Multi-objective optimization-based coal pulverizing system energy-saving adjustment control method and device

Abstract

The invention provides a coal pulverizing system energy-saving adjustment control method and device based on multi-objective optimization, and the method comprises the steps: obtaining the operation parameters of a coal pulverizing system, and carrying out the preprocessing; performing partition identification on the current working condition based on the expert rule base, and dynamically configuring a multi-target optimization weight matrix; carrying out multi-target optimization by utilizing an improved particle swarm optimization algorithm, and realizing balance between global search and local development by adaptively adjusting an inertia weight and a learning factor; an optimization result is input into an expert rule base for hard constraint verification and experience correction, and control parameters meeting safety and energy efficiency requirements are generated; the corrected control parameters are converted into control instructions, closed-loop feedback adjustment is executed, and a periodic optimization control process is formed. The multi-target collaborative optimization control of the coal pulverizing system under different load working conditions can be realized, the coal pulverizing unit consumption is remarkably reduced, the pulverized coal fineness stability is improved, meanwhile, the algorithm convergence speed and the fault early warning advance are improved, and the system safety and the dynamic balance of energy efficiency are guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of control technology, specifically relating to an energy-saving regulation and control method and device for a pulverizing system based on multi-objective optimization. Background Technology

[0002] As a core auxiliary system in coal-fired power plants, the pulverizing system accounts for 20% of the plant's total electricity consumption, directly impacting operating costs and equipment reliability. Existing control technologies for pulverizing systems can be categorized into traditional control, intelligent optimization, and hybrid architectures. Traditional methods primarily rely on PLC central automation systems and PID control; some companies employ closed-loop control, but its coverage is less than 20%. Intelligent optimization technologies include machine learning-based predictive control (such as gradient boosting tree-based condition prediction), fuzzy control (such as multi-dimensional fuzzy controllers adjusting feed rate and damper opening), and multi-objective optimization algorithms. A typical hybrid architecture is the NSGA-Ⅲ+DDPG architecture.

[0003] Existing pulverizing system control technologies have significant shortcomings in energy consumption optimization, dynamic adaptability, and fault early warning. In terms of energy consumption, the power consumption of the central storage steel ball mill pulverizing system accounts for 20% of the plant's total power consumption, and traditional control methods only reduce unit consumption by 3.2%. Regarding dynamic control, traditional PID control has limited accuracy, with mill outlet temperature fluctuations of ±3.5℃, a stabilization time of 32 seconds under varying load conditions, and a significant increase in unit consumption under low load. Optimization algorithms suffer from slow convergence speed and low solution quality; NSGA-II converges slowly, NSGA-III has high complexity, and computational efficiency is poor in large-scale scenarios. Fault early warning response is lagging, with vibration signal diagnosis having an advance lead of only 10 minutes. In terms of modeling, the pure data-driven model has a prediction error of 7.8% under varying coal types and poor interpretability. The evaluation system is limited to a single energy efficiency index, failing to cover dimensions such as reliability and environmental impact; multi-objective optimization is often simplified to a weighted sum, unable to adapt to dynamic changes in operating conditions. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to propose an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization.

[0006] The second objective of this invention is to propose an energy-saving regulation and control device for a pulverizing system based on multi-objective optimization.

[0007] The third objective of this invention is to provide a computer device.

[0008] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0009] To achieve the above objectives, a first aspect of the present invention proposes an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization, comprising: S1. Collect and preprocess the operating parameters of the coal mill. The operating parameters include the mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. S2, based on the operating parameters, identify the current operating condition range, and configure multi-objective optimization weights according to the preset expert rule base, wherein the multi-objective optimization weights correspond to the priority relationship between energy consumption, stability and response speed; S3, The improved particle swarm optimization algorithm is used to dynamically optimize the running parameters. The improved particle swarm optimization algorithm generates preliminary optimized control parameters under preset constraints through a linear decreasing inertial weight strategy and an adaptive adjustment mechanism of the learning factor. S4, input the preliminary optimized control parameters into the expert rule base for hard constraint verification and empirical correction. The hard constraint verification includes that the primary air volume is not lower than the minimum fluidizing air volume and the mill outlet temperature is not higher than the preset threshold. The empirical correction includes triggering an advance adjustment strategy based on the load change rate, outputting the final control command and feeding it back to the actuator to adjust the operating status of the coal mill.

[0010] In one embodiment of the present invention, S1 includes: S11 collects seven core parameters at a sampling interval of 1 minute: mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. S12, the preprocessing stage adopts filtering and noise reduction technology, and ensures that the data length covers at least one week of the full working cycle, including the data distribution characteristics of typical high, medium and low load scenarios.

[0011] In one embodiment of the present invention, S2 includes: S21 divides the unit load into high load range, medium load range and low load range; S22, automatically match the target weight matrix according to the load range, where the stability weight is 0.6 for the high load range, the energy consumption weight is 0.5 for the medium load range, and the response speed weight is 0.5 for the low load range.

[0012] In one embodiment of the present invention, S3 includes: S31 employs a linearly decreasing inertia weight strategy, according to the formula... Calculate the current inertia weight, where , , ; S32, when the fitness of a particle has not improved for 10 consecutive generations, adjust the cognitive factor to 1.5 and the social factor to 2.5 to enhance group information sharing.

[0013] In one embodiment of the present invention, S4 includes: S41, during hard constraint verification, ensure that the primary air volume is not less than 60% of the rated value and the mill outlet temperature is not higher than 75℃; S42, in empirical correction when the load change rate When this happens, an advance adjustment strategy is triggered, and the coal feed rate is pre-adjusted by +5% to avoid the risk of coal blockage.

[0014] In one embodiment of the present invention, the method further includes: S5. Based on local features of the coal mill's operating images, fault diagnosis information is extracted. These local features include shape, texture, and particle size distribution. Seven typical fault types, including accumulation and blockage, insufficient grinding, and dust backflow, are identified using an expert rule base. The fault warning lead time is increased to [missing information]. .

[0015] To achieve the above objectives, a second aspect of the present invention provides an energy-saving regulation and control device for a pulverizing system based on multi-objective optimization, comprising: The operating parameter acquisition and preprocessing module is used to acquire and preprocess the operating parameters of the coal mill. The operating parameters include mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. The working condition identification and weight configuration module identifies the current working condition range based on the operating parameters and configures multi-objective optimization weights according to a preset expert rule base. The multi-objective optimization weights correspond to the priority relationship between energy consumption, stability and response speed. The dynamic optimization module is used to dynamically optimize the running parameters using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm generates preliminary optimized control parameters under preset constraints through a linear decreasing inertial weight strategy and an adaptive adjustment mechanism for the learning factor. The hard constraint verification and experience correction module is used to input the preliminary optimized control parameters into the expert rule base for hard constraint verification and experience correction. The hard constraint verification includes ensuring that the primary air volume is not lower than the minimum fluidizing air volume and that the mill outlet temperature is not higher than a preset threshold. The experience correction includes triggering an advance adjustment strategy based on the load change rate, outputting the final control command, and feeding it back to the actuator to adjust the operating status of the coal mill.

[0016] This invention discloses an energy-saving regulation and control method and device for a pulverizing system based on multi-objective optimization. The method addresses four core technical challenges of pulverizing systems through multi-objective optimization: 1. Energy consumption optimization: While ensuring long-term reliable operation, the reduction rate of pulverizing unit consumption is increased to ≥8%, overcoming the limited reduction effect of existing technologies; 2. Control accuracy: The fluctuation of the coal mill outlet temperature is controlled within ≤±2℃, and the stabilization time under variable load conditions is shortened to ≤22s, overcoming the shortcomings of slow response and insufficient accuracy in traditional control; 3. Algorithm limitations: An improved PSO (Particle Swarm Optimization) algorithm is integrated with an expert rule base to solve the problems of slow convergence and local optima in single intelligent algorithms, achieving globally optimal control under complex operating conditions; 4. System synergy: Fault diagnosis and energy-saving control are integrated, and the fault warning lead time is increased to ≥15min through multi-dimensional feature extraction, ensuring coordinated optimization of safety and energy efficiency.

[0017] To achieve the above objectives, a third aspect of this application provides a computer device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization as described in the first aspect embodiment.

[0018] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization as described in the first aspect embodiment.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization according to an embodiment of the present invention; Figure 2 This is an architecture diagram of an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization according to an embodiment of the present invention; Figure 3 This is a structural diagram of an energy-saving adjustment and control device for a pulverizing system based on multi-objective optimization according to an embodiment of the present invention; Figure 4 It is a computer device according to an embodiment of the present invention. Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0023] The following description, with reference to the accompanying drawings, describes an energy-saving regulation and control method and apparatus for a pulverizing system based on multi-objective optimization, according to an embodiment of the present invention.

[0024] Example 1 Figure 1 This is a flowchart of an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization according to an embodiment of the present invention, such as... Figure 1 As shown, it includes: S1. Collect and preprocess the operating parameters of the coal mill. The operating parameters include the mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature.

[0025] In some implementations, operating parameters include seven key measurement points: mill outlet temperature, primary air volume, coal feed rate, mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. These parameters are acquired in real time via a DCS (Distributed Control System), with a sampling frequency of once per minute. The data acquisition cycle is recommended to cover at least one week to ensure coverage of typical operating conditions such as high, medium, and low loads. The acquired data must have timestamps and data integrity to meet the real-time and reliability requirements of industrial control systems.

[0026] Furthermore, the preprocessing stage includes data filtering, outlier removal, and standardization. Filtering employs moving averages or low-pass filters (such as Butterworth filters with a cutoff frequency of 0.1Hz) to eliminate the interference of transient noise on condition assessment. Outlier removal is based on the 3σ principle or the IQR (interquartile range) method to ensure that the data distribution conforms to normal or approximately normal characteristics. Standardization typically uses the Z-score normalization method, i.e. ,in These are the original parameter values. This is the historical average. The standard deviation is used to eliminate differences in the dimensions of different parameters and improve the convergence efficiency of subsequent optimization algorithms.

[0027] Furthermore, S1 includes: S11 collects seven core parameters at a sampling interval of 1 minute: mill outlet temperature, primary air volume, coal feed rate, mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature.

[0028] In some implementations, this step relies on a distributed control system (DCS) or a supervisory control and data acquisition system (SCADA) to collect data through a sensor network deployed at key nodes of the coal mill. Specifically, the mill outlet temperature is measured by thermocouples or infrared thermometers, typically with an accuracy of ±0.5℃; the primary air volume is calculated by an air velocity sensor combined with the cross-sectional area of ​​the duct, with a range of 15,000–30,000 m³ / h and a sampling frequency of 1Hz; the coal feed rate is acquired in real time via a weighing belt or mass flow meter, with an accuracy of ±1%; the mill current is collected by a current transformer, with a range of 0–500A; the outlet pressure, inlet primary air pressure, and inlet primary air temperature are collected by a pressure transmitter and a temperature sensor, respectively, with accuracies of ±0.1 kPa and ±0.5℃. All data is uploaded to the control platform via OPC, Modbus, or industrial Ethernet protocols to ensure the real-time performance and reliability of data transmission.

[0029] Specifically, the sampling interval was set to 1 minute, which meets the requirements of industrial control systems for balancing dynamic response and data storage efficiency. This time window can effectively capture the changing trends of the coal mill's operating status while avoiding data redundancy. The seven collected parameters cover the thermodynamic, mechanical, and hydrodynamic characteristics of the pulverizing system and are key input variables for constructing the multi-objective optimization model and expert rule base.

[0030] S12, the preprocessing stage adopts filtering and noise reduction technology, and ensures that the data length covers at least one week of the full working cycle, including the data distribution characteristics of typical high, medium and low load scenarios.

[0031] In some implementations, the data acquisition module obtains seven core operating parameters in real time from the coal mill via the DCS system: outlet temperature, primary air volume, coal feed rate, outlet pressure, inlet primary air pressure, inlet primary air temperature, and inlet primary air volume. The sampling frequency is once per minute. The raw data typically contains high-frequency noise and transient outliers; therefore, a low-pass filter (such as a Butterworth filter) is used for noise reduction, with the cutoff frequency set to [value missing]. This preserves the low-frequency trends of operating condition changes while suppressing noise caused by sensor drift or transient disturbances. Additionally, moving average filtering or wavelet threshold denoising can be optionally introduced as supplementary methods to further improve data smoothness.

[0032] Specifically, the preprocessing stage requires data length to cover at least a one-week full operating cycle, ensuring it includes typical operating scenarios such as high load (>80% MCR), medium load (50%-80% MCR), and low load (<50% MCR). The dataset must possess representative load distribution characteristics; for example, the coal feed rate range under high load conditions is... The primary air volume is The outlet temperature is This is to support the adaptive optimization of the PSO algorithm under different operating conditions in subsequent improvements. At the same time, the data must meet the requirements of continuity and completeness, with a missing rate not exceeding 5%, to avoid bias in the optimization model due to incomplete data.

[0033] S2, based on the operating parameters, identify the current operating condition range, and configure multi-objective optimization weights according to the preset expert rule base, wherein the multi-objective optimization weights correspond to the priority relationship of energy consumption, stability and response speed.

[0034] In some implementations, this step involves real-time collection of seven core indicators from the coal mill, including mill outlet temperature, primary air volume, coal feed rate, outlet pressure, inlet primary air temperature and volume, and mill current. These indicators, combined with the unit load rate, are used to classify and identify the current operating status. Specifically, the system divides the unit load into high-load (>80% MCR), medium-load (50%-80% MCR), and low-load (<50% MCR) ranges, and assigns different multi-objective optimization weight matrices to each range to reflect the priority relationship between energy consumption, stability, and response speed.

[0035] Specifically, in the high-load range, stability has a weight of 0.6, energy consumption has a weight of 0.3, and response speed has a weight of 0.1; in the medium-load range, energy consumption has a weight of 0.5, stability has a weight of 0.3, and response speed has a weight of 0.2; and in the low-load range, response speed has a weight of 0.6, energy consumption has a weight of 0.3, and stability has a weight of 0.1. These weightings are based on a fuzzy comprehensive evaluation model and industrial experience, ensuring that the system can prioritize meeting key performance indicators under different load conditions. For example, during high-load operation, the system prioritizes keeping the fluctuation of the coal mill outlet temperature within ±2℃ to avoid overheating of the equipment or insufficient drying of the pulverized coal due to excessively high temperatures.

[0036] Furthermore, S2 includes: S21, S21, divides the unit load into a high load range (>80% MCR), a medium load range (50%-80% MCR), and a low load range (<50% MCR).

[0037] In some implementations, this load zoning method uses real-time acquired unit load signals as the basis for judgment, combined with seven key operating parameters in the DCS system, including pulverizer outlet temperature, primary air volume, and coal feed rate, for auxiliary verification. In specific operation, the system first obtains the current load value through the data acquisition module, and then classifies it according to preset interval division standards. The high load interval is defined as a load greater than 80% MCR, at which point the system operates close to its rated condition, equipment output demand is high, and operational stability must be prioritized; the medium load interval is 50%-80% MCR, the system is in an economical operating window, and the optimization objective focuses on minimizing energy consumption; the low load interval is less than 50% MCR, at which point the system needs to respond quickly to load changes to avoid coal dust accumulation and equipment idling, making response speed the primary optimization objective.

[0038] Specifically, the load partitioning thresholds strictly adhere to industry standards and typical unit operating characteristics. For example, the high-load range is set as >80% MCR, corresponding to mill outlet temperature fluctuations controlled within ±2℃, with a stability weight of 0.6; the medium-load range is set as 50%-80% MCR, with an energy consumption weight of 0.5 and a response speed weight of 0.2; and the low-load range is set as <50% MCR, with a response speed weight of 0.6 and an energy consumption weight of 0.3. This partitioning strategy ensures a reasonable allocation of control objectives under different operating conditions, thereby improving overall control efficiency.

[0039] S22, automatically match the target weight matrix according to the load range, where the stability weight is 0.6 for the high load range, the energy consumption weight is 0.5 for the medium load range, and the response speed weight is 0.5 for the low load range.

[0040] In terms of specific operation, the system first collects the current boiler load value through the DCS system, and then identifies the operating condition by combining it with seven key operating parameters, including the coal mill outlet temperature, primary air volume, and coal feed rate. The identification logic is based on preset partitioning standards in the expert rule base. For example, when the load is higher than 80%, the system determines it to be a high load range. At this time, the optimization objective priority is stability > energy consumption > response speed. In the corresponding objective weight matrix, the stability weight is set to 0.6, and the energy consumption and response speed weights are 0.3 and 0.1, respectively. In the medium load range (50%-80%), the energy consumption weight increases to 0.5, the stability weight is 0.3, and the response speed weight is 0.2; while in the low load range (<50%), the response speed weight is set to 0.5, and the energy consumption and stability weights are 0.3 and 0.2, respectively. This weight matrix serves as the input parameter for improving the PSO algorithm and is used to construct the multi-objective fitness function.

[0041] Furthermore, the weight allocation in this step must meet the constraints of the NSGA-II multi-objective optimization model, such as controlling the mill outlet temperature at 60-80℃ and the primary air pressure at no less than 5kPa. The weight adjustment mechanism ensures that the system can prioritize responses to key performance indicators under different load conditions, avoiding suboptimal solutions caused by single-objective optimization.

[0042] S3, The improved particle swarm optimization algorithm is used to dynamically optimize the operating parameters. The improved particle swarm optimization algorithm generates preliminary optimized control parameters under preset constraints through a linear decreasing inertia weight strategy and an adaptive adjustment mechanism of the learning factor.

[0043] In some implementations, the inertia weight w = w in the improved PSO algorithm init -{(w init -w end The linear decrease is achieved by )×t} / T, where , , This represents the total number of iterations. This represents the current iteration number. This strategy maintains high global search capability in the early stages of the algorithm, helping to escape local optima, while gradually enhancing local exploitation capability in later stages to improve convergence accuracy. Optionally, a cognitive factor... Social factors As a fundamental parameter, furthermore, when the particle's fitness fails to improve for 10 consecutive generations, the algorithm automatically switches to... , This is to enhance information sharing among groups and improve search efficiency.

[0044] Regarding parameter constraints, the upper limit of velocity v max =0.3 × parameter range, where the parameter range is based on key variables of the pulverizing system (such as coal feed rate). Primary air volume Grinding pressure Dynamic settings ensure that particles are searched within the feasible region. The objective function is... ,in This indicates the unit consumption for milling (kWh / t). The fineness of pulverized coal (%) is the decision variable. [Coal feed rate, primary air volume, grinding pressure], constraints include mill outlet temperature. Primary wind pressure .

[0045] Furthermore, S3 includes: S31, using a linear decreasing strategy for inertia weight w = w init -{(w init -w end )×t} / T, where , , .

[0046] In the specific implementation, the dynamic adjustment mechanism of the inertia weight is embedded in the PSO iteration process. The particle velocity update formula is as follows: ,in This is calculated using the linear decreasing formula described above. (By...) The algorithm gradually decreases from 0.9 to 0.4, maintaining strong exploratory capabilities in the first 20 iterations. Subsequently, it gradually transitions to a more refined local search stage, thereby achieving a better combination of control parameters under complex working conditions.

[0047] S32, when the particle's fitness fails to improve for 10 consecutive generations, the cognitive factor will be... Adjusted to 1.5, social factors Adjusted to 2.5 to enhance group information sharing.

[0048] At the technical implementation level, the particle update formula is:

[0049]

[0050] in, For the first The particle in the first During the nth iteration Dimensional speed, For particle position, For the individual's historical best position, To be the globally optimal position , A random number independent of the interval [0,1]. Cognitive factor. The tendency of particles to learn towards their individual optimality, social factors. Control the tendency of particles to learn towards the swarm optimum. Increase the [unclear - possibly "increase" or "enhance"] when a particle's fitness has not been updated for a long time. It can enhance information sharing among groups, making particles more inclined to follow the global optimal solution, thereby escaping the local optimal region and improving the algorithm's exploration capability.

[0051] Specifically, this adjustment strategy is applicable to optimization scenarios with a particle swarm size of 30 and 50 iterations. Velocity constraint v max =0.3 × parameter range, ensuring that particles do not go out of bounds during the search process. The adjustment thresholds for cognitive and social factors are the fitness that has not improved for 10 consecutive generations. This threshold is set based on the convergence characteristics of the control parameters of the milling system, taking into account both algorithm stability and optimization efficiency.

[0052] S4, input the preliminary optimized control parameters into the expert rule base for hard constraint verification and empirical correction. The hard constraint verification includes that the primary air volume is not lower than the minimum fluidizing air volume and the mill outlet temperature is not higher than the preset threshold. The empirical correction includes triggering an advance adjustment strategy based on the load change rate, outputting the final control command and feeding it back to the actuator to adjust the operating status of the coal mill.

[0053] In some implementations, the initial control parameters (such as coal feed rate, primary air volume, and grinding pressure) output by the improved PSO algorithm are first subjected to hard constraint verification. The core of hard constraint verification is to ensure that the system operates within physical and safety boundaries. For example, the primary air volume must not be lower than the minimum fluidizing air volume (e.g., set to 60% of the rated value) to maintain sufficient fluidization and conveying of pulverized coal; at the same time, the pulverizer outlet temperature must not exceed a preset threshold (e.g., 75℃) to prevent pulverized coal coking or equipment overheating and damage. These constraints are verified in real time through the logical judgment module in the expert rule base. If the parameters violate the constraints, a local re-optimization mechanism is triggered to adjust the particle positions and recalculate.

[0054] Specifically, after the hard constraint verification passes, the system further executes an experience-based correction strategy. This experience-based correction dynamically responds based on the load change rate (in % / min). When the load change rate exceeds 2% / min, the system automatically triggers an "advance adjustment coefficient," which pre-adjusts the coal feed rate by +5% based on the optimization results to proactively address the risks of coal powder accumulation or insufficient drying. This strategy effectively improves the system's dynamic adaptability under varying load conditions and shortens the stabilization time of the coal mill outlet temperature to ≤22s, superior to the 32s of traditional PID control.

[0055] Furthermore, S4 includes: S41, during the hard constraint verification, ensure that the primary air volume is not less than 60% of the rated value and the mill outlet temperature is not higher than 75℃.

[0056] In some implementations, the system first receives preliminary control parameters from the improved PSO algorithm, including key variables such as primary air volume and mill outlet temperature. Then, the system performs item-by-item verification based on preset hard constraints. For example, the primary air volume must meet certain conditions. ,in Rated primary air volume of coal mill (unit: If the actual optimized value is lower than this threshold, the system will automatically adjust the particle position and re-perform local optimization to ensure that the air volume is maintained within the safe operating range. Simultaneously, the coal mill outlet temperature must meet the following requirements. If the optimization result exceeds the upper limit, the system will trigger the temperature correction mechanism, which will adjust the hot air damper opening or reduce the coal feeding rate to bring the outlet temperature back to a safe range.

[0057] Specifically, the primary air volume constraint in this step is set based on the fluidization characteristics of the coal mill and the pulverized coal drying requirements. 60% of the rated value is the minimum requirement to ensure the stability of pulverized coal conveying and combustion. The upper limit of 75℃ for the mill outlet temperature is set based on the risk of pulverized coal deflagration and the temperature resistance performance of the equipment to avoid pulverized coal coking or pipeline blockage due to excessive temperature. These parameters all comply with the safety control standards for primary air volume and outlet temperature in the "Operating Procedures for Pulverizing Systems of Thermal Power Plants" (DL / T 514-2012).

[0058] S42, in empirical correction when the load change rate When this happens, an advance adjustment strategy is triggered, and the coal feed rate is pre-adjusted by +5% to avoid the risk of coal blockage.

[0059] From a technical implementation perspective, the calculation of the load change rate is based on the unit load data collected by the DCS system, with a sampling period of 1 minute. The calculation formula is as follows: Where L(t) represents the current unit load percentage, t = 1 min. When When L>2% / min, the system determines it to be a rapidly increasing load condition. At this time, the "advance adjustment coefficient" mechanism will be activated to pre-adjust the coal feed rate by +5%, that is: ,in To improve the baseline value of coal feed obtained by the PSO algorithm optimization, This is the corrected actual control value.

[0060] At the parameter level, the trigger threshold of this strategy This setting is based on the dynamic response characteristics of the coal mill and the physical limitations of the pulverized coal conveying system, ensuring timely adjustment of fuel supply before sudden load changes. The +5% correction margin is the optimal pre-adjustment ratio verified through historical operating data and simulation models, which can avoid coal blockage without significantly increasing the energy consumption burden of the coal mill.

[0061] S5. Based on local features of the coal mill's operating images, fault diagnosis information is extracted. These local features include shape, texture, and particle size distribution. Seven typical fault types, including accumulation and blockage, insufficient grinding, and dust backflow, are identified using an expert rule base. The fault warning lead time is increased to [missing information]. .

[0062] In some implementations, image acquisition typically involves using industrial cameras or infrared imaging devices to acquire real-time image data of the coal powder flow state inside the coal mill. Image preprocessing includes operations such as grayscale conversion, histogram equalization, and noise filtering (e.g., median filtering or Gaussian filtering) to enhance image contrast and feature clarity. Subsequently, edge detection algorithms (e.g., Canny or Sobel) are used to extract the shape contour of the coal powder accumulation area. Texture analysis methods (e.g., Gray-Level Co-occurrence Matrix, GLCM) are used to quantify the surface roughness and arrangement of coal powder particles. Simultaneously, image segmentation and particle size statistics techniques (e.g., particle size distribution analysis based on K-means clustering) are employed to obtain the coal powder particle size distribution characteristics. These features serve as input vectors for fault diagnosis, used to match fault patterns in an expert rule base.

[0063] Specifically, it is recommended to set the image acquisition frequency to 1 frame per minute to ensure the ability to capture dynamic changes. In texture analysis, the window size of GLCM is typically [size missing]. or The number of gray levels is set to To balance computational efficiency and feature accuracy, particle size classification in particle size distribution analysis can be achieved using... The intervals are divided, and the proportion of particles in each interval is statistically analyzed to form a particle size distribution histogram. The response time for fault diagnosis is controlled within... Early warning can be achieved by matching image features with historical fault samples.

[0064] An energy-saving regulation and control method for a pulverizing system based on multi-objective optimization according to an embodiment of the present invention can realize multi-objective collaborative optimization control of the pulverizing system under different load conditions, significantly reduce pulverizing unit consumption and improve the stability of coal powder fineness, while improving the algorithm convergence speed and fault warning advance, and ensuring the dynamic balance between system safety and energy efficiency.

[0065] Example 2 The following describes in detail an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization, according to an embodiment of the present invention, with reference to the accompanying drawings.

[0066] The core technical solution of this invention is to use an improved particle swarm optimization (PSO) combined with an expert rule base, and through the innovative combination of dynamic parameter adjustment and operating condition zoning rules, to achieve energy-saving regulation and control of the pulverizing system. The following is an introduction to the corresponding modules of this invention: 1. Improved PSO algorithm module: The improved PSO algorithm lies in dynamic parameter design to resolve the conflict between energy consumption and pulverized coal quality in the pulverizing system. Specifically, the inertia weight w is initially 0.9 and linearly decreases to 0.4, the learning factors c1=2.0 and c2=2.0 (adaptively adjusted), and the velocity constraint v... max=0.3 × parameter range. This algorithm introduces the concept of Pareto optimal solution and performs multi-objective optimization through the NSGA-II algorithm. It can dynamically adjust parameters according to the operating condition range and output optimized control commands.

[0067] 2. Expert Rule Base Construction Module: The expert rule base is built upon a fuzzy comprehensive evaluation model and industrial practice experience, and includes detection parameters, operating condition zoning standards, weight values, and correction thresholds. Specific rules are as follows: If the boiler load is >80%, then the optimization priority is: stability > energy consumption > response speed; If the boiler load is 50%-80%, then the optimization priorities are: energy consumption > stability > response speed. If the optimization priorities are: response speed > energy consumption > stability.

[0068] The rule base covers key parameters such as coal mill current, outlet temperature, and primary air volume. For example, it prioritizes stability under high load, focuses on energy consumption optimization under medium load, and sets correction mechanisms such as mill outlet temperature threshold.

[0069] 3. Collaborative Integration Process: The two components combine to form a closed-loop control process: the data acquisition module obtains seven key parameters, including the mill current and instantaneous coal feed rate; the operating condition identification module divides the operating condition regions based on an expert rule base; the improved PSO algorithm dynamically adjusts parameters according to the regions; the output optimized parameters are verified by the rule base (such as mill outlet temperature threshold correction); and finally, the execution module applies the optimized parameters to the equipment. The specific steps are: 1) Initialize the particle swarm (30 particles, 50 iterations); 2) Determine the target weight based on the current operating condition using expert rules; 3) Obtain preliminary control parameters through PSO optimization; 4) Verify and correct the parameters using the expert rule base; 5) Output the final control command.

[0070] like Figure 2 As shown, based on the above modules, the embodiments of the present invention have the following implementation methods.

[0071] S10. Specific design of the improved PSO algorithm: To achieve multi-objective optimization control of the pulverizing system, the improved PSO algorithm design needs to balance optimization accuracy and engineering feasibility. The core parameter configuration and optimization strategy are as follows: S101. Dynamic parameter adaptive adjustment mechanism: Inertia weight (w): Employs a linear decreasing strategy to balance global exploration and local exploitation capabilities. The initial value is set to 0.9, and it decreases according to the formula (w = w) during iteration. init -{(w init -w endThe value of )×t} / T decreases to 0.4, where (t) is the current iteration number and (T=50) is the total number of iterations. This design maintains strong global search capability in the early stages of iteration, and focuses on fine-grained local optimization in the later stages.

[0072] Learning Factors and Velocity Constraints: Under the basic configuration of cognitive factors (c1=2.0) and social factors (c2=2.0), an adaptive adjustment mechanism is introduced: when the fitness of a particle does not improve for 10 consecutive generations, it switches to (c1=1.5) and (c2=2.5) to enhance information sharing among the group. The velocity constraint is set to (vmax=0.3×parameter range) to prevent particles from leaving the feasible region due to excessive velocity. The parameter range is dynamically set for key variables of the pulverizing system (coal feed rate 20-50t / h, primary air volume 15000-30000 m³ / h, grinding pressure 8-12MPa).

[0073] S102. Multi-objective optimization framework: By introducing the Pareto optimal solution set and the NSGA-II algorithm, a bi-objective optimization model is constructed: Objective function: (min f (X) = [E, R90]), where (E) is the coal pulverization unit consumption (kWh / t), and (R90) is the coal powder fineness (%). Decision variables: (X = [coal feed rate, primary air volume, grinding pressure]); Constraints: Mill outlet temperature 60-80℃, primary air pressure ≥5 kPa.

[0074] S103. Population initialization strategy: Thirty initial particles were generated using Latin hypercube sampling to ensure uniform coverage of the parameter space. The particle position boundaries were set according to the actual operating range of the pulverizing system. For example, the coal feed rate was initialized to a random value of [25, 45] t / h, and the grinding pressure was initialized to a random number in the range of [9, 11] MPa to avoid getting trapped in local optima.

[0075] Through the above improvements, the algorithm achieved a reduction of pulverizing unit consumption of 4.2 kWh / t and stable control of coal powder fineness R90 at 8±0.5% in the 600 MW unit pulverizing system test, verifying the engineering effectiveness of the parameter design.

[0076] S20. Construction of Expert Rule Base: The expert rule base is built upon industrial practical experience, integrating multi-dimensional rules such as testing parameters, operating condition zoning, control strategies, and fault diagnosis to form a systematic decision support system. The core of the rule base includes the following four key clauses.

[0077] S201. Detection parameter system: The rule base explicitly includes seven core detection parameters, such as coal mill current (A), instantaneous coal feed rate (t / h), outlet air-coal mixture temperature (°C), outlet pressure (kPa), inlet primary air pressure (kPa), inlet primary air temperature (°C), and inlet primary air volume (m³ / h), as the basis for judging operating conditions and controlling and adjusting them.

[0078] S202. Operating Condition Zoning and Target Priority Rules: Based on the unit load, the operating conditions are divided into three intervals, and differentiated optimization objectives are matched accordingly: High load area (>80% of rated load): Prioritize system stability to avoid equipment failure. The optimization priority is stability > energy consumption > response speed, with a stability weight of 0.6. Medium load zone (50%-80% of rated load): Focus on energy consumption optimization, with priority adjusted to energy consumption > stability > response speed, and energy consumption weight increased to 0.5; Low load area (<50% of rated load): Response speed is the primary objective, with the priority order being response speed > energy consumption > stability.

[0079] S203. Control Parameter Correction Rules: Based on fuzzy control experience, key parameter correction thresholds are set for different load ranges. For example, the upper limit of hot air damper opening is constrained to 85% under high load to balance drying output and system safety margin. In terms of fuel management, the rule base needs to cover the combustion characteristic parameters, mixing ratios, and calorific value compensation strategies of various alternative fuels such as waste oil, animal powder, and plastics to ensure constant energy input. The number of coal mills in operation is determined based on the boiler coal consumption and the rated coal capacity of a single mill, and the difference in the number of layers of the burner walls of the coal mills in operation should not exceed 2 layers. At the same time, at least one permanent magnet motor direct-drive coal mill must be ensured to operate as a regulating mill to achieve flexible coal allocation.

[0080] S204 Fault Diagnosis Rules: Based on the local features (shape, texture, particle size distribution, etc.) of coal mill operation images, a fault type judgment rule is established, which can identify 7 typical faults such as dust accumulation and blockage, insufficient crushing, and dust backflow, providing decision support for abnormal operating conditions.

[0081] To visually represent the weight allocation logic, the optimization target weights for each load interval are summarized in Table 1: Table 1

[0082] S30. Improve the process of integrating PSO with expert rule base: The improved integration process of PSO (Power Sorting System) and expert rule base achieves dynamic energy-saving adjustment of the pulverizing system through the organic synergy of data acquisition, operating condition identification, intelligent optimization, rule verification, and closed-loop control. This process, with "data-driven, rule-guided, algorithm optimization, and closed-loop feedback" as its core logic, involves the following specific steps.

[0083] S301 Data Acquisition and Preprocessing: Key operating parameters of the coal mill are acquired in real time through the DCS system, including seven key measuring points such as mill outlet temperature, primary air volume, and coal feed rate. The sampling interval is 1 minute, and the data length must cover at least one week of the entire operating cycle to ensure that the data distribution characteristics of typical scenarios such as high, medium, and low loads are included. The raw data is filtered and denoised before being used as the basic input for operating condition identification and optimization calculations.

[0084] S302 Operating Condition Identification and Target Weight Configuration: Based on the expert rule base's zoning criteria, the unit load is divided into three zones: high (>80% MCR), medium (50%~80% MCR), and low (<50% MCR). The system automatically matches the target weight matrix for the corresponding zone according to the current load. For example, under high load, priority is given to ensuring pulverizing output (weight 0.6), while under low load, the focus is on optimizing unit consumption indicators (weight 0.5), achieving multi-objective optimization and adaptive adjustment of operating conditions.

[0085] S303 Dynamic Parameter PSO Optimization: Using the current operating parameters as the initial particle swarm (30 particles, 50 iterations), an improved PSO algorithm is employed for multi-objective optimization. The algorithm dynamically adjusts the inertia weight and learning factor, outputting preliminary optimized values ​​for coal feed rate, primary air volume, and grinding pressure under the constraints of the objective function (comprehensively considering pulverizing power consumption, output, and stability). Compared to traditional PSO, the dynamic parameter mechanism enhances the algorithm's global search capability under varying operating conditions.

[0086] S304 Expert Rule Validation and Parameter Correction: Optimized parameters must undergo dual verification by the expert rule base: Hard constraint verification: Ensure that the primary air volume is greater than or equal to the minimum fluidizing air volume (e.g., 60% of the rated value) and the mill outlet temperature is less than or equal to 75°C, etc. Experience-based correction rules: For example, during variable load processes, when the load change rate > 2% / min, the "advanced adjustment coefficient" (+5% coal feed pre-adjustment) is automatically triggered to avoid the risk of coal mill blockage. If the optimized value exceeds the rule range, the system will readjust the particle position and initiate local optimization.

[0087] S305 control output and feedback iteration: The optimized parameters that pass the verification are converted into control commands and sent to the actuators to adjust the valve opening and motor frequency. The system repeats the above process every 5 minutes, forming a closed-loop control to ensure that the pulverizing system continues to operate within the optimal range under disturbances such as load fluctuations and changes in coal quality.

[0088] Example 3 To achieve the above embodiments, such as Figure 3 As shown, this embodiment also provides an energy-saving adjustment and control device 10 for a pulverizing system based on multi-objective optimization. The device 10 includes an operating parameter acquisition and preprocessing module 100, an operating condition identification and weight configuration module 200, a dynamic optimization module 300, and a hard constraint verification and experience correction module 400.

[0089] The operating parameter acquisition and preprocessing module 100 is used to acquire and preprocess the operating parameters of the coal mill. The operating parameters include mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. The working condition identification and weight configuration module 200 identifies the current working condition range based on the operating parameters and configures multi-objective optimization weights according to a preset expert rule base. The multi-objective optimization weights correspond to the priority relationship between energy consumption, stability and response speed. The dynamic optimization module 300 is used to dynamically optimize the running parameters using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm generates preliminary optimized control parameters under preset constraints through a linear decreasing inertial weight strategy and an adaptive adjustment mechanism for the learning factor. The hard constraint verification and experience correction module 400 is used to input the preliminary optimized control parameters into the expert rule base for hard constraint verification and experience correction. The hard constraint verification includes that the primary air volume is not lower than the minimum fluidizing air volume and the mill outlet temperature is not higher than a preset threshold. The experience correction includes triggering an advance adjustment strategy based on the load change rate, outputting the final control command and feeding it back to the actuator to adjust the operating status of the coal mill.

[0090] Furthermore, the aforementioned operating parameter acquisition and preprocessing module 100 is also used for: Seven core parameters were collected at a sampling interval of 1 minute: mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. The preprocessing stage employs filtering and noise reduction techniques and ensures that the data length covers at least one week of the full operating cycle, including the data distribution characteristics of typical high, medium, and low load scenarios.

[0091] Furthermore, the aforementioned working condition identification and weight configuration module 200 is also used for: The unit load is divided into high load range, medium load range and low load range; The target weight matrix is ​​automatically matched based on the load range, where the stability weight is 0.6 for the high load range, the energy consumption weight is 0.5 for the medium load range, and the response speed weight is 0.5 for the low load range.

[0092] Furthermore, the aforementioned dynamic optimization module 300 is also used for: Using a linear decreasing inertia weight strategy, according to the formula Calculate the current inertia weight, where , , ; If a particle fails to improve its fitness for 10 consecutive generations, adjust the cognitive factor to 1.5 and the social factor to 2.5 to enhance group information sharing.

[0093] Furthermore, the aforementioned hard constraint verification and experience correction module 400 is also used for: During hard constraint verification, ensure that the primary air volume is not less than 60% of the rated value and the mill outlet temperature is not higher than 75℃; In empirical correction when the load change rate When this happens, an advance adjustment strategy is triggered, and the coal feed rate is pre-adjusted by +5% to avoid the risk of coal blockage.

[0094] An energy-saving adjustment and control device for a pulverizing system based on multi-objective optimization according to an embodiment of the present invention can realize multi-objective collaborative optimization control of the pulverizing system under different load conditions, significantly reduce pulverizing unit consumption and improve the stability of coal powder fineness, while improving the algorithm convergence speed and fault warning advance, and ensuring the dynamic balance between system safety and energy efficiency.

[0095] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 4 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the method described above.

[0096] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0097] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0098] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. An energy-saving regulation and control method for a pulverizing system based on multi-objective optimization, characterized in that, include: S1. Collect and preprocess the operating parameters of the coal mill. The operating parameters include the mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. S2, based on the operating parameters, identify the current operating condition range, and configure multi-objective optimization weights according to the preset expert rule base, wherein the multi-objective optimization weights correspond to the priority relationship between energy consumption, stability and response speed; S3, The improved particle swarm optimization algorithm is used to dynamically optimize the running parameters. The improved particle swarm optimization algorithm generates preliminary optimized control parameters under preset constraints through a linear decreasing inertial weight strategy and an adaptive adjustment mechanism of the learning factor. S4, input the preliminary optimized control parameters into the expert rule base for hard constraint verification and empirical correction. The hard constraint verification includes that the primary air volume is not lower than the minimum fluidizing air volume and the mill outlet temperature is not higher than the preset threshold. The empirical correction includes triggering an advance adjustment strategy based on the load change rate, outputting the final control command and feeding it back to the actuator to adjust the operating status of the coal mill.

2. The method as described in claim 1, characterized in that, S1 includes: S11 collects seven core parameters at a sampling interval of 1 minute: mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. S12, the preprocessing stage adopts filtering and noise reduction technology, and ensures that the data length covers at least one week of the full working cycle, including the data distribution characteristics of typical high, medium and low load scenarios.

3. The method as described in claim 1, characterized in that, S2 includes: S21 divides the unit load into high load range, medium load range and low load range; S22, automatically match the target weight matrix according to the load range, where the stability weight is 0.6 for the high load range, the energy consumption weight is 0.5 for the medium load range, and the response speed weight is 0.5 for the low load range.

4. The method as described in claim 1, characterized in that, The S3 further includes: S31 employs a linearly decreasing inertia weight strategy, according to the formula... Calculate the current inertia weight, where , , ; S32, when the fitness of a particle has not improved for 10 consecutive generations, adjust the cognitive factor to 1.5 and the social factor to 2.5 to enhance group information sharing.

5. The method as described in claim 1, characterized in that, The S4 includes: S41, during hard constraint verification, ensure that the primary air volume is not less than 60% of the rated value and the mill outlet temperature is not higher than 75℃; S42, in empirical correction when the load change rate When this happens, an advance adjustment strategy is triggered, and the coal feed rate is pre-adjusted by +5% to avoid the risk of coal blockage.

6. The method as described in claim 1, characterized in that, Also includes: S5. Based on local features of the coal mill's operating images, fault diagnosis information is extracted. These local features include shape, texture, and particle size distribution. Seven typical fault types, including accumulation and blockage, insufficient grinding, and dust backflow, are identified using an expert rule base. The fault warning lead time is increased to [missing information]. .

7. An energy-saving regulation and control device for a pulverizing system based on multi-objective optimization, characterized in that, include: The operating parameter acquisition and preprocessing module is used to acquire and preprocess the operating parameters of the coal mill. The operating parameters include mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. The working condition identification and weight configuration module identifies the current working condition range based on the operating parameters and configures multi-objective optimization weights according to a preset expert rule base. The multi-objective optimization weights correspond to the priority relationship between energy consumption, stability and response speed. The dynamic optimization module is used to dynamically optimize the running parameters using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm generates preliminary optimized control parameters under preset constraints through a linear decreasing inertial weight strategy and an adaptive adjustment mechanism for the learning factor. The hard constraint verification and experience correction module is used to input the preliminary optimized control parameters into the expert rule base for hard constraint verification and experience correction. The hard constraint verification includes ensuring that the primary air volume is not lower than the minimum fluidizing air volume and that the mill outlet temperature is not higher than a preset threshold. The experience correction includes triggering an advance adjustment strategy based on the load change rate, outputting the final control command, and feeding it back to the actuator to adjust the operating status of the coal mill.

8. The apparatus as claimed in claim 7, characterized in that, The operating parameter acquisition and preprocessing module is also used for: Seven core parameters were collected at a sampling interval of 1 minute: mill outlet temperature, primary air volume, coal feed rate, coal mill current, outlet pressure, inlet primary air pressure, and inlet primary air temperature. The preprocessing stage employs filtering and noise reduction techniques and ensures that the data length covers at least one week of the full operating cycle, including the data distribution characteristics of typical high, medium, and low load scenarios.

9. A computer device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the energy-saving regulation and control method for a pulverizing system based on multi-objective optimization as described in any one of claims 1-6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements an energy-saving regulation and control method for a pulverizing system based on multi-objective optimization as described in any one of claims 1-6.

Images