660MW once-through boiler slag dryer system based on gradient cooling and intelligent regulation and control

Abstract

The invention relates to a 660MW once-through boiler slag dryer system based on gradient cooling and intelligent regulation and control. The invention aims to solve the technical problems of slag blockage, steel belt damage and low energy efficiency caused by non-uniform cooling of high-temperature slag. According to the technical scheme, the device comprises a stepped multi-stage cooling cavity which is divided into high-temperature sections (gt; 700 DEG C), a medium temperature section (400-700 DEG C) and a low temperature section (lt; through deep coupling of the structure, the algorithm and the monitoring, gradient cooling provides a physical basis for intelligent regulation and control, the MPC algorithm realizes advanced control by using a magnetic grid high-precision speed signal, the slag blocking early warning model output directly drives an air volume redistribution strategy, and the slag blocking early warning model output directly drives the air volume redistribution strategy. And closed-loop optimization is formed.

Description

Technical Field

[0001] The present invention belongs to the technical field of boiler slag removal equipment in thermal power plants, and specifically relates to a 660MW direct current boiler slag removal system based on gradient cooling and intelligent regulation, which solves the problems of slag blockage, steel belt damage and low energy efficiency caused by uneven cooling of high-temperature slag. Background Art

[0002] Traditional dry slag machines have three major defects:

[0003] Low cooling efficiency: The single-stage cooling structure cannot adapt to the phase change characteristics of slag (>700℃ molten state → 400℃ glass transition), resulting in insufficient heat exchange in the high-temperature section and overcooling and bonding in the low-temperature section;

[0004] Control lag: Manual valve adjustment response is slow (>30 seconds), and over-temperature events occur frequently (an average of 28 times per month for a 660MW unit);

[0005] Unreliable monitoring: Contact steel belt detection is prone to wear and tear, and the fault location error is greater than 1m.

[0006] The existing improvement scheme does not solve the coordination problem between temperature zoning and cooling device, and the accuracy drops sharply in a vibration environment. Summary of the Invention

[0007] The purpose of the present invention is to solve the above technical problems and provide a 660MW direct current boiler slag dryer system based on gradient cooling and intelligent control.

[0008] A 660MW once-through boiler slag dryer system based on gradient cooling and intelligent control, including:

[0009] The stepped multi-stage cooling chamber is divided into a high-temperature section (>700°C), a medium-temperature section (400-700°C) and a low-temperature section (<400°C) with mutually exclusive temperature ranges along the slag conveying direction;

[0010] Differentiated cooling device:

[0011] The high-temperature section is equipped with a double-layer staggered array of directional nozzles with dynamically adjustable inclination angles. The upper nozzles have an inclination angle of 30°-60°, while the lower nozzles have an inclination angle of 15°-45° in the opposite direction, forming a vortex counter-flow with a swirl intensity of ≥0.6.

[0012] The medium temperature section is equipped with a honeycomb guide plate with an aperture of 50-100mm and a swirl nozzle with a swirl angle of 40°-70°, which prolongs the slag retention time by 30%-40%;

[0013] The low temperature section is equipped with a porous plate flow-distributing layer with an opening rate of 30%-50%, so that the standard deviation of the outlet temperature is ≤5℃;

[0014] Intelligent control system:

[0015] The dynamic compensation mechanism based on model predictive control (MPC) takes outlet temperature, slag flow rate and particle size as inputs, and the response time is less than 2 seconds to compensate the air volume set value. The objective function is:

[0016] Min J=∑[Tset(t)-Tactual(t)] 2 +λ•Δu(t) 2 ;

[0017] Where, J: objective function (performance index), which needs to be minimized;

[0018] Tset(t): set temperature of the dry slag machine outlet (°C);

[0019] Tactual(t): actual measured temperature at the outlet of the dry slag machine (°C);

[0020] Δu(t): The adjustment range of the air volume control amount (i.e., the control increment);

[0021] λ: Weight coefficient, used to balance the relative importance between temperature deviation and control action amplitude;

[0022] The summation term ∑: usually represents the accumulation in the prediction time domain (under the MPC framework);

[0023] The YOLOv5 slag flow morphology recognition model, improved based on transfer learning, replaces the backbone network with ResNet34. It inputs slag layer thickness, slag block contour, and temperature distribution data fused from a dual-band infrared sensor and outputs the slag blockage risk probability. When the risk probability is greater than 80%, the air volume in the high-temperature section is reduced by 10% per level, while the air volume in the medium-temperature section is increased by 5%.

[0024] The PLC uses an incremental PID algorithm to close the loop and control the electric proportional control valve, with a correction cycle of 1 second;

[0025] Non-contact monitoring device: Magnetic encoders with a resolution of ±0.1mm are installed on both sides of the steel belt, with a sampling frequency of ≥100Hz. They can diagnose steel belt breakage (speed = 0), deviation (displacement > 5mm) and drive failure (acceleration > 3g) in real time, with a positioning accuracy of ±10cm.

[0026] Furthermore, the vortex counter-flow covers more than 90% of the cavity cross section, the nozzle spacing is 1.5-2 times the diameter, and the inclination angle is continuously adjusted by a servo motor according to the temperature gradient of the infrared thermal imaging in the high-temperature section.

[0027] Furthermore, the YOLOv5 model training data includes 1,000 sets of labeled samples, and the labeling standard is: slag layer thickness > 200 mm or slag block diameter > 100 mm is defined as slag blocking risk, and the warning lead time is ≥ 5 minutes.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. This invention solves long-standing technical problems of dry slag machines, such as low cooling efficiency, high risk of slag blockage, and short equipment life, through the collaborative innovation of gradient cooling structure design, intelligent control algorithm, and non-contact monitoring device.

[0030] 2. The present invention achieves a deep coupling of structure, algorithm and monitoring: gradient cooling provides a physical basis for intelligent control, the MPC algorithm uses the high-precision speed signal of the magnetic grid to achieve advanced control, and the output of the slag blockage warning model directly drives the air volume redistribution strategy, forming a closed-loop optimization;

[0031] 2. The present invention uses a three-level gradient cooling structure to match the slag phase change characteristics and combines it with the MPC dynamic compensation algorithm to reduce the maximum outlet temperature and improve temperature uniformity.

[0032] 3. The present invention adopts a differentiated air volume distribution strategy (vortex crushing in high temperature section + swirl delay in medium temperature section + equal flow in low temperature section) to enhance air flow penetration in high temperature section, reduce cooling air volume, and improve heat exchange adequacy in medium temperature section.

[0033] 4. Active suppression of slag blockage risk: This invention integrates dual-band infrared data through the improved YOLOv5 model (ResNet34+CBAM), which improves the accuracy of slag blockage warning and reduces the incidence of slag blockage;

[0034] 5. Sharp reduction in equipment failure rate: The present invention reduces steel belt-related failures and shortens fault location time through non-contact magnetic grid monitoring combined with fault diagnosis logic. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0036] Figure 2 This is a schematic diagram of the stepped cooling cavity structure of the present invention;

[0037] Figure 3 Schematic diagram of the intelligent air volume distribution device of the present invention;

[0038] Figure 4 This is a schematic diagram of the slag blocking and temperature early warning device of the present invention;

[0039] Figure 5 Schematic diagram of the non-contact operating condition monitoring device of the present invention;

[0040] Figure 6 Schematic diagram of the air volume adjustment process of Example 1 of the present invention;

[0041] In the figure: 1- double-layer staggered array of directional nozzles, 2- honeycomb guide plate, 3- porous plate flow layer, 4- data collection point, 5- slag bin, 6- non-contact encoder, 7- fan, 8- electric proportional control valve, 9- alarm device. DETAILED DESCRIPTION

[0042] The present invention will be further described below with reference to the embodiments and the accompanying drawings.

[0043] Example 1 A 660MW supercritical unit

[0044] Gradient cooling parameters

[0045] High temperature section: nozzle spacing is 150mm, and the inclination angle is continuously adjusted by a servo motor (the inclination angle increases by 10° when the temperature gradient is >50℃ / m);

[0046] Medium temperature section: guide plate inclination 45°, swirl nozzle coverage 80%;

[0047] Low temperature section: The porous plate is 12mm thick and is installed 200mm away from the steel belt.

[0048] Intelligent control process

[0049] Data acquisition: Dual-band infrared sensor 4 sampling at 10 Hz, resolution 1920 × 1080;

[0050] Slag blocking warning: When the slag layer thickness is greater than 200mm or the slag block diameter is greater than 100mm, the model outputs P risk ;

[0051] MPC compensation: solve the objective function J every 5 seconds and output the optimal Δu.

[0052] MPC weight coefficient λ value

[0053] Historical data regression analysis

[0054] Collected three months of operating data from a 660MW unit (including 12,000 sets of temperature-air volume control records)

[0055] Construct the cost function:

[0056] J(λ)=α∑|T set -T a |+β∑|Δu| (α=0.6, β=0.4 are operating weights)

[0057] Find the value of λ that minimizes J(λ)

[0058] Physical Constraint Verification

[0059]

[0060] Dynamic adjustment mechanism

[0061] When the slag flow rate is >5kg / s: λ is automatically reduced to 0.7 (enhanced temperature tracking);

[0062] When the steel belt acceleration is >1.5g: λ increases to 1.0 (suppressing mechanical shock).

[0063] Test results show that the ResNet34 fine-tuning parameter design improves the model's mAP by 7.1% in a vibration environment, and the dynamic assignment strategy of λ=0.8 extends the mechanical life of the damper to 2.3 times that of the original system (from 12 months to 28 months).

[0064] Troubleshooting

[0065] Magnetic grating data is transmitted to the PLC in real time. When the deviation is greater than 5mm, an audible and visual alarm is triggered. The positioning error is ±5cm.

[0066] Drive failure (acceleration > 3g) automatically cuts off the fan power supply.

[0067] Fine-tuning parameters for ResNet34 transfer learning

[0068] Transfer learning uses the pre-trained ResNet34 as the backbone and fine-tunes it for the slag flow morphology recognition task:

[0069] Network structure adjustment:

[0070] After removing the original classification layer and adding the Convolutional Block Attention Module (CBAM) to the C3 and C4 layers, the output layer is replaced with a 3-channel detection head (slag layer thickness / slag block contour / temperature distribution);

[0071] Fine-tuning parameter settings

[0072]

[0073] Migration Performance Comparison

[0074]

[0075] The test results of this embodiment and the traditional system are shown in the following table:

[0076]

Claims

1. A 660MW direct current boiler slag dryer system based on gradient cooling and intelligent control, characterized in that: include: The stepped multi-stage cooling chamber is divided into a high-temperature section (>700°C), a medium-temperature section (400-700°C) and a low-temperature section (<400°C) with mutually exclusive temperature ranges along the slag conveying direction; Differentiated cooling device: The high temperature section is provided with a double-layer staggered array directional nozzle (1) with dynamically adjustable inclination angles, the upper nozzle inclination angle being 30°-60° and the lower nozzle inclination angle being 15°-45° in the opposite direction, forming a vortex counter-flow with a swirl intensity ≥0.6; The medium temperature section is provided with a honeycomb guide plate (2) with an aperture of 50-100 mm and a swirl nozzle with a swirl angle of 40°-70°, which prolongs the slag particle retention time by 30%-40%; The low temperature section is provided with a porous plate flow-distributing layer (3) with an opening rate of 30%-50%, so that the standard deviation of the outlet temperature is ≤5°C; Intelligent control system: The dynamic compensation mechanism based on model predictive control (MPC) takes outlet temperature, slag flow rate and particle size as inputs, and the response time is less than 2 seconds to compensate the air volume set value. The objective function is: Min J=∑[Tset(t)-Tactual(t)] 2 +λ•Δu(t) 2 ; Where, J: objective function (performance index), which needs to be minimized; Tset(t): set temperature of the dry slag machine outlet (°C); Tactual(t): actual measured temperature at the outlet of the dry slag machine (°C); Δu(t): The adjustment range of the air volume control amount (i.e., the control increment); λ: Weight coefficient, used to balance the relative importance between temperature deviation and control action amplitude; The summation term ∑: usually represents the accumulation in the prediction time domain (under the MPC framework); Based on the improved YOLOv5 slag flow morphology recognition model of transfer learning, the Backbone network is replaced by ResNet34, and the slag layer thickness, slag block contour and temperature distribution data fused by the dual-band infrared sensor (4) are input to output the slag blocking risk probability; when the risk probability is >80%, the air volume in the high-temperature section is reduced by 10% per level, while the air volume in the medium-temperature section is increased by 5%; The PLC uses an incremental PID algorithm to close the loop and control the electric proportional control valve (8), with a correction period of 1 second; Non-contact monitoring device: Magnetic encoders (6) with a resolution of ±0.1mm are installed on both sides of the steel belt, with a sampling frequency of ≥100Hz, which can diagnose steel belt breakage (speed = 0), deviation (displacement > 5mm) and drive failure (acceleration > 3g) in real time, with a positioning accuracy of ±10cm.

2. A 660MW once-through boiler slag dryer system based on gradient cooling and intelligent control according to claim 1, characterized in that: The vortex counter-flow covers more than 90% of the cavity cross section, the nozzle spacing is 1.5-2 times the diameter, and the inclination angle is continuously adjusted by a servo motor according to the temperature gradient of the infrared thermal imaging in the high-temperature section.

3. The 660MW direct current boiler slag dryer system based on gradient cooling and intelligent control according to claim 1 is characterized in that: The YOLOv5 model training data contains 1000 sets of labeled samples. The labeling standard is: slag layer thickness > 200 mm or slag block diameter > 100 mm is defined as slag blocking risk, and the warning lead time is ≥ 5 minutes.

Images