abnormal combustion line sub-library, and a generated extremely abnormal combustion line flame image sub-library;

[0156] (2) Select typical images from the complete image library to form a typical template library and train the twin convolutional neural network;

[0157] (3) Extract the burning line features from the new flame image and use the similarity metric of the twin convolutional neural network to achieve adaptation with the flame image of the “typical template library”;

[0158] (4) Non-adaptive new flame image, using the nearest neighbor criterion to achieve combustion line quantification;

[0159] (5) Based on the redundant discrimination mechanism and combined with the experience of domain experts, the “typical template library” is adaptively updated using non-adaptive images.

[0160] Visually driven combustion state recognition module: The flame combustion state is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve visually driven combustion state recognition include:

[0161] (1) The flame image is dehazed and denoised using a dehazing algorithm based on artificial multi-exposure image fusion, feature normalization, notch filtering, median filtering and other preprocessing methods to obtain a clear image;

[0162] (2) Extract multiple features with physical meanings such as brightness, texture, and color of flame images to represent the image from multiple viewpoints, and simplify these features based on mutual information;

[0163] (3) The above simplified features are used as inputs to image classifiers such as support vector machines, deep forests, and convolutional neural networks to establish visually driven combustion state recognition models for the left and right grates respectively;

[0164] (4) For each new flame image, the combustion state represented by it is recognized.

[0165] Multimodal data-driven process parameter prediction module: Achieving one-step or even multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration is crucial for domain experts. The specific steps for achieving multimodal data-driven process parameter prediction include:

[0166] (1) Using the multimodal data stored in the M2 module, the flame image features are extracted according to the M4 module;

[0167] (2) Combining flame characteristics and process data serially into new features for training key process parameter prediction models;

[0168] (3) The multimodal process data newly collected by the M2 module is input into the above-mentioned key process parameter prediction model to obtain one-step or even multi-step prediction outputs of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration.

[0169] Soft measurement module for difficult-to-measure process parameters: This module implements soft measurement modeling for difficult-to-measure parameters such as product quality parameter slag thermal loss rate and environmental indicator parameter dioxin. This is crucial for single-objective / multi-objective operation optimization of the MSWI process. The specific steps for achieving prediction of difficult-to-measure process parameters include:

[0170] (1) Organize the difficult-to-measure process parameter data recorded in various production reports as the output true value of the soft sensor model;

[0171] (2) Select the multimodal data stored in the M2 module based on the true value of the soft measurement output of the difficult-to-measure process parameter, that is, obtain the multimodal data time period corresponding to the true value of the soft measurement output;

[0172] (3) For the flame data within the time period corresponding to the true value of the soft measurement output, the flame image features are extracted according to the M4 module, and the features are serially combined with the process data to form new features for the soft measurement model input;

[0173] (4) Using the above simplified features as input to regression models such as support vector machines, deep forests, and deep neural networks to establish soft measurement models for difficult-to-detect process parameters;

[0174] (5) The multimodal process data newly collected by the M2 module is input into the above-mentioned soft measurement model of difficult-to-measure process parameters to obtain soft measurement values ​​of difficult-to-measure process parameters such as slag thermal loss rate and dioxin concentration.

[0175] The data forward acquisition isolation module, multimodal data acquisition module, flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module, difficult-to-measure process parameter soft measurement module, MSWI process single-objective / multi-objective operation optimization module, operation parameter reverse transmission module and operation parameter auxiliary decision module together constitute an industrial field auxiliary decision and operation optimization subsystem. Process data is transmitted to the forward server of the data acquisition forward isolation module through OPC, and is transmitted to the multimodal data acquisition module after physical isolation and forward acquisition. After processing by the flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module, difficult-to-measure process parameter soft measurement module and MSWI process single-objective / multi-objective operation optimization module, multiple categories of key operation parameter prediction values ​​and soft measurement values ​​as well as optimized operation parameter values ​​are obtained, and then transmitted to the reverse server of the operation parameter reverse transmission module. After the data is transmitted to the server, it is physically isolated and transmitted back to the operating parameter reverse receiving server. After the operating parameter auxiliary decision-making module performs auxiliary decision-making analysis, the optimized operating parameters are transmitted to the on-site monitoring system and PLC / DCS system using the OPC protocol or OCR recognition method according to the safety requirements of the MSWI plant. The control quantity is then transmitted to the actual MSWI process actuator in the form of analog output, acting on the actual objects consisting of solid waste storage and transportation, solid waste combustion, waste heat exchange, flue gas purification and flue gas emission stages. The data is then collected through analog input by the instrument device to the PLC / DCS system and on-site monitoring system, and then transmitted to the data acquisition forward isolation module and multimodal data acquisition module through OPC, and then fed back to the MSWI process single-objective / multi-objective operation optimization module, completing the implementation of the operation optimization algorithm for the actual industrial site. The functions of each module of the industrial site auxiliary decision-making and operation optimization subsystems are introduced respectively:

[0176] Data forward acquisition isolation module: This module collects process data from the PLC / DCS control system network during the MSWI process through the standard OPC client provided by the industrial field DCS system manufacturer, and publishes the collected real-time data through the OPC communication protocol through physical isolation. The steps for implementing data forward acquisition isolation are as follows:

[0177] (1) Using the OPC server protocol of the PLC / DCS manufacturer, process data is collected from the on-site process monitoring system to the forward server and published externally in the form of an OPC server;

[0178] (2) Connect the forward server and the physically isolated forward acquisition machine to the same local area network through a switch;

[0179] (3) Using unidirectional optical fiber transmission to transmit the collected process data to the forward data analysis server in a physically isolated forward transmission mode, including process variable grouping, naming, and sampling time setting;

[0180] (4) The physically isolated forward transmission and forward data analysis servers are connected to the same local area network through a switch, and the forward data analysis server provides data services to the M2 module in the form of OPC services.

[0181] Multimodal data acquisition module: Incinerator flame and process data are important bases for field experts to judge the stability of the MSWI process. Real-time acquisition and preprocessing of the above multimodal data are particularly important. The specific steps of multimodal data acquisition include:

[0182] (1) Using two identical sets of camera equipment to respectively capture the left and right flame videos of the industrial site in real time;

[0183] (2) Transmitting the camera collected information to the video capture card via coaxial cable;

[0184] (3) Install the video acquisition card into the multimodal data acquisition computer and realize the preliminary display of flame information through video decoding;

[0185] (4) Using image recognition software in a computer, the collected flame video is preprocessed by denoising and enhancing;

[0186] (5) Collect process data from the OPC server to the acquisition computer of the M2 module via industrial Ethernet in the form of an OPC client;

[0187] (6) The production reports are collected into the multimodal data collection computer by manual input by domain experts, and the sampling period is reasonably set to realize the synchronous storage of multimodal data after collection.

[0188] Flame combustion line quantification module: The flame combustion line is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve quantification include:

[0189] (1) Using image processing technology, conditional generative adversarial networks, and cycle-consistent generative adversarial networks, a complete image library is constructed, which includes a real normal combustion line sub-library, a real / generated abnormal combustion line sub-library, and a generated extremely abnormal combustion line flame image sub-library;

[0190] (2) Select typical images from the complete image library to form a typical template library and train the twin convolutional neural network;

[0191] (3) Extract the burning line features from the new flame image and use the similarity metric of the twin convolutional neural network to achieve adaptation with the flame image of the “typical template library”;

[0192] (4) Non-adaptive new flame image, using the nearest neighbor criterion to achieve combustion line quantification;

[0193] (5) Based on the redundant discrimination mechanism and combined with the experience of domain experts, the “typical template library” is adaptively updated using non-adaptive images.

[0194] Visually driven combustion state recognition module: The flame combustion state is the controlled variable required to achieve stable control of the MSWI process. The specific steps to achieve visually driven combustion state recognition include:

[0195] (1) The flame image is dehazed and denoised using a dehazing algorithm based on artificial multi-exposure image fusion, feature normalization, notch filtering, median filtering and other preprocessing methods to obtain a clear image;

[0196] (2) Extract multiple features with physical meanings such as brightness, texture, and color of flame images to represent the image from multiple viewpoints, and simplify these features based on mutual information;

[0197] (3) The above simplified features are used as inputs to image classifiers such as support vector machines, deep forests, and convolutional neural networks to establish visually driven combustion state recognition models for the left and right grates respectively;

[0198] (4) For each new flame image, the combustion state represented by it is recognized.

[0199] Multimodal data-driven process parameter prediction module: Achieving one-step or even multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration is crucial for domain experts. The specific steps for achieving multimodal data-driven process parameter prediction include:

[0200] (1) Using the multimodal data stored in the M2 module, the flame image features are extracted according to the M4 module;

[0201] (2) Combining flame characteristics and process data serially into new features for training key process parameter prediction models;

[0202] (3) The multimodal process data newly collected by the M2 module is input into the above-mentioned key process parameter prediction model to obtain one-step or even multi-step prediction outputs of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, CO emission concentration, and NOx emission concentration.

[0203] Soft measurement module for difficult-to-measure process parameters: This module implements soft measurement modeling for difficult-to-measure parameters such as product quality parameter slag thermal loss rate and environmental indicator parameter dioxin. This is crucial for single-objective / multi-objective operation optimization of the MSWI process. The specific steps for achieving prediction of difficult-to-measure process parameters include:

[0204] (1) Organize the difficult-to-measure process parameter data recorded in various production reports as the output true value of the soft sensor model;

[0205] (2) Select the multimodal data stored in the M2 module based on the true value of the soft measurement output of the difficult-to-measure process parameter, that is, obtain the multimodal data time period corresponding to the true value of the soft measurement output;

[0206] (3) For the flame data within the time period corresponding to the true value of the soft measurement output, the flame image features are extracted according to the M4 module, and the features are serially combined with the process data to form new features for the soft measurement model input;

[0207] (4) Using the above simplified features as input to regression models such as support vector machines, deep forests, and deep neural networks to establish soft measurement models for difficult-to-detect process parameters;

[0208] (5) The multimodal process data newly collected by the M2 module is input into the above-mentioned soft measurement model of difficult-to-measure process parameters to obtain soft measurement values ​​of difficult-to-measure process parameters such as slag thermal loss rate and dioxin concentration.

[0209] The MSWI process single-objective / multi-objective operation optimization module: In different scenarios, considering single or multiple objectives such as flue gas emission indicators, economic benefits, and slag heat reduction rate, the set values ​​of key controlled variables such as furnace temperature, flue gas oxygen content, and steam flow rate are adjusted in a timely manner as the MSWI process changes dynamically. This is very important for achieving optimal control of the MSWI process. The specific steps for achieving single-objective / multi-objective operation optimization of the MSWI process include:

[0210] (1) Based on the MSWI process analysis for single-objective / multi-objective optimization, a single-objective / multi-objective optimization model was established with key controlled variables such as furnace temperature, boiler steam flow rate, and flue gas oxygen content as decision variables to minimize pollutant emission indicators and heat reduction rate and maximize combustion efficiency and economic indicators;

[0211] (2) Based on the above-mentioned single / multi-objective optimization model, the multimodal data collected in real time by the M2 module and the soft measurement model of difficult-to-measure process parameters of the M6 ​​module are used to obtain the optimized set values ​​of key controlled variables using intelligent optimization algorithms such as genetic algorithm, particle swarm optimization algorithm, and differential evolution algorithm.

[0212] Operation parameter reverse transmission module: obtains the single-step / multi-step prediction values ​​and soft measurement values ​​of the key operation parameter model and the optimized operation parameter values ​​through the reverse server, transmits the above data through physical isolation, and then uses the reverse data analysis server to analyze the above data in the form of OPC. The implementation steps of the M7-operation parameter reverse transmission module are as follows:

[0213] (1) Collecting and storing the single-step / multi-step prediction values ​​and soft measurement values ​​of the key operating parameter model and the optimized operating parameter values ​​through the reverse server computer;

[0214] (2) Connecting the reverse server and the physically isolated reverse acquisition machine to the same local area network through a switch, and then transmitting data to the physically isolated reverse acquisition machine;

[0215] (3) Using unidirectional optical fiber transmission to transmit the collected process data to the reverse data analysis server in a physically isolated reverse transmission mode, including process variable grouping, naming, and sampling time setting;

[0216] (4) The physically isolated reverse transmission and reverse data analysis servers are connected to the same local area network through a switch, and the reverse data analysis server provides data services to the M9 module using OPC services.

[0217] Operating parameter decision-making support module: The single-step / multi-step prediction values ​​and soft measurement values ​​of key operating parameter models are crucial for domain experts to execute control strategies and also determine whether to apply optimized operating parameters to the control of the MSWI process. The steps to implement operating parameter decision-making support are as follows:

[0218] (1) Collect the single-step / multi-step prediction values ​​and soft measurement values ​​of the key operating parameter model from the M8 module to the operating parameter reverse receiving server;

[0219] (2) Based on the single-step / multi-step prediction values ​​and soft measurement values ​​of the key operating parameter models in different time periods, a statistically significant comparative analysis is conducted with the actual values ​​of these key operating parameters. Active learning by domain experts and automatic discrimination by setting thresholds are used to assist in the decision-making of the credibility of the single-step / multi-step prediction values ​​and soft measurement values ​​of these key operating parameter models.

[0220] (3) Based on the credibility of the single-step / multi-step prediction values ​​and soft measurement values ​​of the above-mentioned key operating parameter models and the optimized operating parameters, two methods are adopted for data transmission: when the industrial site fully trusts the isolation device of the system, the optimized operating parameters decided by domain experts or automatically are transmitted to the on-site process control system; when the industrial site requires a higher level of security issues, the operating parameters are optimized based on the operating parameter OCR recognition technology and transmitted to the on-site process control system.

[0221] The multi-purpose modular operation optimization system for the municipal solid waste incineration process provided by the present invention includes a multi-modal historical data synchronization drive module, an MSWI process virtual control object module, an MSWI process loop control module, an MSWI process monitoring module, an operation parameter auxiliary decision module, a data acquisition forward isolation module, an operation parameter reverse transmission module, an MSWI process single-objective / multi-objective operation optimization module, a difficult-to-measure process parameter soft measurement module, a multi-modal data-driven process parameter prediction module, a visually driven combustion state recognition module, a flame combustion line quantification module and a multi-modal data acquisition module. The system can be applied to both laboratories and industrial sites, avoiding the traditional simulation and simulation of the system. The shortcomings of real experimental systems that are too complex or too simple can be overcome by modular construction based on specific needs, and can also effectively meet the needs of isolated data collection and software transplantation in industrial sites. The system can effectively integrate multimodal data synchronous prediction and operation optimization control. The system takes into account the security requirements of different MSWI companies and can choose OPC server or OCR recognition method to transmit operation parameters. It has strong practicality and adaptability. The system can realize laboratory operation optimization algorithm verification, laboratory process parameter modeling algorithm simulation real-time verification, industrial site data collection and process parameter modeling, industrial site decision support and operation optimization and other functions, and is easy to use.

[0222] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A multi-purpose modular operation optimization system for municipal solid waste incineration process, characterized in that: include: A multimodal historical data synchronization drive module, an MSWI process virtual control object module, an MSWI process loop control module, an MSWI process monitoring module, an operation parameter auxiliary decision module, a data acquisition forward isolation module, an operation parameter reverse transmission module, an MSWI process single-objective / multi-objective operation optimization module, a difficult-to-measure process parameter soft measurement module, a multimodal data-driven process parameter prediction module, a visually driven combustion state recognition module, a flame combustion line quantification module and a multimodal data acquisition module. The multimodal historical data synchronization drive module is connected to the multimodal data acquisition module, and the multimodal data acquisition module is connected to the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data-driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the MSWI process single-objective / multi-objective operation optimization module. The flame combustion line quantification module, the visually driven The combustion state identification module, the multimodal data-driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module and the MSWI process single-objective / multi-objective operation optimization module are connected to the operation parameter auxiliary decision module and the operation parameter reverse transmission module, the operation parameter reverse transmission module is connected to the operation parameter auxiliary decision module, the operation parameter auxiliary decision module is connected to the MSWI process monitoring module and the field process monitoring system, the field process monitoring system is connected to the data acquisition forward isolation module, the data acquisition forward isolation module is connected to the multimodal data acquisition module, the MSWI process monitoring module is connected to the MSWI process loop control module, the MSWI process loop control module is connected to the MSWI process virtual control object module, the field process monitoring system is connected to the field loop control system, and the field loop control system is connected to the actuator and instrumentation device; The multimodal historical data synchronization driving module is used to provide a multimodal data source for the MSWI process; The MSWI process virtual control object module is used to realize the simulation of the MSWI process built in the laboratory; The MSWI process loop control module is used to implement loop control of the virtual MSWI process; The MSWI process monitoring module is used to monitor the virtual MSWI process; The operating parameter auxiliary decision module is used to obtain operating parameters and conduct comparative analysis and decision making on them; The data acquisition forward isolation module is used to realize data acquisition of all process variables in the monitoring module of the virtual MSWI process by means of physical isolation; The operating parameter reverse transmission module is used to reversely transmit the operating parameter optimization values ​​obtained from the MSWI process single-objective / multi-objective operation optimization module, the operating parameters of the flame combustion line quantification module, the visual-driven combustion state recognition module, the multi-modal data-driven process parameter prediction module, and the operating parameter detection values ​​of the difficult-to-measure process parameter soft measurement module in a physically isolated manner; The MSWI process single-objective / multi-objective operation optimization module is used to optimize the MSWI process operation parameters based on multi-modal data and hard-to-measure parameter soft sensor models; The difficult-to-measure process parameter soft measurement module is used to implement soft measurement modeling of difficult-to-detect parameters based on multimodal data and production reports; The multimodal data driven process parameter prediction module is used to realize single-step / multi-step prediction of process parameters such as furnace temperature, flue gas oxygen content, steam flow rate, etc. based on multimodal data; The visually driven combustion state recognition module is used to realize the recognition of the combustion state in the furnace by imitating the domain expert recognition mechanism for the MSWI process; The flame burning line quantification module is used to realize the quantification of the flame burning line based on the domain expert recognition mechanism for the MSWI process; The multimodal data acquisition module is used to realize the collection of left grate flame video, right grate flame video and historical process data of simulated and actual MSWI processes with multimodal data, as well as the processing and entry of various production reports related to product quality, environmental protection indicators and economic indicators generated by the actual process.

2. The multi-purpose modular operation optimization system for municipal solid waste incineration process according to claim 1 is characterized in that: The multimodal historical data synchronization drive module, the multimodal data acquisition module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module, the MSWI process single-objective / multi-objective operation optimization module, the operation parameter auxiliary decision module, the MSWI process monitoring module, the MSWI process loop control module and the MSWI process virtual control object module together constitute a laboratory operation optimization algorithm verification subsystem, which obtains the synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module through the multimodal data acquisition module, and obtains the synchronized left grate and right grate flame image data and historical process data through the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module, the difficult-to-measure process parameter soft measurement module, the MSWI process single-objective / multi-objective operation optimization module, the operation parameter auxiliary decision module, the MSWI process monitoring module, the MSWI process loop control module and the MSWI process virtual control object module The process parameter soft measurement module and the difficult-to-measure process parameter soft measurement module are processed to obtain various categories of operating parameters, which are transmitted to the MSWI process monitoring module through OPC and downloaded to the MSWI process loop control module based on real PLC / DCS equipment. The control quantity is then transmitted to the virtual actuator of the MSWI process virtual control object module in the form of analog output. The output of the virtual actuator acts on the virtual object to generate a controlled variable output, which is transmitted to the MSWI process loop control module in the form of analog input through the virtual instrument device, and then transmitted to the MSWI process monitoring module, and then transmitted to the multimodal data acquisition module through OPC, and then fed back to the difficult-to-measure process parameter soft measurement module to complete the laboratory-oriented operation optimization algorithm verification.

3. The multi-purpose modular operation optimization system for municipal solid waste incineration process according to claim 1 is characterized in that: The multimodal historical data synchronization drive module, the multimodal data acquisition module, the flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module and the difficult-to-measure process parameter soft measurement module together constitute a laboratory process parameter modeling algorithm simulation real-time verification subsystem. The synchronized left grate and right grate flame image data and historical process data from the multimodal historical data synchronization drive module are obtained through the multimodal data acquisition module. The flame combustion line quantification module, the visually driven combustion state recognition module, the multimodal data driven process parameter prediction module and the difficult-to-measure process parameter soft measurement module are used to realize the process parameter prediction, combustion state recognition, combustion line quantification and process parameter soft measurement results simulated by the real-time multimodal data synchronization released in the industrial field.

4. The multi-purpose modular operation optimization system for municipal solid waste incineration process according to claim 1 is characterized in that: The data forward acquisition isolation module, multimodal data acquisition module, flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module and difficult-to-measure process parameter soft measurement module together constitute an industrial field data acquisition and process parameter modeling subsystem, which transmits the industrial field data to the forward server via OPC, and then transmits it to the multimodal data acquisition module after physical isolation forward acquisition, and realizes process parameter prediction, combustion state recognition, combustion line quantification and process parameter soft measurement results based on the synchronous release of real-time multimodal data of the industrial field through the flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module and difficult-to-measure process parameter soft measurement module.

5. The multi-purpose modular operation optimization system for municipal solid waste incineration process according to claim 1 is characterized in that: The data forward acquisition isolation module, multimodal data acquisition module, flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module, difficult-to-measure process parameter soft measurement module, MSWI process single-objective / multi-objective operation optimization module, operation parameter reverse transmission module and operation parameter auxiliary decision module together constitute an industrial field auxiliary decision and operation optimization subsystem. The process data is transmitted to the forward server of the data acquisition forward isolation module through OPC, and is transmitted to the multimodal data acquisition module after physical isolation and forward acquisition. After being processed by the flame combustion line quantification module, visually driven combustion state recognition module, multimodal data driven process parameter prediction module, difficult-to-measure process parameter soft measurement module and MSWI process single-objective / multi-objective operation optimization module, various categories of key operation parameter prediction values ​​and soft measurement values ​​as well as optimized operation parameter values ​​are obtained. After being transmitted to the reverse server of the operating parameter reverse transmission module, it is reversely transmitted to the operating parameter reverse receiving server through physical isolation. After the operating parameter auxiliary decision-making module performs auxiliary decision-making analysis, the optimized operating parameters are transmitted to the on-site monitoring system and PLC / DCS system using the OPC protocol or OCR recognition method according to the safety requirements of the MSWI plant. The control quantity is then transmitted to the actual MSWI process actuator in the form of analog output, acting on the actual object composed of solid waste storage and transportation, solid waste combustion, waste heat exchange, flue gas purification and flue gas emission stages, and then collected to the PLC / DCS system and the on-site monitoring system through the instrument device through analog input, and then transmitted to the data acquisition forward isolation module and the multimodal data acquisition module through OPC, and then fed back to the MSWI process single-objective / multi-objective operation optimization module to complete the implementation of the operation optimization algorithm for the actual industrial site.

Images