Thermal power generating unit dry-wet state conversion intelligent regulation and control system based on multi-mode sensing fusion
By using a multimodal sensing fusion system, the problems of inaccurate state identification and deterioration of control quality during the dry-wet state transition of the boiler steam-water system in thermal power units have been solved. This system achieves high-precision state perception and multi-level safety protection, thereby improving the stability and safety of the system.
Patent Information
- Application Number
- CN202511667791.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-14
- Publication Date
- 2026-03-06
AI Technical Summary
Existing technologies rely on single or limited sensor signals during the dry-wet state transition process of boiler steam-water systems in thermal power units, leading to inaccurate state identification. Traditional control strategies are rigid and unable to adapt to dynamic changes, resulting in deteriorated control quality and insufficient safety.
A multimodal sensing fusion system is adopted, including modules for multimodal sensing and monitoring, data transmission, fusion, intelligent identification, control decision-making, adaptive optimization, and safety monitoring. Sensor data time alignment is achieved through the IEEE 1588 protocol, bias is eliminated through DS evidence fusion and least squares calibration, and state identification and control are performed by combining support vector machines and fuzzy inference algorithms. PID parameters are dynamically adjusted, and multi-layer safety protection is implemented.
It achieves high-precision status perception, reduces the false judgment rate, improves the stability and security of the system, reduces the number of unplanned downtimes, and extends the equipment life.
Smart Images
Figure CN121613709A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automatic control and condition monitoring technology for thermal power units, specifically to an intelligent control system for the dry and wet transition states of thermal power units based on multimodal sensor fusion. Background Technology
[0002] With the advancement of new power system construction and large-scale grid connection of new energy sources, the demand for grid load regulation has significantly increased. Thermal power units need to frequently participate in deep peak shaving and flexibility upgrades, leading to increasingly frequent dry-wet state transitions in boiler steam-water systems. This transition process involves complex phase changes and heat and mass transfer, characterized by strong nonlinearity, large time lag, and multivariable coupling. Existing technologies mainly rely on single or limited sensor signals (such as temperature or pressure) for state judgment, which is prone to inaccurate state identification due to sensor measurement deviations, signal interference, or discrepancies between local characteristics and the global state. At the control level, traditional control strategies (such as fixed-parameter PID) are difficult to adapt to the dynamically changing characteristics of the object during the transition process, often resulting in control quality deterioration problems such as main steam temperature overshoot and severe fluctuations in separator liquid level. In severe cases, the unit may be forced to exit automatic control mode. Furthermore, existing systems lack in-depth protection for the reliability of sensor data and the overall safety status of the control system, resulting in insufficient operational safety under extreme conditions and the risk of unplanned shutdowns. Therefore, there is an urgent need for a dry-wet transition intelligent control system that can achieve high-precision state perception, adaptive intelligent control, and multi-level safety protection to ensure the safe, stable, and efficient operation of thermal power units within a wide load range. Summary of the Invention
[0003] (a) Technical problems to be solved
[0004] To address the shortcomings of existing technologies, this invention provides an intelligent control system for the dry-wet transition of thermal power units based on multimodal sensor fusion. It has the advantages of high-precision state perception, adaptive intelligent control, and multi-level safety protection, and solves the problems of inaccurate state identification, deterioration of control quality during transition, and insufficient safety caused by single sensor information and rigid control strategies in traditional methods.
[0005] (II) Technical Solution
[0006] To achieve the above objectives, the present invention provides the following technical solution: a multimodal sensor fusion-based intelligent control system for dry and wet state transitions in thermal power units, comprising a multimodal sensor monitoring module, a multimodal data transmission module, a multimodal sensor fusion module, a dry and wet state intelligent identification module, an intelligent control decision module, an adaptive optimization module, a control execution module, and a safety monitoring and protection module;
[0007] The multimodal sensing and monitoring module is responsible for collecting heterogeneous sensing signals covering the boiler steam-water system, and preprocessing the collected raw signals by filtering, denoising and scaling transformation. It also achieves microsecond-level time alignment and spatial coordinate unification of multi-channel sensor data through the IEEE 1588 precision clock protocol.
[0008] The multimodal data transmission module encrypts and transmits the pre-processed raw signal to the multimodal sensing fusion module;
[0009] The multimodal sensing fusion module performs basic probability allocation and confidence synthesis on conflicting evidence from multiple sensors regarding dry and wet states based on the fusion target proposition result value; and dynamically estimates and compensates for sensor drift errors based on the calibrated optimal parameter result value.
[0010] The dry / wet state intelligent identification module is based on the confidence vector and state feature parameters output by the multimodal sensing fusion module. It adopts a hybrid algorithm of support vector machine and fuzzy rule reasoning to identify the current dry / wet state and transition trend of the unit in real time, and outputs the state identification result and state confidence index.
[0011] The intelligent control decision module generates fuzzy control rules for coal-water ratio, desuperheating water, and feedwater pump speed online based on the state recognition results and load commands. It outputs PID parameters for the main steam temperature and separator level control loop based on fuzzy formulas to achieve seamless switching of composite control strategies.
[0012] The adaptive optimization module, based on a model reference adaptive mechanism, optimizes the fuzzy rule weights and PID correction coefficients online according to the dynamic deviation between the actual unit response and the reference model, and outputs the optimized control weight vector. , and ;
[0013] The control execution module receives the comprehensive control command vector generated by the intelligent control decision module, distributes the control commands to the execution mechanism, and monitors the feedback status of the actuator.
[0014] The safety monitoring and protection module monitors the reliability of fused data, control output deviation, and key parameter over-limit situations in real time through multi-level residual analysis and dynamic threshold determination algorithms, triggering rapid load reduction or shutdown protection.
[0015] Preferably, the multimodal sensing and monitoring module includes a temperature sensing unit, a pressure sensing unit, a flow sensing unit, and a humidity detection unit; the multimodal sensing fusion module includes a DS evidence fusion unit and a least squares calibration unit; and the intelligent control decision-making module includes a fuzzy inference unit and a PID control unit.
[0016] Preferably, the temperature sensing unit collects the temperature field distribution of the unit's thermal process by using K-type / E-type thermocouples and PT100 resistance thermometers installed at key nodes of the boiler superheater, reheater, economizer, and steam-water pipeline.
[0017] Preferably, the pressure sensing unit collects steam pressure, feedwater pressure and furnace negative pressure through ceramic piezoresistive / capacitive pressure transmitters distributed in each pressure vessel of the steam-water system.
[0018] Preferably, the flow sensing unit collects the volumetric / mass flow rate of steam and feedwater through a standard throttling device and a matching differential pressure transmitter installed in the main steam pipeline and feedwater pipeline.
[0019] Preferably, the humidity detection unit collects the trace moisture content in superheated steam and saturated steam using an online humidity analyzer based on the resistance-capacitance method or laser absorption spectroscopy.
[0020] Preferably, the DS evidence fusion unit calculates the fusion target proposition using the following formula: In the formula, This indicates the fusion objective proposition. , Let A and B represent the basic probability assignment functions for the propositions regarding dry and wet states, respectively. , Let A and B represent the propositional subsets of sensors A and B, respectively. This represents the target proposition after fusion. This represents the empty set.
[0021] Preferably, the least squares calibration unit calculates the optimal parameters after calibration, and the calculation formula is as follows: In the formula, , These represent the optimal parameters for the calibrated slope and intercept, respectively. Indicates the number of samples. , These represent the true value of the i-th reference standard and the measured value of the sensor to be calibrated, respectively.
[0022] Preferably, the fuzzy inference unit employs the Mamdani inference method, and the output control quantity is synthesized from the fuzzy relationship between the input error and the rate of change of the error, and its calculation formula is as follows: In the formula, Indicates control output. Indicates the scaling factor. This represents the membership function value of the input variable corresponding to the i-th rule. Let represent the conclusion of the i-th fuzzy rule. Indicates the total number of activated fuzzy rules;
[0023] The PID control unit employs an adaptive PID control calculation formula to perform closed-loop PID control of the main steam temperature or separator liquid level, with a proportional gain of [missing information]. ,integral With differential parameters Output based on fuzzy reasoning Automatic adjustment to changes in operating conditions:
[0024] Its parameter tuning adopts a fuzzy adaptive correction strategy: Proportional gain: ; Points Time: ; Differential time: ; in the formula This indicates the output of the PID controller. Indicates control deviation. , , These represent the baseline PID parameters, Dynamic correction based on wall temperature deviation Adaptive based on the rate of change of liquid level The differential action is enhanced based on the membership degree of dry and wet states. This represents the amplitude of the control quantity output by fuzzy inference. , and They automatically adjust according to changes in operating conditions.
[0025] Preferably, the dry / wet state intelligent identification module is based on the fusion target proposition calculated by the multimodal sensing fusion module. calibrated slope With intercept The optimal parameters, combined with the state feature library constructed from historical operating data, are used to identify the current dry and wet status and transition trend of the unit in real time by employing a hybrid algorithm of support vector machine and fuzzy rule reasoning.
[0026] The intelligent control decision module calculates the control output according to the formula. The resulting values are used to generate fuzzy control rules for the coal-water ratio, desuperheating water, and feedwater pump speed online through fuzzy rule base mapping and operating condition weight allocation algorithm. These rules are then used to output PID controller rules. The resulting values employ a dynamic parameter tuning and non-disruptive switching strategy. Through feedforward compensation and incremental output smoothing, the PID parameters of the main steam temperature and separator level control loops are automatically adjusted to achieve non-disruptive switching of the composite control strategy.
[0027] Compared with existing technologies, this invention provides an intelligent control system for the dry-wet transition of thermal power units based on multimodal sensor fusion, which has the following beneficial effects:
[0028] 1. This invention achieves the beneficial effect of high-precision state determination through multimodal sensing fusion and intelligent recognition algorithms. By integrating temperature, pressure, humidity and other heterogeneous sensor data, it uses DS evidence theory + least squares calibration to eliminate single sensor bias, and supports a hybrid algorithm of support vector machine (SVM) and fuzzy inference to achieve millisecond-level recognition of the state during dry / wet / transition. This significantly improves the confidence of the final state of the system while greatly reducing the false judgment rate.
[0029] 2. This invention achieves stable control under complex operating conditions through a composite control strategy and adaptive optimization mechanism. It dynamically tunes PID parameters using fuzzy inference, combined with model reference adaptive correction. , , Under complex scenarios involving sudden load changes and coal blending, the system can stabilize the fluctuation range of the main steam temperature and reduce the overshoot of the separator level.
[0030] 3. This invention achieves high reliability under extreme conditions through a multi-layered safety protection system. By implementing three-level residual analysis (sensor level → fusion level → decision level) through the safety monitoring module, the response time of RB protection actions is reduced. Combined with the redundant actuator switching strategy, the number of unplanned shutdowns is reduced, ultimately extending the equipment life. Attached Figure Description
[0031] Figure 1 This is a system flowchart of the present invention. Detailed Implementation
[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0033] Please see Figure 1The intelligent control system for dry and wet transition states of thermal power units based on multimodal sensor fusion includes a multimodal sensor monitoring module, a multimodal data transmission module, a multimodal sensor fusion module, a dry and wet state intelligent identification module, an intelligent control decision module, an adaptive optimization module, a control execution module, and a safety monitoring and protection module.
[0034] The multimodal sensing and monitoring module is responsible for collecting heterogeneous sensing signals covering the boiler steam-water system, and preprocessing the collected raw signals by filtering, denoising and scaling transformation. It achieves microsecond-level time alignment and spatial coordinate unification of multi-channel sensor data such as temperature, pressure, flow rate, liquid level, wall temperature, vibration and acoustic through the IEEE 1588 precise clock protocol.
[0035] The multimodal data transmission module encrypts and transmits the pre-processed raw signal to the multimodal sensing fusion module, enabling real-time and reliable transmission of sensing data and ensuring data integrity and timeliness.
[0036] The multimodal sensing fusion module performs basic probability allocation and confidence synthesis on conflicting evidence from multiple sensors regarding dry and wet states based on the fusion target proposition result value; it dynamically estimates and compensates for sensor drift errors based on the calibrated optimal parameter result value, thereby improving the accuracy of fused data.
[0037] The dry / wet state intelligent identification module is based on the confidence vector and state feature parameters output by the multimodal sensing fusion module. It adopts a hybrid algorithm of support vector machine (SVM) and fuzzy rule reasoning to identify the current dry / wet state and transition trend of the unit in real time, and outputs the state identification result (wet / dry / transitioning) and state confidence index.
[0038] The intelligent control decision module generates fuzzy control rules for coal-water ratio, desuperheating water, and feedwater pump speed online based on the state recognition results and load commands. It outputs PID parameters for the main steam temperature and separator level control loop based on fuzzy formulas to achieve seamless switching of composite control strategies.
[0039] The adaptive optimization module, based on a model reference adaptive mechanism, optimizes the fuzzy rule weights and PID correction coefficients online according to the dynamic deviation between the actual unit response and the reference model, and outputs the optimized control weight vector. , , Enhance adaptability to complex working conditions;
[0040] The control execution module receives the comprehensive control command vector (including feedwater pump speed command, fuel quantity adjustment command, and desuperheating water valve opening command) generated by the intelligent control decision module, distributes the control commands to the actuators such as feedwater pump hydraulic coupling, fuel regulating valve, and desuperheating water regulating valve, and monitors the actuator feedback status.
[0041] The safety monitoring and protection module monitors the reliability of fused data, control output deviation, and key parameter exceedance in real time through multi-level residual analysis and dynamic threshold determination algorithms, triggering rapid load reduction (RB) or shutdown protection to ensure the safety of the transition process.
[0042] This invention achieves the beneficial effects of improving the safety and economy of unit operation through full-module closed-loop collaboration (monitoring-fusion-identification-decision-optimization-execution-protection) and multi-level safety mechanisms. It effectively avoids faults such as over-temperature and liquid level exceeding limits during the transition process, reduces coal consumption, reduces equipment wear, and extends the service life of the unit.
[0043] The multimodal sensing and monitoring module includes a temperature sensing unit, a pressure sensing unit, a flow sensing unit, and a humidity detection unit; the multimodal sensing fusion module of the intelligent control system for dry-wet transition of thermal power units based on multimodal sensing fusion includes a DS evidence fusion unit and a least squares calibration unit; the intelligent control decision module of the intelligent control system for dry-wet transition of thermal power units based on multimodal sensing fusion includes a fuzzy inference unit and a PID control unit.
[0044] The temperature sensing unit collects the temperature field distribution of the unit's thermal process by installing K-type / E-type thermocouples and PT100 resistance thermometers at key nodes of the boiler superheater, reheater, economizer, and steam-water pipeline. The collection range covers thermocouple arrays, resistance thermometer arrays, signal conditioning circuits, and cold junction compensation modules.
[0045] The pressure sensing unit collects steam pressure, feedwater pressure and furnace negative pressure through ceramic piezoresistive / capacitive pressure transmitters distributed in various pressure vessels of the steam-water system. The collection range covers high-stability pressure transmitters, isolation diaphragm assemblies and HART protocol communication modules.
[0046] The flow sensing unit collects the volumetric / mass flow rate of steam and feedwater through standard throttling devices (such as orifice plates and nozzles) and matching differential pressure transmitters installed in the main steam pipeline and feedwater pipeline. The collection range covers throttling devices, high-precision differential pressure transmitters, temperature and pressure compensation modules, and flow totalizers.
[0047] The humidity detection unit collects trace moisture content in superheated steam and saturated steam using an online humidity analyzer based on the resistance-capacitance method or laser absorption spectroscopy. The collection range covers a high-frequency capacitive sensor / tunable laser, photoelectric conversion module, and calibration gas path system.
[0048] The advantages are: through the various dedicated units of the above-mentioned multimodal sensing and monitoring module, comprehensive and highly synchronous monitoring of multiple physical quantities (temperature, pressure, flow rate, humidity) of the boiler steam-water system can be achieved, thereby providing a high-quality, spatiotemporally unified raw data foundation for subsequent data fusion and status identification, fundamentally improving the breadth and accuracy of system perception.
[0049] The DS evidence fusion unit calculates the fusion target proposition (state to be determined), and its calculation formula is as follows: In the formula, This represents the fusion target proposition (the state to be determined, whether wet or dry). , Let A and B represent the basic probability assignment functions for the propositions regarding dry and wet states, respectively. , Let A and B represent the propositional subsets of sensors A and B, respectively. The denominator represents the target proposition after fusion (dry / wet state), with a conflict coefficient correction term in the denominator to avoid distortion of the fusion result. This represents the empty set (a set containing no elements).
[0050] The advantages are: the DS evidence fusion unit calculates the fusion target proposition (the state to be determined), performs basic probability allocation and confidence synthesis on conflicting evidence from multiple sensors regarding dry and wet states, effectively solves the risk of misjudgment caused by false alarms from a single sensor, suppresses interference from contradictory evidence through the conflict coefficient correction term, significantly improves the robustness of dry / wet state classification, and enables the state recognition results to have quantifiable confidence indicators.
[0051] The least squares calibration unit calculates the optimal parameters after calibration, dynamically estimates and compensates for sensor drift errors, and improves the accuracy of fused data. Its calculation formula is as follows: In the formula, , These represent the optimal parameters for the calibrated slope (gain coefficient) and intercept (zero offset), respectively. Indicates the number of samples. , Let the true values of the i-th reference base (such as the median value of redundant sensors or the calibration value of standard instruments) and the measured values of the sensor to be calibrated (including the raw output of drift error) be represented respectively. The result is achieved by minimizing the sum of squared residuals. Achieve optimal parameter estimation.
[0052] The advantages are: the least squares calibration unit calculates the optimal parameters after calibration, dynamically estimates and compensates for sensor drift error, improves the accuracy of fused data, thereby eliminating zero drift caused by environmental temperature changes and component aging in real time, achieving the global optimal solution of gain and offset by minimizing the sum of squared residuals, reducing data accumulation error in long-term operation, and ultimately ensuring the benchmark consistency of cross-modal data fusion.
[0053] The fuzzy inference unit uses the Mamdani inference method, and the output control quantity is synthesized from the fuzzy relationship between the input error and the error change rate. Its calculation formula is as follows: In the formula, Indicates control output. This represents the scaling factor, used to adjust the output range to suit the actuator. This represents the membership function value of the input variable corresponding to the i-th rule. This represents the conclusion of the i-th fuzzy rule (such as the center value of the fuzzy set, such as "negative large" or "zero"). Indicates the total number of activated fuzzy rules;
[0054] The advantages are: by using the Mamdani inference method through the fuzzy inference unit, the output control quantity is synthesized from the fuzzy relationship between the input error and the error change rate. Based on the fuzzy output, the PID parameters of the main steam temperature and separator liquid level control loop are automatically adjusted to achieve seamless switching of the composite control strategy. This overcomes the limitation of traditional PID fixed parameters being difficult to adapt to strongly coupled working conditions. The membership function weight allocation is used to achieve multi-rule collaborative decision-making. The scaling factor K is used to scale the control quantity range to match the characteristics of different actuators, ensuring the smoothness of the dry-wet transition period.
[0055] The PID control unit uses an adaptive PID control calculation formula to perform closed-loop PID control of the main steam temperature or separator liquid level. Its proportional, integral, and derivative parameters ( , , Automatically adjust based on fuzzy inference output U and changes in operating conditions:
[0056] Its parameter tuning adopts a fuzzy adaptive correction strategy: Proportional gain: ; Points Time: ; Differential time: ; in the formula This indicates the output of the PID controller (feed pump speed command % or desuperheating water valve opening %). This indicates the control deviation. For the level circuit, it represents the separator level deviation (actual - set value), and for the temperature circuit, it represents the main steam temperature deviation (°C). , , These represent the baseline PID parameters (tuned by typical operating condition tests). Based on dynamic correction of wall temperature deviation, the temperature is reduced by 20% to suppress oscillations when the temperature exceeds the limit. Based on the adaptive liquid level change rate, the integral time is increased during the transition phase to avoid integral saturation. By enhancing the differential action based on the membership degree of dry and wet states, the predictability of transition processes can be improved. This represents the amplitude of the control quantity output by fuzzy inference; it enhances the proportional effect when there is a large deviation. , , They automatically adjust according to changes in operating conditions.
[0057] The advantages are: The PID control unit employs an adaptive PID control calculation formula to perform closed-loop PID control of the main steam temperature or separator liquid level, introducing fuzzy inference output U as a feedforward compensation signal, and dynamically correcting based on wall temperature deviation. Suppress over-temperature oscillations and adjust according to the rate of liquid level change. To prevent integral saturation, membership enhancement is combined with dry and wet states. It has predictive capabilities and achieves a self-mapping relationship between PID parameters and operating condition complexity.
[0058] The dry / wet state intelligent identification module is based on the fusion target proposition calculated by the multimodal sensing fusion module. calibrated slope With intercept The optimal parameters, combined with a state feature library constructed from historical operating data (such as load and coal type), are used to employ a hybrid algorithm of Support Vector Machine (SVM) and fuzzy rule inference to identify the current dry / wet state and transition trend of the unit in real time. The specific workflow is as follows:
[0059] S1.1. Support Vector Machine (SVM) is used to classify high-dimensional fused sensor data, and its structural risk minimization principle is used to obtain preliminary and robust classification results for dry / wet states.
[0060] S1.2 Then, fuzzy rule reasoning is used to finely characterize the classification results and continuous dynamic process of state transition of SVM, defining rules such as "if the superheat of the steam at the separator outlet is high and its rate of change is positive, then the confidence of belonging to the dry state is enhanced".
[0061] S1.3 The final output is a structured state identification result (wet / dry / in transition) and its corresponding state confidence index (represented by probability values between 0 and 1), and also outputs dynamic parameters used to describe the trend, such as the state transition rate.
[0062] By combining the DS evidence fusion and least squares calibration of the multimodal sensing fusion module with the hybrid algorithm of the dry and wet state intelligent identification module, the system achieves the beneficial effect of overcoming the limitations of a single sensor and accurately identifying dry and wet states and transition trends. The state identification confidence level is ≥0.9, and the state transition prediction is 2-3 minutes in advance, providing reliable support for control decisions.
[0063] The intelligent control decision module of the intelligent control system for thermal power units based on multimodal sensor fusion calculates the control output according to the formula. The resulting values are used to generate fuzzy control rules for the coal-water ratio, desuperheating water, and feedwater pump speed online through fuzzy rule base mapping and operating condition weight allocation algorithm. These rules are then used to output PID controller rules. The resulting values employ a dynamic parameter tuning and non-disruptive switching strategy. Through feedforward compensation and incremental output smoothing, the PID parameters of the main steam temperature and separator level control loops are automatically adjusted to achieve non-disruptive switching of the composite control strategy.
[0064] The advantages are: by linking the fuzzy inference unit and the PID regulation unit of the intelligent control decision module, and with the online parameter optimization of the adaptive optimization module, the beneficial effects of adapting to complex operating condition fluctuations, achieving seamless switching between dry and wet states and high-precision control are achieved. The main steam temperature fluctuation range is reduced, the separator liquid level regulation error is reduced, and thus the stability of the transition process is greatly improved.
[0065] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A dry-wet state intelligent regulation system for a thermal power unit based on multi-modal sensor fusion, characterized in that, The system comprises a multi-modal sensor monitoring module, a multi-modal data transmission module, a multi-modal sensor fusion module, a dry-wet state intelligent identification module, an intelligent control decision module, a self-adaptive optimization module, a control execution module, and a safety monitoring and protection module. The multi-modal sensor monitoring module is responsible for collecting heterogeneous sensor signals covering the boiler steam-water system, and performing filtering, denoising, and scaling transformation preprocessing on the collected raw signals. The multi-modal data transmission module transmits the preprocessed raw signals to the multi-modal sensor fusion module. The multi-modal sensor fusion module performs basic probability assignment and confidence synthesis on the conflict evidence of multi-sensors regarding the dry-wet state based on the fusion target proposition result value. The dry-wet state intelligent identification module uses a support vector machine and a fuzzy rule reasoning hybrid algorithm to identify the current dry-wet state and transition trend of the unit in real time based on the confidence vector and state characteristic parameters output by the multi-modal sensor fusion module, and outputs the state identification result and state confidence index. The intelligent control decision module generates fuzzy control rules for coal-water ratio, desuperheating water, and feedwater pump speed based on the state identification result and load instruction, and outputs PID parameters for automatically adjusting the main steam temperature and separator level control loop based on fuzzy formulas, thereby realizing disturbance-free switching of the composite control strategy. The control execution module receives the comprehensive control instruction vector generated by the intelligent control decision module, distributes the control instructions to the actuators, and monitors the feedback state of the actuators. The adaptive optimization module is based on a model reference adaptive mechanism, and according to the dynamic deviation of the actual response of the unit from the reference model, online optimizes the fuzzy rule weight and the PID correction coefficient, and outputs the optimized control weight vector 、 and ; The safety monitoring and protection module monitors the fusion data reliability, control output deviation, and key parameter out-of-limit conditions in real time through multi-level residual analysis and dynamic threshold determination algorithm, and triggers fast load reduction or shutdown protection. The multi-modal sensor monitoring module comprises a temperature sensing unit, a pressure sensing unit, a flow sensing unit, and a humidity detection unit.
2. The multi-modal sensor fusion-based intelligent regulation system for dry-wet state transition of a thermal power generating unit according to claim 1, characterized in that: The temperature sensing unit collects the temperature field distribution of the thermal process of the unit through K-type / E-type thermocouples and PT100 thermal resistors installed at key nodes of the boiler superheater, reheater, economizer, and steam-water pipeline.
3. The multi-modal sensor fusion-based intelligent regulation system for dry-wet state transition of a thermal power generating unit according to claim 2, characterized in that: The pressure sensing unit collects steam pressure, feedwater pressure, and furnace negative pressure through ceramic piezoresistive / capacitive pressure transmitters distributed in each pressure vessel of the steam-water system.
4. The multi-modal sensor fusion-based intelligent regulation system for dry-wet state transition of a thermal power generating unit according to claim 2, characterized in that: The flow sensing unit collects the volume / mass flow of steam and feedwater through standard throttling devices installed in the main steam pipeline and feedwater pipeline and the corresponding differential pressure transmitters.
5. The multi-modal sensor fusion-based intelligent regulation system for dry-wet transition of a thermal power generating unit according to claim 2, characterized in that: The humidity detection unit collects the trace moisture content in superheated steam and saturated steam through online humidity analyzers based on the resistance-capacitance method or laser absorption spectroscopy principle.
6. The multi-modal sensor fusion based intelligent regulation system for dry-wet transition state of thermal power generating unit according to claim 2, characterized in that: The D-S evidence fusion unit calculates the fusion target proposition, and the calculation formula is:
7. The multi-modal sensor fusion-based intelligent regulation system for dry-wet transition of a thermal power generating unit according to claim 2, characterized in that: The least squares calibration unit calculates the calibrated optimal parameters, and the calculation formula is: In the formula, denotes the fusion target proposition, , denote the basic probability assignment functions of the sensor A, B about the dry-wet state proposition, , denote the proposition subsets of the sensor A, B, denotes the fused target proposition, denotes the empty set.
8. The multi-modal sensor fusion-based intelligent regulation system for dry-wet state transition of a thermal power generating unit according to claim 2, characterized in that: In the formula, , respectively represent the calibrated slope and intercept optimal parameters, represents the sample number, , respectively represent the i-th group of reference benchmark true values and the to-be-calibrated sensor measurement values.
9. The multi-modal sensor fusion-based intelligent regulation system for dry-wet transition of a thermal power generating unit according to claim 2, characterized in that: The fuzzy inference unit adopts Mamdani inference method, and the output control quantity is the fuzzy relation synthesis of input error and error change rate, and the calculation formula is: in the formula, represents a control output, represents a proportional factor, represents the input variable membership function value corresponding to the ith rule, represents the conclusion of the ith fuzzy rule, represents the total number of activated fuzzy rules; The PID adjusting unit adopts adaptive PID adjusting calculation formula to carry out closed loop PID adjustment of the main steam temperature or the separator liquid level, the proportion , integration and differential parameters are automatically adjusted according to fuzzy inference output and working condition change: The parameter setting adopts fuzzy self-adaptive correction strategy: Proportional gain: ; Integration time: ; Differential time: ; In the formula represents the PID controller output, represents the control deviation, respectively represent the reference PID parameters, is dynamically corrected according to the wall temperature deviation, is self-adaptive according to the liquid level change rate, enhances the differential action according to the dry and wet state membership degree, represents the control amount amplitude of the fuzzy inference output, and are automatically adjusted according to the working condition change, respectively. 10. The multi-modal sensor fusion based intelligent regulation system for dry-wet transition state of thermal power generating unit according to claim 1, characterized in that: The dry-wet state intelligent identification module is based on a fusion target proposition calculated by a multi-modal sensor fusion module , a calibrated slope and an intercept , an optimal parameter, and a state feature library constructed in combination with historical working condition data, a support vector machine and a fuzzy rule reasoning hybrid algorithm are used to identify the current dry-wet state and conversion trend of the unit in real time; The intelligent control decision module calculates the control output according to the formula The result value is mapped with the working condition weight distribution algorithm by the fuzzy rule base to generate fuzzy control rules of the coal-water ratio, desuperheating water, and feed pump rotating speed online, and is based on the PID controller output The result value is automatically adjusted by the PID parameters of the main steam temperature and the separator liquid level control loop through feedforward compensation and incremental output smoothing processing, and the disturbance-free switching of the compound control strategy is realized.