Two-phase flow drying control method and system self-adaptive to boiler load change

Abstract

The invention discloses a two-phase flow drying control method and system adaptive to boiler load change, and the method comprises the following steps: collecting boiler load parameters and two-phase flow drying system operation parameters in real time, and constructing an input vector; inputting the input vector to a least square support vector machine prediction model, and outputting an optimal set value adaptive to the current boiler load by establishing a nonlinear mapping relationship between the boiler working condition and the optimal control parameter; according to a preset process security constraint, performing boundary verification and correction on the optimal set value to generate a final control instruction; and the final control instruction is controlled and executed, two-phase flow parameters are adjusted, closed-loop control is formed based on system output feedback, and self-adaptive optimization of the two-phase flow drying process is achieved. According to the method, the least square support vector machine is adopted as a core algorithm, so that the adaptability, the prediction precision and the automation degree of the system are improved, and stable and efficient operation of the two-phase flow drying process under various working conditions is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of desulfurization wastewater zero discharge automation, and particularly relates to a two-phase flow drying control method and system self-adapting to boiler load changes. BACKGROUND

[0002] In the operation process of a coal-fired power plant boiler, the waste heat of boiler flue gas is often used to dry desulfurization wastewater through a two-phase flow drying system. Two-phase flow drying refers to a technology that uses liquid and compressed gas to realize liquid droplet atomization. The typical device is a two-fluid nozzle: liquid is mixed with high-speed gas through the nozzle to form fine liquid droplets, thereby increasing the liquid-gas contact area and using the flue gas heat to rapidly evaporate the liquid droplets to complete the drying process. Compared with a single-fluid (pressure type) nozzle, the two-fluid nozzle reduces the liquid supply pressure by introducing compressed air and generates finer atomized droplets. The fine droplets are beneficial to intensify heat and mass transfer and improve the evaporation rate.

[0003] However, the load of a coal-fired boiler often changes (load reduction means changes in flue gas temperature, flow rate, and other conditions), which directly affects the effect of two-phase flow drying. When the boiler load decreases, the flue gas temperature and flow rate decrease, and the sprayed liquid droplets may not be completely evaporated, resulting in reduced drying efficiency and even the risk of unevaporated liquid droplets adhering to the pipe wall and causing corrosion. When the load increases, if not properly controlled, too much moisture may be sprayed, causing a sharp drop in flue gas temperature or wasting energy. Therefore, how to adaptively adjust the spray parameters according to the change in boiler load to maintain the stability and efficiency of the atomization drying process has become a technical problem to be solved.

[0004] Traditional two-phase flow drying systems often use fixed parameters or simple proportional control methods, which cannot fully adapt to the complex nonlinear heat and mass transfer process and the dynamic changes in load. This easily leads to conservative setting of operating parameters, low drying efficiency, or frequent manual intervention. With the development of machine learning and intelligent control technology, a data-driven control model can be introduced to realize intelligent optimization control of the two-phase flow drying system, so that it can maintain the best operating state under different boiler loads. SUMMARY

[0005] The present application provides a two-phase flow drying control method self-adapting to changes in boiler load, which solves the problem that the two-phase flow drying process is difficult to maintain optimal process parameters when the boiler load changes, thereby reducing drying efficiency, increasing energy consumption, causing equipment corrosion, and making the system unstable.

[0006] In a first aspect, the present application provides a two-phase flow drying control method self-adapting to changes in boiler load, comprising the following steps:

[0007] Real-time acquisition of boiler load parameters and two-phase flow drying system operating parameters to construct an input vector;

[0008] inputting the input vector into a pre-trained least squares support vector machine prediction model, the least squares support vector machine prediction model directly outputting an optimal set value suitable for a current boiler load by establishing a nonlinear mapping relationship between a boiler working condition and the optimal control parameter;

[0009] performing boundary checking and correction on the optimal set value according to a preset process safety constraint to generate a final control instruction;

[0010] controlling the final control instruction to be executed to adjust two-phase flow parameters and form a closed-loop control based on system output feedback, thereby realizing adaptive optimization of a two-phase flow drying process.

[0011] A further optimization scheme is that the construction of the least squares support vector machine prediction model is based on a structural risk minimization principle, and an optimization problem thereof is defined as:

[0012] ;

[0013] An equality constraint thereof is:

[0014] ;

[0015] In the formula, C is a regularization parameter for balancing model complexity and training error. is a training error of the ith sample.

[0016] A further optimization scheme is that a regression function of the least squares support vector machine prediction model is:

[0017] ;

[0018] In the formula, is a kernel function satisfying a Mercer condition.

[0019] A further optimization scheme is that the kernel function is a radial basis function, and an expression thereof is:

[0020] ;

[0021] In the formula, sigma is a kernel function parameter for controlling a radial action range of the function.

[0022] A further optimization scheme is that the regularization parameter and the kernel function parameter are determined by a particle swarm optimization algorithm to adapt to a fitness function minimized as an objective, wherein is a root mean square error of K-fold cross-validation.

[0023] A further optimization scheme is that the least squares support vector machine prediction input vector includes at least: boiler load rate, flue gas inlet temperature, flue gas flow rate, drying tower outlet temperature, drying tower outlet humidity, spray water volume of the previous control cycle, and gas-liquid ratio of the previous control cycle.

[0024] Secondly, this application provides an adaptive two-phase flow drying control system for boiler load changes, used to implement the method described above, including:

[0025] The sensor module is used to collect boiler load parameters and operating parameters of the two-phase flow drying system in real time to construct the input vector;

[0026] The controller module integrates the least squares support vector machine prediction model and the rule correction unit, and is communicatively connected to the sensor module. It is used to predict the optimal control setpoint based on the input parameters and complete the safety correction.

[0027] The actuator module is communicatively connected to the controller module and is used to control the execution of the final control command, adjust the two-phase flow parameters, and form a closed-loop control based on system output feedback to achieve adaptive optimization of the two-phase flow drying process.

[0028] The sensor module, controller module, and actuator module together form a closed-loop control circuit.

[0029] A further optimization scheme is that the controller module is a programmable logic controller or an embedded industrial control computer, which internally stores the parameter set and kernel function parameters of the least squares support vector machine model obtained through offline training.

[0030] A further optimized solution is that the actuator module includes a variable frequency pump or regulating valve for adjusting the wastewater flow rate, and a compressed air regulating valve or variable frequency compressor for adjusting the pressure and flow rate of atomized air.

[0031] A further optimization scheme is that the system is also equipped with a data buffer unit to store operating conditions and control command data, so as to support the periodic online updates and iterative optimization of the least squares support vector machine prediction model.

[0032] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0033] This application uses least squares support vector machine as the core algorithm to improve the system's adaptability, prediction accuracy and automation, and ensure the stable and efficient operation of the two-phase flow drying process under various working conditions. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0035] Figure 1 A flowchart of the two-phase flow drying control method for adaptive boiler load changes provided in the embodiments of this application;

[0036] Figure 2 A functional block diagram of the adaptive boiler load change two-phase flow drying control system provided in the embodiments of this application. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0038] First, some of the technical terms used in this application will be explained to help those skilled in the art understand this application.

[0039] PSO: Particle Swarm Optimization;

[0040] LS-SVM: Least Squares Support Vector Machine;

[0041] PLC: Programmable Logic Controller;

[0042] DCS: Distributed Control System;

[0043] RBF: Radial Basis Function;

[0044] MAPE: Mean Absolute Percentage Error;

[0045] RMSE: Root Mean Square Error (hereinafter referred to as RMSE) (Appearing in form);

[0046] KKT: Karush-Kuhn-Tucker, Karush-Kuhn-Tucker condition.

[0047] In a first aspect, this application provides a two-phase flow drying control method that adapts to changes in boiler load, comprising the following steps:

[0048] Step S1: Real-time acquisition of boiler load parameters and two-phase flow drying system operating parameters to construct an input vector;

[0049] Step S2: Input the input vector into the pre-trained least squares support vector machine prediction model. The least squares support vector machine prediction model directly outputs the optimal setpoint adapted to the current boiler load by establishing a nonlinear mapping relationship between the boiler operating conditions and the optimal control parameters.

[0050] Step S3: Based on the preset process safety constraints, perform boundary verification and correction on the optimal set value to generate the final control command;

[0051] Step S4: The system executes the final control command, adjusts the two-phase flow parameters, and forms a closed-loop control based on the system output feedback to achieve adaptive optimization of the two-phase flow drying process.

[0052] This implementation uses a least squares support vector machine prediction model to effectively handle the complex nonlinear relationship between boiler operating conditions and optimal control parameters. Compared with traditional PID control, this model has stronger robustness to noise data and lower computational complexity. It can quickly output the optimal setpoint adapted to the current load, thereby improving the accuracy and response speed of control.

[0053] By constructing input vectors through real-time data acquisition and combining them with rule correction units for boundary verification and correction, the system's adaptability and operational safety under variable load conditions are effectively enhanced, ensuring that control commands always meet process constraints.

[0054] The closed-loop control based on system output feedback enables continuous optimization of the two-phase flow drying process, which helps stabilize product quality and effectively overcomes the effects of nonlinear time delays such as boiler load fluctuations.

[0055] In one embodiment, the construction process of the least squares support vector machine prediction model in step S2 is as follows:

[0056] First, operational data covering different load conditions was collected from the power plant's distributed control system and historical database, including key variables such as boiler load rate, flue gas temperature and flow rate, dryer tower outlet status, and historical spray operation parameters. The raw data was then cleaned and normalized to eliminate the influence of dimensions and improve the stability of model training, resulting in a pre-processed training sample set. ,in, For input data, ; To output data, ; The number of training samples;

[0057] Based on the training sample set, the goal of the LS-SVM regression model is to construct a function that maps input data to output, as shown in the following equation:

[0058]

[0059] in: b is the weight vector; b is the bias value. A nonlinear mapping function that maps input data to a high-dimensional feature space;

[0060] The training process of the transformation model involves solving an optimization problem with equality constraints, defined as follows:

[0061] ;

[0062] Its constraints are:

[0063] ;

[0064] In the formula: C is the regularization parameter, used to balance model complexity and training error; Let be the training error of the i-th sample.

[0065] Introducing Lagrange multipliers Based on the Karush-Kuhn-Tucker (KKT) conditions, the above optimization problem is transformed into a system of linear equations to be solved, yielding the regression function expression for the least squares support vector machine prediction model as follows:

[0066] ;

[0067] in: To satisfy the Mercer condition, a radial basis function is used, with the following expression:

[0068] ;

[0069] In the formula: σ is the kernel function parameter, which controls the radial range of the function.

[0070] In the training process of the least squares support vector machine prediction model, the key hyperparameters of the model, namely the regularization parameter C and the kernel function parameter σ, have a decisive impact on the prediction performance. To obtain the optimal model performance, this invention employs the particle swarm optimization (PSO) algorithm to automatically optimize the parameter combination. Specifically, the PSO algorithm uses the fitness function... Minimize as the objective, where To minimize the root mean square error (RMSE) of K-fold cross-validation, an iterative search is performed to find the parameter combination that minimizes this metric, ensuring the trained model possesses optimal generalization ability and prediction accuracy. Ultimately, the optimized model accurately captures the complex nonlinear relationship between the input real-time operating parameters and the optimal spray control quantity. After thorough validation on both the training and test sets, and achieving high goodness of fit and low prediction error, the model can be deployed to a real-world control system for operation. Furthermore, the system establishes a regular update mechanism, incorporating new operational data into the model for incremental learning, enabling it to continuously adapt to dynamic changes in operating conditions.

[0071] This training method transforms complex optimization problems into linear system solutions, ensuring both the uniqueness and global optimality of the solution, and significantly improving training efficiency. This enables the model to effectively learn the nonlinear mapping relationship between boiler operating conditions and spray parameters from the data.

[0072] The constructed least squares support vector machine prediction model has the following effects:

[0073] (1) LS-SVM is based on the principle of minimizing structural risk, has better generalization ability, can effectively avoid overfitting, and has stronger adaptability to complex working conditions such as boiler load changes.

[0074] (2)LS-SVM solves a convex optimization problem, which guarantees the global optimal solution and avoids the problem of getting trapped in local minima during neural network training, making the control model more stable and reliable.

[0075] (3)LS-SVM has unique advantages in handling small sample and nonlinear problems. It can also establish high-precision prediction models for industrial scenarios where it is difficult to obtain a large amount of effective operational data.

[0076] (4) The least squares support vector machine prediction model has fewer hyperparameters (only regularization parameter C and kernel function parameter σ), which reduces the complexity of model selection and optimization, and facilitates engineering implementation and maintenance.

[0077] In one embodiment, the least squares support vector machine prediction input vector in step S2 includes at least the boiler load rate, flue gas inlet temperature, flue gas flow rate, drying tower outlet temperature, drying tower outlet humidity, spray water volume of the previous control cycle, and gas-liquid ratio of the previous control cycle.

[0078] In one embodiment, step S3 is specifically implemented as follows:

[0079] The rule correction unit receives the model predictions and performs verification and correction based on preset process safety constraints; the verification content includes:

[0080] 1) Ensure that the set values ​​are within the operating range of the spray water pump frequency converter, the valve opening (0-100%), and the upper and lower limits of the compressed air pressure;

[0081] 2) The core constraint for process safety is that the outlet temperature of the drying tower must not be lower than 180℃ (the safety threshold is set at 120℃).

[0082] If the predicted value exceeds the limit, a correction strategy is applied to adjust it; specifically, values ​​exceeding the equipment boundary are limited; when the predicted spray water volume may cause the outlet temperature to be lower than 120°C, the water volume is forcibly reduced according to the heat balance model.

[0083] After calibration, the final control commands are generated to ensure the safe operation of the system.

[0084] Secondly, such as Figure 2 As shown, this application provides an adaptive two-phase flow drying control system for boiler load changes, used to implement the above-described method, comprising:

[0085] The sensor module 100 is used to collect boiler load parameters and operating parameters of the two-phase flow drying system in real time to construct the input vector;

[0086] The controller module 200 integrates the least squares support vector machine prediction model and the rule correction unit, and is communicatively connected to the sensor module 100. It is used to predict the optimal control setpoint based on the input parameters and complete the safety correction.

[0087] The actuator module 300 is communicatively connected to the controller module 200 and is used to control the execution of the final control command, adjust the two-phase flow parameters, and form a closed-loop control based on system output feedback to achieve adaptive optimization of the two-phase flow drying process.

[0088] The sensor module, controller module, and actuator module together form a closed-loop control circuit.

[0089] In a more specific embodiment, sensor modules are arranged at the boiler and drying tower flue gas duct locations to monitor boiler load and two-phase flow drying-related parameters in real time. The flue gas inlet temperature is obtained via thermocouples and flow meters. and flue gas flow Obtain the current boiler load rate from the boiler distributed control system (DCS). By monitoring the flue gas temperature at the dryer tower outlet and humidity The effect of wastewater drying and evaporation in the drying tower was evaluated. Pressure sensors and flow meters were installed on the wastewater and compressed air pipelines to obtain the spray water volume W and compressed air supply pressure. Information such as these multi-source real-time data, after feature extraction, serves as the input features for the LS-SVM control model;

[0090] The controller uses a programmable logic controller (PLC) or an embedded industrial computer as its hardware platform. Internally, it stores the parameter set and kernel function parameters of a least squares support vector machine (LS-SVM) model obtained through offline training and runs an intelligent control algorithm based on LS-SVM. This least squares support vector machine prediction model, through learning from historical operating data, can effectively establish a complex nonlinear mapping relationship between boiler operating conditions and optimal spray control parameters. The model's input variables comprehensively consider current boiler load-related parameters (such as boiler load factor, flue gas inlet temperature, flue gas flow rate, dryer tower outlet temperature and humidity) and historical operating values ​​from the previous control cycle (such as the spray water volume and gas-liquid ratio at the previous moment), thus comprehensively capturing the dynamic characteristics of the system. The model's output is the predicted optimal control setpoint, including the optimal spray water volume. Optimal gas-liquid ratio and optimal compressed air pressure Equal to the set value;

[0091] The actuators include variable frequency pumps or regulating valves for adjusting wastewater flow, and compressed air regulating valves or variable frequency compressors for adjusting atomizing air pressure and flow. The controller, based on control commands output from the least squares support vector machine prediction model, drives each actuator to achieve precise control of the spray process parameters.

[0092] Specifically, the system changes the wastewater flow rate into the tower by adjusting the frequency converter of the wastewater pump or the opening of the outlet valve; simultaneously, it adjusts the supply pressure and flow rate of the atomizing air by adjusting the opening of the compressed air valve or the compressor power, thereby achieving precise control of the target gas-liquid ratio. When a significant change in spray volume is required, the controller can also start and stop multiple spray guns in stages according to control needs to achieve coarse adjustment of the total spray volume, and then achieve precise control of the total spray volume by finely adjusting the opening of the water supply valve of each nozzle.

[0093] In one embodiment, the system is further configured with a data buffer unit for storing operating conditions and control command data to support the periodic online updates and iterative optimization of the least squares support vector machine prediction model.

[0094] In one specific embodiment, the system is designed as follows:

[0095] The two-phase flow drying system for desulfurization wastewater is designed with a maximum wastewater evaporation rate of 3 t / h. The control system adopts the LS-SVM control model described in this invention.

[0096] 1. Data Acquisition and Preprocessing: 68 sets of operational data covering 50% to 100% of boiler load were collected from the power plant's DCS and historical database. Of these, 60 sets were used as training samples and 8 sets were used as test samples.

[0097] Seven variables were selected as model inputs:

[0098] Boiler load rate (%), flue gas inlet temperature (°C), flue gas flow rate (m³) 3 / h), drying tower outlet temperature (°C), outlet humidity (%), spray water volume at the previous moment (kg / s), and gas-liquid ratio at the previous moment.

[0099] The output variables are three: optimal spray water volume (kg / s), optimal gas-liquid ratio, and optimal compressed air pressure (MPa). To eliminate the influence of dimensions, all data are normalized.

[0100] Model Training and Performance: Independent LS-SVM prediction models were built for each of the three output variables, and the parameters were optimized using the PSO algorithm. The performance of the trained models is shown in the table below:

[0101] Table 1 Performance Evaluation of Least Squares Support Vector Machine Prediction Model

[0102]

[0103] Note: MAPE is the mean absolute percentage error, and R² is the absolute fractional variance.

[0104] The results show that all three LS-SVM sub-models have extremely high fitting accuracy (R² close to 1) and good predictive performance (MAPE of all models on the test set is less than 5%).

[0105] Based on the high-performance least squares support vector machine prediction model described above, the boiler load adaptive two-phase flow drying control process is as follows:

[0106] (1) When the controller starts, it loads the parameters (C, σ, α, b) of the three trained least squares support vector machine prediction models and safety control limits (such as maximum and minimum spray water volume, and the lower limit of the outlet temperature of 180°C).

[0107] (2) When the control cycle (e.g. every 30 seconds) begins, the controller reads the current status of the boiler and drying system from the sensors, including boiler load rate, flue gas inlet temperature, flue gas flow rate, drying tower outlet temperature and humidity, etc.

[0108] (3) The collected data and the historical operation values ​​of the previous period are organized into an input vector X and normalized.

[0109] (4) Input the feature vector X into three least squares support vector machine prediction models respectively to obtain the normalized output value of the prediction, and then calculate the physical suggested value through inverse normalization: spray water volume. Gas-liquid ratio and compressed air pressure ;

[0110] (5) The recommended values ​​are corrected by a safety rule controller to meet process safety constraints. For example, if the predicted outlet temperature is lower than 120°C, the spray water volume is forcibly reduced; if the predicted value exceeds the adjustable range of the actuator, the boundary value is used. The final control command after correction is as follows: , , ;

[0111] (6) The controller sends the final control command to the actuator to adjust the water pump / valve and compressed air valve so that the actual spray parameters gradually approach the set value;

[0112] (7) After a response delay, new sensor data reflects the change in system state. The controller continuously monitors the drying effect indicators. If there is a deviation, the algorithm will automatically adjust it in the next control cycle. At the same time, the current operating conditions and control actions can be stored as new data in the buffer for periodic online updates of the model.

[0113] The following is an example of the prediction result of a test sample to illustrate the prediction effect of the least squares support vector machine prediction model provided in this application.

[0114] Table 2. Example of prediction results for test samples (spray water volume)

[0115]

[0116] Test results show that the LS-SVM-based control model constructed in this application, namely the least squares support vector machine prediction model, can accurately predict the optimal spray parameters under different boiler loads, and achieve adaptive, stable and efficient control of the two-phase flow atomization drying process.

[0117] Thirdly, embodiments of this application provide an adaptive two-phase flow drying control device for boiler load changes. The adaptive two-phase flow drying control device for boiler load changes can be a personal computer (PC), laptop computer, server, or other device with data processing capabilities.

[0118] In this embodiment, the two-phase flow drying control device that adapts to boiler load changes may include a processor, a memory, a communication interface, and a communication bus.

[0119] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0120] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces for interconnecting internal components of the two-phase flow drying control equipment that adapts to boiler load changes, as well as interfaces for interconnecting the two-phase flow drying control equipment with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0121] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0122] The processor can be a general-purpose processor, which can call the adaptive boiler load change two-phase flow drying control program stored in the memory and execute the adaptive boiler load change two-phase flow drying control method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the adaptive boiler load change two-phase flow drying control program is called can refer to the various embodiments of the adaptive boiler load change two-phase flow drying control method of this application, which will not be repeated here.

[0123] Fourthly, embodiments of this application also provide a readable storage medium.

[0124] This application has a readable storage medium storing an adaptive boiler load change two-phase flow drying control program, wherein when the adaptive boiler load change two-phase flow drying control program is executed by a processor, it implements the steps of the adaptive boiler load change two-phase flow drying control method as described above.

[0125] The method implemented when the adaptive boiler load change two-phase flow drying control program is executed can be referred to in various embodiments of the adaptive boiler load change two-phase flow drying control method of this application, and will not be repeated here.

[0126] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A two-phase flow drying control method that adapts to changes in boiler load, characterized by, The method comprises the following steps: Real-time acquisition of boiler load parameters and two-phase flow drying system operating parameters to construct an input vector; Inputting the input vector into a pre-trained least squares support vector machine prediction model, which directly outputs the optimal set value suitable for the current boiler load by establishing a nonlinear mapping relationship between the boiler operating condition and the optimal control parameter; According to the preset process safety constraints, the optimal set value is checked and corrected at the boundary to generate a final control instruction; Control execution of the final control instruction to adjust the two-phase flow parameters, and form a closed-loop control based on system output feedback to realize adaptive optimization of the two-phase flow drying process.

2. The adaptive boiler load change two-phase flow drying control method according to claim 1, characterized by, The construction of the least squares support vector machine prediction model is based on the principle of structural risk minimization, and its optimization problem is defined as: ； The equality constraint condition is: ； where C is a regularization parameter balancing model complexity and training error; is the training error for the i-th sample.

3. The adaptive boiler load change two-phase flow drying control method according to claim 2, wherein, The regression function of the least squares support vector machine prediction model is: ； In the formula: is a kernel function that satisfies the Mercer condition.

4. The adaptive boiler load change two-phase flow drying control method according to claim 3, wherein, The kernel function is a radial basis function, and its expression is: ； In the formula, σ is the kernel function parameter, and the radial function range is controlled.

5. The adaptive boiler load change two-phase flow drying control method according to claim 3, wherein, The regularization parameter and kernel function parameter are determined by a particle swarm optimization algorithm to adapt to a fitness function is minimized, where is the root mean square error of K-fold cross-validation.

6. The two-phase flow drying control method for adaptive boiler load variation as claimed in claim 1 wherein, The input vector of the least squares support vector machine prediction model at least includes: boiler load rate, flue gas inlet temperature, flue gas flow, drying tower outlet temperature, drying tower outlet humidity, last control cycle spray water volume, and last control cycle gas-liquid ratio.

7. A two-phase flow drying control system that is self-adapting to boiler load variations for implementing the method of any one of claims 1 to 6, characterized in that, It comprises: A sensor module for real-time acquisition of boiler load parameters and two-phase flow drying system operating parameters to construct the input vector; A controller module integrated with the least squares support vector machine prediction model and a rule correction unit, which is in communication connection with the sensor module, and is used for predicting the optimal control set value according to the input parameters and completing safety correction; An actuator module in communication connection with the controller module, which is used for controlling execution of the final control instruction to adjust the two-phase flow parameters, and forming a closed-loop control based on system output feedback to realize adaptive optimization of the two-phase flow drying process; The sensor module, the controller module and the actuator module constitute a closed-loop control loop.

8. The two-phase flow drying control system for adaptive boiler load changes of claim 7, wherein, The controller module is a programmable logic controller or an embedded industrial computer, which internally stores the least squares support vector machine model parameter set and the kernel function parameter obtained by offline training.

9. The two-phase flow drying control system for adaptive boiler load changes of claim 7, wherein, The actuator module includes a variable frequency pump or a regulating valve for adjusting wastewater flow, and a compressed air regulating valve or a variable frequency compressor for adjusting the pressure and flow of atomizing air.

10. The two-phase flow drying control system for adaptive boiler load changes of claim 7, wherein, The system is also provided with a data buffer unit for storing operating conditions and control instruction data to support periodic online updating and iterative optimization of the least squares support vector machine prediction model.

Images