A control method of a pulverizing system considering multi-parameter cooperation

Abstract

The application discloses a control method of a pulverizing system considering multi-parameter cooperation and relates to the technical field of thermal power generation. Operation parameters of a thermal power generating unit at a current time point are acquired; the operation parameters include parameters of a boiler combustion process, parameters of a pulverizing system, storage state information of an external powder bin and parameter information of a boiler variable load; the operation parameters are input into a data correlation model to obtain initial control instructions of the pulverizing system; the data correlation model is obtained based on fusion training of fuzzy reasoning and neural network algorithms; the initial control instructions are taken as initial positions of a population, a maximum comprehensive evaluation function value is taken as a target, and the initial control instructions are iteratively optimized in a combination mode of a particle swarm algorithm and a genetic algorithm, so that a global optimal solution of the last iteration is taken as the control instruction of the pulverizing system; the comprehensive evaluation function is constructed based on the performance of the pulverizing system; and the pulverizing system is controlled to operate through the control instruction. The method can realize efficient operation of the pulverizing system.

Description

Technical Field

[0001] This invention relates to the field of thermal power generation technology, and in particular to a control method for a pulverizing system that considers multi-parameter coordination. Background Technology

[0002] Currently, with the advancement of "dual carbon" targets and the large-scale grid connection of new energy sources, thermal power units are undertaking more arduous peak-shaving tasks, requiring flexible, efficient, and environmentally friendly operation across a wide load range. Positive pressure direct-fired pulverizing systems, due to their compact structure and relatively fast response speed, are widely used in thermal power units. Meanwhile, external pulverized coal storage silos are gradually becoming an important means of enhancing the peak-shaving capacity of these units.

[0003] However, current thermal power unit pulverizing systems and related supporting facilities still face numerous problems during operation. On the one hand, traditional control methods primarily regulate boiler combustion, pulverizing system operation, and external pulverized coal silo management independently, lacking coordinated consideration of key parameters such as boiler combustion efficiency, pollutant emissions, pulverizing system energy consumption, and silo storage status. This independent control model prevents subsystems from achieving synergistic effects, making it difficult to improve unit peak-shaving and load response capabilities while simultaneously optimizing overall system performance. On the other hand, existing control strategies fail to adequately explore the complex relationships between multiple parameters, relying heavily on experience-based settings or simple linear control models. When faced with factors such as coal quality changes, environmental fluctuations, and equipment aging, timely adjustments to control parameters are impossible, leading to poor system stability. This directly impacts boiler combustion characteristics and pulverizing system energy consumption, resulting in low operating efficiency for the pulverizing system. Summary of the Invention

[0004] Therefore, it is necessary to provide a control method for a pulverizing system that considers multi-parameter coordination to address the above-mentioned technical problems. This method can achieve efficient operation of the pulverizing system.

[0005] The present invention adopts the following technical solution: This invention provides a control method for a pulverizing system considering multi-parameter coordination, comprising: Obtain the operating parameters of the thermal power unit at the current moment; the operating parameters include the parameters of the boiler combustion process, the parameters of the pulverizing system, the storage status information of the external pulverizing silo, and the boiler variable load parameter information; the external pulverizing silo is the pulverized coal storage device of the pulverizing system; The operating parameters are input into the data association model to obtain the initial control commands for the pulverizing system; the data association model is obtained through training based on the fusion of fuzzy inference and neural network algorithms. The initial control command is used as the initial position of the population. With the goal of maximizing the comprehensive evaluation function value, the initial control command is iteratively optimized using a combination of particle swarm optimization and genetic algorithms. The global optimal solution of the last iteration is used as the control command of the milling system. The comprehensive evaluation function is constructed based on the performance of the milling system. The operation of the pulverizing system is controlled by control commands.

[0006] Optionally, the training process of the data association model includes: Acquire historical operational data and preprocess the historical operational data to obtain training data; The initial neural network is initially trained using training data, and the connection weights of the initial neural network are continuously adjusted using the backpropagation method. Based on the preset fuzzy rules and membership functions, the training data is fuzzified to obtain fuzzy output results, and the fuzzy output results are used as supervision information to update the connection weights of the initial neural network. The initial neural network after training is identified as the data association model.

[0007] Optionally, the training data is fuzzified according to preset fuzzy rules and membership functions to obtain fuzzy output results, and these fuzzy output results are used as supervision information to update the weights of the initial neural network, including: Based on the preset fuzzy rules and membership functions, the training data is fuzzified to obtain fuzzy output results; The fuzzy output result is defuzzified using the centroid method to obtain the adjustment coefficient; The initial neural network is trained using adjusted coefficients and training data, and the connection weights of the initial neural network are updated.

[0008] Optionally, the expression for the comprehensive evaluation function F is: ; in, For load change response time, For boiler combustion efficiency, The unit power consumption of the pulverizing system. The comprehensive pollutant emission index, w 1. w 2. w 3. w 4 represents the weight of each sub-item.

[0009] Optionally, in the iterative optimization process, the combination of particle swarm optimization and genetic algorithm can be achieved by having the genetic algorithm and particle swarm optimization perform parallel computations and exchange optimal solutions with each other in each optimization process.

[0010] Optionally, the method further includes: Perform anomaly detection on operating parameters to obtain anomaly information; The abnormal information is used for fault detection to obtain the fault detection results of the thermal power unit; the fault detection results include the fault mode, fault cause and fault phenomenon.

[0011] This invention provides a control system for a powder-making system that considers multi-parameter coordination, comprising: The data acquisition module is used to obtain the operating parameters of the thermal power unit at the current moment. The operating parameters include parameters of the boiler combustion process, parameters of the pulverizing system, storage status information of the external pulverizing silo, and boiler load change parameter information. The external pulverizing silo is a pulverized coal storage device for the pulverizing system. The intelligent computing module is used to input operating parameters into the data association model to obtain the initial control commands for the milling system. The data association model is obtained by training based on a fusion of fuzzy inference and neural network algorithms. The initial control commands are used as the initial position of the population. With the goal of maximizing the comprehensive evaluation function value, the initial control commands are iteratively optimized using a combination of particle swarm optimization and genetic algorithms. The global optimal solution of the last iteration is used as the control command for the milling system. The comprehensive evaluation function is constructed based on the performance of the milling system. The execution control module is used to control the operation of the pulverizing system through control commands.

[0012] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described control method for a pulverizing system considering multi-parameter coordination.

[0013] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described control method for a powder-making system that considers multi-parameter coordination.

[0014] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects: In this invention, by comprehensively analyzing multiple aspects of operating parameters, including boiler combustion process parameters, pulverizing system operating parameters, external pulverizer storage status information, and boiler load variation parameters, the real-time status of the pulverizing system can be accurately reflected. A data association model is used to generate initial control commands. This model combines the advantages of fuzzy reasoning in handling uncertain information with the powerful nonlinear mapping and learning capabilities of neural networks. This allows for accurate handling of complex and uncertain operating parameter relationships, thereby accurately predicting initial control commands suitable for the current operating state. This provides a more optimized starting point for the pulverizing system control, reducing the process of blind trial and error. Finally, by combining particle swarm optimization and genetic algorithms, the advantages of each are fully utilized to efficiently search the solution space of control commands, continuously optimizing the control commands to obtain the optimal control commands that better meet the requirements of efficient pulverizing system operation. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0016] Figure 1 A schematic diagram of a control system for a powder-making system considering multi-parameter coordination is provided for the present invention; Figure 2 A schematic flowchart of a control method for a powder-making system considering multi-parameter coordination provided by the present invention; Figure 3 This is a schematic diagram of a computer device for implementing a control method for a pulverizing system that considers multi-parameter coordination, as provided by the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0018] On the one hand, traditional control methods primarily regulate boiler combustion, pulverizing system operation, and external pulverized coal silo management independently, lacking coordinated consideration of key parameters such as boiler combustion efficiency, pollutant emissions, pulverizing system energy consumption, and silo storage status. For example, during peak shaving, in pursuit of rapid response to load changes, the output of the pulverizing system is excessively adjusted, leading to a surge in pulverizer energy consumption, a decrease in combustion efficiency, and excessive emissions of pollutants such as nitrogen oxides. Alternatively, in ensuring combustion efficiency, the silo storage status is ignored, resulting in silo blockage or pulverized coal waste. This independent control model prevents subsystems from achieving synergistic effects, making it difficult to improve the unit's peak shaving and load response capabilities while simultaneously optimizing the overall system performance.

[0019] On the other hand, existing control strategies lack sufficient understanding of the complex relationships between multiple parameters, relying heavily on empirical settings or simple linear control models. When faced with factors such as changes in coal quality, fluctuations in environmental conditions, and equipment aging, they cannot adjust control parameters in a timely manner, resulting in poor system stability and high operating costs. For example, the significant differences in volatile matter and calorific value of different coal types directly affect boiler combustion characteristics and pulverizing system energy consumption. However, traditional control methods struggle to adapt quickly to these changes, making it difficult for the unit to achieve optimal operating conditions under various circumstances.

[0020] Based on this, the present invention provides a control method for a pulverizing system that considers multi-parameter coordination, which can achieve safe, efficient and green operation of thermal power units.

[0021] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] First, the control system of the powder-making system considering multi-parameter coordination provided by the present invention will be described, such as... Figure 1 As shown, the system includes a data acquisition module, an intelligent computing module, an execution control module, and a monitoring and feedback module, specifically including: The data acquisition module is used to obtain the operating parameters of the thermal power unit at the current moment; the operating parameters include the parameters of the boiler combustion process, the parameters of the pulverizing system, the storage status information of the external pulverizing silo, and the boiler variable load parameter information; the external pulverizing silo is the pulverized coal storage device of the pulverizing system.

[0023] The intelligent computing module is used to input operating parameters into the data association model to obtain the initial control commands of the pulverizing system. The data association model is obtained by training based on the fusion of fuzzy inference and neural network algorithms. The initial control commands are used as the initial position of the population. With the goal of maximizing the comprehensive evaluation function value, the initial control commands are iteratively optimized by combining particle swarm optimization and genetic algorithms. The global optimal solution of the last iteration is used as the control command of the pulverizing system. The comprehensive evaluation function is constructed based on the performance of the pulverizing system.

[0024] The execution control module is used to control the operation of the pulverizing system through control commands.

[0025] Specifically, the data acquisition module consists of distributed sensor nodes and a high-performance data acquisition terminal. The sensor nodes are encapsulated in a metal shell, capable of withstanding the harsh environment of a thermal power plant characterized by high temperatures, high dust levels, and strong electromagnetic interference. Each sensor node has a built-in microprocessor, supporting both Modbus TCP / IP and Profibus DP communication protocols, and can automatically switch communication modes according to network load to ensure reliable data transmission. The data processing software performs preliminary processing on the acquired data, including real-time filtering, outlier detection, and data compression, and stores the processed data in a local cache. When the network is normal, data is transmitted to the intelligent computing module via industrial Ethernet; when the network fails, the data is stored on the local hard drive and automatically retransmitted after the network is restored, ensuring no data loss.

[0026] An external pulverized coal silo is a dynamic pulverized coal storage device. Addressing the issues of slow load response and insufficient peak-shaving capacity of existing thermal power units under grid load fluctuations, it constructs a dynamic storage and on-demand release system for pulverized coal in an external silo. This mechanism, through real-time monitoring of boiler operating parameters and pulverizing system status data, combined with grid load forecasting results, uses intelligent algorithms to pre-plan the pulverized coal storage volume and release strategy within the silo. When grid load fluctuates, the external silo can rapidly release the stored pulverized coal according to the preset strategy, providing fuel support for rapid load changes in the unit. This significantly improves the unit's load response rate and peak-shaving capacity, effectively enhancing the unit's operational flexibility and grid stability. Simultaneously, it optimizes the pulverizing system's operating efficiency and reduces unit energy consumption, resulting in significant economic and social benefits.

[0027] Intelligent Computing Module: Built on a high-performance industrial control computer, running Windows Server 2022. The module integrates a centralized control model, big data analytics software, and intelligent optimization algorithms. The big data analytics software uses a distributed file system (Hadoop Distributed File System, HDFS) to store historical data and leverages the MapReduce framework for distributed computing, enabling rapid data processing. The intelligent optimization algorithm program implements Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) by calling the Scikit-Optimize and DEAP libraries. Acquired data and model parameters are stored and managed through the InfluxDB real-time database, which uses time-series data storage and supports high-concurrency read / write operations. The intelligent computing module uses multi-threading technology to achieve parallel operation of data processing, model calculation, and algorithm optimization, improving system response speed. Upon receiving data from the data acquisition module, the module first standardizes the data, then inputs it into the centralized control model and big data analytics model for calculation. Combined with the intelligent optimization algorithm, it solves for the optimal values ​​of each key control parameter, generates control commands, and sends them to the execution control module.

[0028] The intelligent computing module also includes a high-performance industrial control computer as its core device, running Windows Server 2022 Datacenter Edition. Big data analytics software based on the Hadoop 3.3.4 distributed computing framework is deployed on the computer, and a distributed file system cluster containing NameNode and DataNode nodes is built to store and process historical operating data from the thermal power units. Simultaneously, an Anaconda 3 environment is installed to run intelligent optimization algorithm programs developed using Python 3.9. InfluxDB 2.6 is used as a real-time database to store and manage the collected real-time operating data and model parameters, balancing data storage and analytical needs.

[0029] The execution control module, centered on a programmable logic controller (PLC) and equipped with frequency converters, servo drives, and other actuators, precisely controls the operating parameters of the coal feeder, coal mill, fan, and the powder collection and delivery equipment in the pulverizing system. The PLC employs a redundant CPU configuration, supporting hot standby switching to ensure system reliability. The frequency converter and servo drive are connected to the PLC via a Profinet fieldbus with a communication rate of up to 100Mbps. After receiving control commands from the intelligent computing module, the PLC uses pulse width modulation (PWM) technology and vector control algorithms to precisely control the operating parameters of the coal feeder, coal mill, fan, and the powder collection and delivery equipment in the external powder silo. For example, for a coal feeder, the inverter output frequency is adjusted according to control commands to achieve precise control of the coal feed rate within the range of 0~100 t / h, with a control accuracy of ±0.5 t / h. For a coal mill, the mill speed is controlled between 500~1500 rpm by adjusting the inverter output voltage and frequency, with a speed control accuracy of ±1 rpm. Simultaneously, the PLC also has fault diagnosis and protection functions. When faults such as overload, overheating, or short circuit are detected, the equipment operation is immediately stopped, and a fault signal is sent to the monitoring feedback module.

[0030] Optionally, the execution control module also includes: a redundant CPU as the core controller, paired with an ET 200SP distributed I / O module. In terms of equipment control for the pulverizing system, the coal feeder drive motor is equipped with a frequency converter, communicating with the PLC via the Profinet protocol to receive frequency control signals output by the PLC, achieving variable frequency speed regulation of 0~50 Hz, corresponding to a coal feed rate of 0~100 t / h; the main motor of the coal mill is equipped with a frequency converter, achieving speed regulation of 500~1500 rpm through a vector control algorithm; the primary air fan is driven by a compact frequency converter, maintaining stable primary air pressure through PID closed-loop control. The external powder silo's powder-collecting screw motor and powder-feeding Roots blower are controlled by servo drives and frequency converters respectively, achieving precise start / stop and speed regulation. All control circuits for the actuators use shielded cables and are grounded to reduce electromagnetic interference.

[0031] The system also includes a monitoring and feedback module, which is used to detect anomalies in operating parameters and obtain abnormal information; to perform fault detection on the abnormal information and obtain fault detection results for the thermal power unit; the fault detection results include fault mode, fault cause and fault phenomenon.

[0032] Specifically, the monitoring and feedback module consists of a data monitoring unit and a fault diagnosis unit. The data monitoring unit uses a data acquisition module to collect system operating parameters in real time via the OPC protocol and compares them with preset safety thresholds. These safety thresholds are set based on equipment design parameters, operating experience, and industry standards; for example, the main steam pressure safety range is set at 16~18MPa, and the upper limit of the pulverizing system equipment current is set at 120% of the rated current. When a parameter exceeds the threshold, the data monitoring unit immediately sends the abnormal information to the fault diagnosis unit. The fault diagnosis unit uses a Bayesian network algorithm to pre-construct a Bayesian network model containing equipment fault modes, fault causes, and fault phenomena. By analyzing the abnormal information and calculating the probability of each fault occurring, when an abnormality is detected, an audible and visual alarm is immediately triggered, and a fault handling solution is pushed to the operator through the human-machine interface. Simultaneously, the system operating data is fed back to the intelligent computing module, which adjusts the control strategy based on the feedback data to achieve closed-loop control.

[0033] Monitoring and Feedback Module and Interactive Interface Deployment: The data monitoring unit employs a data acquisition module, communicating with the PLC and intelligent computing module via the OPC UA protocol to collect system operating parameters at 500 ms intervals. The fault diagnosis unit runs on the industrial control computer of the intelligent computing module, developing a Bayesian network algorithm program based on the MATLAB 2022b platform, and training the network using a pre-imported training set of fault sample data. The human-machine interface uses an industrial touchscreen, with three levels of user permissions: operator, engineer, and administrator, corresponding to different operation and setting permissions. It establishes a connection with the PLC and intelligent computing module via the PC UA protocol to achieve real-time data interaction and display.

[0034] The data acquisition module is used to collect multi-dimensional data on the operation of the pulverizing system, external pulverizing silos, and boiler in real time through a sensor network. The intelligent computing module is used to construct a unified control model, using fuzzy inference and neural network algorithms to train historical operating data under different load conditions and coal quality conditions, build a data association model, and discover key parameters for optimizing system performance. Then, particle swarm optimization and genetic algorithms are combined to optimize the obtained key parameters and calculate the optimal balance point for each key control parameter. The execution control module receives control commands output by the intelligent computing module and drives the pulverizing system and external pulverizing silos to operate collaboratively. The monitoring and feedback module monitors the system's operating status in real time and feeds back the operating data to the intelligent computing module to assist in adjusting the control strategy.

[0035] The sensors in the data acquisition module include a pressure sensor, a temperature sensor, a flow sensor, a component analyzer, and a weighing sensor, which are used to collect data on pressure, temperature, flow rate, gas composition, and powder storage capacity, respectively. The sensors also have data filtering and calibration functions.

[0036] The above describes a control system for a milling system considering multi-parameter coordination, provided by one or more embodiments of the present invention. Based on the same concept, the present invention also provides a corresponding control method for a milling system considering multi-parameter coordination. Specific limitations of the control system for a milling system considering multi-parameter coordination can be found in the limitations of the control method for a milling system considering multi-parameter coordination described below, and will not be repeated here. Each module in the above-described control system for a milling system considering multi-parameter coordination can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in hardware or independently of the processor in a computer device, or stored in software in the memory of a computer device, so that the processor can call and execute the operations corresponding to each module.

[0037] Figure 2 This is a schematic diagram of a control method for a powder-making system considering multi-parameter coordination according to the present invention, which specifically includes the following steps: S101, obtain the operating parameters of the thermal power unit at the current moment; the operating parameters include the parameters of the boiler combustion process, the parameters of the pulverizing system, the storage status information of the external pulverizing silo, and the boiler variable load parameter information; the external pulverizing silo is the pulverized coal storage device of the pulverizing system.

[0038] The boiler combustion process parameters include fuel composition and flow rate, air flow rate and temperature, furnace temperature and pressure, etc.; pulverizing system parameters include the speed of the coal mill, inlet and outlet pressure difference, coal powder humidity and particle size distribution, etc.; storage status information of the external pulverizing silo includes silo material level height, silo temperature and humidity, etc.; and boiler variable load parameter information.

[0039] These operating parameters can be acquired in real time by sensors and need to be preprocessed, such as normalization and data cleaning, to ensure the quality and consistency of the data so that it can be processed by the neural network model.

[0040] Operational data acquisition includes: establishing a sensor network covering the entire operation process of the pulverizing system, external pulverizing silos, and boiler. In the pulverizing system, a dual-weighing bridge structure load cell is installed at the pulverizer inlet, coupled with a high-precision weighing instrument. After on-site calibration, the coal feed measurement error is ensured to be within ±0.18%. Near the pulverizer bearing housing, a magnetoelectric speed sensor is vertically installed, with the sensor probe's distance from the rotating bearing component adjusted to 3mm to ensure a resolution of 0.1rpm. It is connected to the data acquisition terminal via redundant wiring, improving reliability through redundant design. A thermal gas mass flow meter is horizontally installed on the primary air duct, with a temperature-compensated platinum resistance thermometer installed upstream to eliminate the influence of temperature on flow measurement. After calibration, the flow measurement accuracy reaches ±0.9%. The temperature sensor at the separator outlet uses an armored thermocouple with a high-temperature resistant alloy protective tube, achieving an accuracy of ±0.5℃. Inside the external powder silo, a high-frequency radar level gauge is installed at the top center. This gauge can penetrate dust to accurately measure the powder storage volume, with a blind zone of 0.5m and a range of 0~30m. It can accurately measure the powder storage volume even with a dust concentration of 500g / m³. A high-sensitivity piezoresistive pressure sensor is installed on the top side wall, with a range of 0~5 kPa and an accuracy of ±0.08 kPa. Above the powder silo outlet, a capacitive humidity sensor is installed with a response time of less than 5 seconds, which can be 3 seconds. The measurement range is 0~100% RH, and the accuracy is ±2% RH. Differential pressure transmitters are installed on the boiler steam-water system and flue, and on the main steam pipeline. The range covers 0~32MPa, and the accuracy is ±0.07%. The positive pressure chamber is connected to the main steam pipeline, and the negative pressure chamber is open to the atmosphere. A zirconia flue gas oxygen analyzer is installed in the flue gas, using a zirconia probe. The probe is inserted to a depth of 1 / 3 of the flue gas diameter, with a measurement range of 0~25% and an accuracy of ±0.15%. All sensors are connected to a redundantly configured industrial Ethernet switch via double-shielded network cables, forming a ring network topology to ensure reliable data transmission. Data is acquired at a 1-second interval, and time synchronization technology is used to ensure data consistency. The acquired data is first processed by median filtering at the data acquisition terminal to remove random impulse interference, and then further optimized using a Kalman filter algorithm to eliminate system noise and measurement errors, ensuring data accuracy and stability.

[0041] After the system powers on, it automatically completes sensor self-tests and network connections, collecting data from each sensor at a 1-second interval. The collected data is first processed by median filtering at the data acquisition terminal with a window size of 5 to remove random impulse interference. Then, it is further optimized using a Kalman filter algorithm, dynamically estimating and correcting the data based on the sensor's measurement noise covariance and the system's process noise covariance. The processed data is stored in a local cache and synchronously transmitted to the InfluxDB real-time database in the intelligent computing module.

[0042] S102, input the operating parameters into the data association model to obtain the initial control commands of the pulverizing system; the data association model is obtained by training based on the fusion of fuzzy inference and neural network algorithms.

[0043] Optionally, the training process of the data association model includes: acquiring historical running data and preprocessing the historical running data to obtain training data; performing preliminary training on the initial neural network using the training data and continuously adjusting the connection weights of the initial neural network using the backpropagation method; performing fuzzification processing on the training data according to preset fuzzy rules and membership functions to obtain fuzzy output results, and using the fuzzy output results as supervision information to update the connection weights of the initial neural network; and determining the initial neural network after training as the data association model.

[0044] Optionally, the training data is fuzzified according to preset fuzzy rules and membership functions to obtain fuzzy output results, and the fuzzy output results are used as supervision information to update the weights of the initial neural network. This includes: fuzzifying the training data according to preset fuzzy rules and membership functions to obtain fuzzy output results; defuzzifying the fuzzy output results using the centroid method to obtain adjustment coefficients; and training the initial neural network with the adjustment coefficients and training data to update the connection weights of the initial neural network.

[0045] Specifically, big data analysis and parameter correlation mining: machine learning algorithms are used to conduct in-depth analysis of historical operating data of thermal power units. Historical data is collected, covering operating records under different load conditions (from deep peak shaving conditions at 30% rated load to 100% rated load), coal quality (including anthracite, lean coal, bituminous coal, lignite, etc.), and different environmental conditions. A multi-layer initial neural network is constructed using the TensorFlow deep learning framework, with the network structure including an input layer, multiple hidden layers, and an output layer. The input layer receives preprocessed multi-dimensional data, and the hidden layers use the ReLU activation function for feature extraction. The network parameters are trained using the backpropagation algorithm and the Adam optimizer. During training, cross-validation is used to divide the training and test sets, and the network structure and parameters are continuously adjusted until the mean squared error (MSE) of the model on the test set is less than 0.03. This allows for accurate mining of nonlinear relationships between various parameters, such as the discovery of complex coupling relationships between coal mill output and primary air velocity, pulverized coal fineness, and boiler load, as well as the nonlinear influence of coal volatile matter on combustion efficiency and pollutant emissions, providing data support for optimized control.

[0046] A multi-parameter linkage model was trained using historical operating data of the unit. The data was divided into training and test sets in an 8:2 ratio. A multi-layer neural network model with three hidden layers was constructed based on the TensorFlow 2.8 framework, with 128, 64, and 32 neurons in the hidden layers, and ReLU activation function was used for all layers. The training batch size was set to 128, the initial learning rate was 0.001, and the Adam optimizer was used for parameter updates. The model was validated on the test set every 10 training epochs. After 300 training epochs, the mean squared error (MSE) of the model on the test set stabilized at 0.028, meeting the accuracy requirements. The trained model parameters were saved for subsequent real-time calculations.

[0047] When the generating unit receives a load change command signal from the power grid dispatch center, the intelligent computing module immediately acquires the current load command and the actual operating load of the unit. If the load change rate exceeds 2% of the rated load per minute, it is determined to enter the rapid load change mode; if the unit load drops below 40% of the rated load, it is determined to enter the deep peak shaving mode. At the same time, the collected operating parameters of the pulverizing system, external pulverizing silo, and boiler are input into the neural network model and fuzzy control module.

[0048] In the fuzzy control module, fuzzy rules are first established based on experience and understanding of the system, usually in the form of "if-then". Factors such as the current coal powder quantity in the coal powder silo, changes in load commands, and deviations in coal feed rate need to be comprehensively considered. The following fuzzy rules are formulated: The amount of pulverized coal in the pulverized coal bin is divided into five levels: {-3, -2, 0, 2, 3}. The load command change is also divided into the same level, and then converted into fuzzy quantities through triangular membership functions.

[0049] (1) When the amount of pulverized coal is high and the load is stable If the coal powder quantity level of the coal powder silo is "3" (high) and the load command change level is "0" (no change), then reduce the coal feeder's coal feed rate, reduce the speed of the coal conveying fan, and reduce the amount of coal powder fed into the coal powder silo.

[0050] (2) When the amount of pulverized coal is high and the load decreases If the coal powder quantity level of the coal powder silo is "3" (high) and the load command change level is "-3" (significant decrease), the coal feeder's coal feed rate will be significantly reduced, the coal conveying fan speed will be significantly reduced, and the coal powder silo discharge operation can be started if necessary.

[0051] (3) When the amount of pulverized coal is low and the load increases If the coal powder quantity level in the coal powder silo is "-3" (low) and the load command change level is "3" (significant increase), then the coal feeder's coal feed rate will be significantly increased, the coal conveying fan speed will be increased, and coal powder will be replenished to the coal powder silo more quickly.

[0052] (4) When the amount of pulverized coal is low and the load is stable If the coal powder quantity level in the coal powder silo is "-3" (low) and the load command change level is "0" (no change), increase the coal feeder's coal feed rate and appropriately increase the speed of the coal conveying fan.

[0053] (5) The amount of pulverized coal is moderate and the load increases significantly. If the coal powder quantity level in the coal powder silo is "0" (moderate) and the load command change level is "3" (significant increase), then increase the coal feeder's coal feed rate and increase the speed of the coal conveying fan to reserve coal powder in advance to meet the load demand.

[0054] (6) The amount of pulverized coal is moderate and the load drops significantly. If the coal powder quantity level in the coal powder silo is "0" (moderate) and the load command change level is "-3" (significant decrease), reduce the coal feeder's coal feed rate and lower the coal conveying fan speed to avoid coal powder accumulation.

[0055] (7) Rules for considering coal feed deviation If the coal powder quantity level in the coal powder silo is "-2" (lower) and the coal feed deviation level is "-3" (coal feed is much lower than expected), the coal feeder should be increased quickly, the coal feed rate should be increased, the coal conveying fan speed should be increased, and an abnormal warning for the coal feeding equipment should be issued.

[0056] If the coal powder quantity level in the coal powder silo is "2" (relatively high) and the coal feed deviation level is "3" (the coal feed is much higher than expected), quickly reduce the coal feeder's coal feed rate, reduce the speed of the coal conveying fan, and check the operating status of the coal feeding equipment.

[0057] Then, inference is performed based on fuzzy rules. For each rule, its activation strength is calculated using the method described above, and its contribution to the output fuzzy set is determined based on the activation strength. The output obtained through fuzzy inference is a fuzzy quantity, which needs to be converted into a specific numerical value through defuzzification to actually control the actuator. Defuzzification is performed using the centroid method, and the centroid coordinates of the output fuzzy set are calculated as the defuzzified result.

[0058] Fuzzy reasoning is performed based on pre-set fuzzy rules to obtain fuzzy outputs of the output adjustment coefficient of the powder making system and the powder taking amount adjustment coefficient of the external powder hopper. Then, the center of gravity method is used for defuzzification to obtain accurate adjustment coefficients.

[0059] The adjustment coefficients from the fuzzy control output, real-time operating parameters, and historical data from the previous three time steps are input into the data association model. After receiving the data, the input layer of the LSTM network in the data association model performs time-series feature extraction through a two-layer LSTM layer. Then, through a fully connected layer, it outputs the adjustment values ​​for the coal feeder's coal rate, the pulverizer's speed, the external pulverizer's powder intake, and the pulverizing fan's speed. The calculated adjustment values ​​are then converted into corresponding initial control commands.

[0060] S103 uses the initial control command as the initial position of the population, and aims to maximize the comprehensive evaluation function value. It iteratively optimizes the initial control command by combining particle swarm optimization and genetic algorithm, and uses the global optimal solution of the last iteration as the control command of the pulverizing system. The comprehensive evaluation function is constructed based on the performance of the pulverizing system.

[0061] Key parameters such as boiler combustion efficiency, pollutant emissions, pulverizing system energy consumption, and external pulverized coal storage status are organically integrated. To measure boiler combustion efficiency, in addition to main steam temperature, pressure, and thermal efficiency, parameters such as flame stability coefficient and furnace temperature field uniformity are introduced, with data acquired through infrared thermography and flame image monitoring equipment installed in the furnace. For pollutant emission monitoring indicators, in addition to nitrogen oxides, sulfur dioxide, and particulate matter, carbon monoxide, mercury and its compounds are added, with real-time detection using online flue gas analyzers. Pulverizing system energy consumption analysis not only relates to pulverizer power and feeder power consumption but also considers factors such as fan power consumption and system air leakage rate, with data collected through power sensors and anemometers. In monitoring the external pulverized coal storage status, in addition to pulverized coal storage volume, silo pressure, and pulverized coal humidity, parameters such as pulverized coal flowability index and silo level change rate are added, measured using ultrasonic level gauges and rheometers. Each key parameter is connected to the centralized control model through a data interface. Mathematical correlation equations between the parameters are established based on the principles of thermodynamics, fluid mechanics, and chemical reaction kinetics to form a complete control system.

[0062] Combining Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), parameter optimization is performed with the goal of achieving optimal overall system performance. A comprehensive evaluation function F is established, with the following expression:

[0063] ; in, For load change response time, For boiler combustion efficiency, The unit power consumption of the pulverizing system. The comprehensive pollutant emission index ( , a , b , c (These are the weighting coefficients for each pollutant).w 1. w 2. w 3. w 4 represents the weight of each sub-item, which is dynamically adjusted according to different operational needs and environmental protection policies. For example, during periods of strict environmental protection requirements, the weight of w4 is appropriately increased.

[0064] In the PSO algorithm, the particle swarm size is set to 50, the maximum number of iterations is 100, the inertia weight decreases linearly from 0.9 to 0.4, and the learning factor is... c 1 and c Both are 2; in the GA algorithm, the population size is set to 60, the crossover probability is 0.8, the mutation probability is 0.01, and the number of iterations is 120.

[0065] In the iterative optimization process, the combination of particle swarm optimization (PSO) and genetic algorithm involves parallel computation of the two algorithms, with the optimal solutions being exchanged between them in each optimization process. Specifically, the two algorithms compute in parallel, periodically exchanging optimal solutions. By continuously iterating to find the maximum value of the comprehensive evaluation function, the optimal balance point of each key control parameter (such as coal feed rate of the coal feeder, speed of the coal mill, primary air velocity, and powder intake of the external powder silo) is dynamically calculated, and corresponding control commands are output to achieve multi-parameter collaborative optimization control.

[0066] After receiving the control commands from the pulverizing system, they can be sent to the PLC of the execution control module via the Profinet protocol.

[0067] S104 controls the operation of the pulverizing system through control commands.

[0068] After receiving control commands, the PLC adjusts the inverter output frequency using PWM technology according to preset control logic to control the operation of equipment such as the coal feeder, coal mill, and fan. For example, if the coal feeder's feed rate is adjusted to +5 t / h, the PLC will increase the inverter output frequency by the corresponding value, gradually increasing the coal feeder's feed rate to the target value. Simultaneously, the PLC monitors equipment operating status parameters such as motor current and bearing temperature in real time. When an abnormality is detected, it immediately triggers a protective shutdown and sends a fault signal to the monitoring feedback module.

[0069] Optionally, the method further includes: performing anomaly detection on operating parameters to obtain anomaly information; performing fault detection on the anomaly information to obtain fault detection results for the thermal power unit; the fault detection results include fault mode, fault cause, and fault phenomenon.

[0070] Specifically, operating parameters are compared with preset safety thresholds. These safety thresholds are set based on equipment design parameters, operating experience, and industry standards; for example, the main steam pressure safety range is set at 16-18 MPa, and the upper limit of the pulverizing system equipment current is set at 120% of the rated current. A Bayesian network algorithm is used to pre-construct a Bayesian network model containing equipment fault modes, fault causes, and fault phenomena. When parameters exceed the thresholds, abnormal information is analyzed to calculate the probability of each fault occurring. When an anomaly is detected, an audible and visual alarm is immediately triggered, and a fault handling solution is pushed to the operator through the human-machine interface. Simultaneously, system operating data is fed back to the intelligent computing module, which adjusts the control strategy based on the feedback data to achieve closed-loop control.

[0071] The system can collect operating parameters every 500 ms to calculate deviations between actual fuel supply and unit load demand, boiler combustion efficiency, and pollutant emission concentration. When any deviation exceeds a preset threshold, the intelligent calculation module is triggered to initiate secondary optimization.

[0072] The secondary optimization employs a genetic algorithm for parameter fine-tuning, with a population size of 40, a crossover probability of 0.85, a mutation probability of 0.015, and 30 iterations. Aiming to maximize the comprehensive evaluation function value, key control parameters of the milling system and external milling silo are locally optimized. The optimized control commands are then sent back to the execution control module to dynamically adjust the control strategy until the system operating parameters stabilize within the preset range.

[0073] The operating parameters, abnormal information, and fault detection results of the aforementioned thermal power units can be displayed through a human-machine interface (HMI). Specifically, the HMI uses an industrial touchscreen and is developed based on the WinCC V8.0 software platform. The interface design follows ergonomic principles and is divided into functional modules such as system overview, parameter settings, real-time curves, alarm records, and historical data query. In the system overview interface, a dynamic flowchart visually displays the operating status of the thermal power unit's pulverizing system, external pulverizer silo, and boiler. Key parameters are displayed using numbers and color changes; for example, green indicates normal operation, and red indicates abnormal operation. The parameter settings interface allows operators to manually adjust control strategy parameters such as the weight coefficients and control thresholds of the intelligent optimization algorithm, and also has parameter backup and recovery functions. The real-time curve interface can simultaneously display the real-time change curves of multiple parameters, facilitating operators' observation of the relationships and trends between parameters. The alarm record interface records in detail the time, type, location, and handling status of faults, and supports queries based on time, type, and other conditions. The historical data query interface allows querying operating data for any past time period and generates reports and charts for convenient data analysis and performance evaluation. The human-machine interface can also interact with the distributed control system (DCS) of the power plant via the OPC UA protocol to achieve integrated management. Operators can monitor the operating status of this system on the DCS system, and can also view the relevant parameters of the DCS system on the human-machine interface of this system.

[0074] In one embodiment, the specific flow of the control method for a powder-making system considering multi-parameter coordination provided by the present invention is as follows: S1. Construct a centralized control model and set boiler combustion efficiency, pollutant emissions, pulverizing system energy consumption, and external pulverizing silo storage status as key control parameters.

[0075] S2 utilizes a sensor network covering the entire process of the pulverizing system, external powder silos, and boiler operation to collect multi-dimensional data on temperature, pressure, flow rate, and composition in real time.

[0076] The multi-dimensional data includes the pulverizer speed, coal feed rate, primary air velocity, separator outlet temperature, external powder storage capacity and pressure, and boiler main steam pressure, main steam temperature and flue gas oxygen content of the pulverizing system.

[0077] S3. Utilize big data analytics to uncover the potential correlations and influence patterns between the key control parameters and multi-dimensional data.

[0078] Big data analytics technology uses machine learning algorithms to train historical operating data and build a data association model. The historical operating data covers operating data under different load conditions and coal quality conditions.

[0079] S4 combines particle swarm optimization and genetic algorithms to dynamically calculate the optimal balance point of each key control parameter with the goal of optimizing the overall system performance, and outputs the corresponding control commands.

[0080] With the goal of optimizing the overall system performance, this involves establishing a comprehensive evaluation function that includes the peak-shaving capacity of thermal power units, combustion efficiency, energy consumption indicators, and pollutant emission indicators. The maximum value of the comprehensive evaluation function is then obtained through an optimization algorithm to determine the optimal balance point of each key control parameter.

[0081] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In existing technologies, the pulverizing system, boiler combustion, and external pulverized coal storage management of thermal power units mostly adopt independent control modes, lacking consideration of the synergistic relationship between multiple parameters. For example, when adjusting the output of the pulverizing system, the impact on boiler combustion efficiency and pollutant emissions is often ignored, which can easily lead to unstable combustion and excessive emissions. This invention constructs a centralized control model that incorporates key parameters such as boiler combustion efficiency, pollutant emissions, pulverizing system energy consumption, and external pulverized coal storage status. Based on the principles of thermodynamics and fluid mechanics, mathematical correlation equations between parameters are established to achieve multi-parameter synergistic optimization control. This model can comprehensively consider the changes in each parameter while adjusting the pulverizing system, ensuring optimal overall system performance.

[0082] 2. Existing technologies for data acquisition are mostly limited to conventional operating parameters, and data processing methods are simple, failing to delve into the complex relationships between parameters. This invention constructs a sensor network covering the entire process, collecting parameters that not only include traditional indicators but also new key parameters such as flame stability coefficient and pulverized coal flowability index. The acquisition frequency reaches 1 second / time, and median filtering and Kalman filtering techniques are used to ensure data accuracy. In terms of data analysis, multi-condition, multi-coal-quality, and multi-environment data are utilized, and a multi-layer neural network model is constructed using the TensorFlow deep learning framework. After cross-validation training, it can accurately uncover the nonlinear relationships between parameters, such as precisely analyzing the complex influence of coal volatile matter on combustion efficiency and pollutant emissions, providing strong support for precise control.

[0083] 3. Traditional control strategies often rely on empirical settings or simple linear models, making it difficult to adapt to complex operating conditions. This invention combines Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) to establish a comprehensive evaluation function that includes indicators such as peak-shaving capability, combustion efficiency, energy consumption, and emissions. It dynamically adjusts weights to adapt to different operational needs and environmental policies. In terms of algorithm parameter settings, the particle swarm size, number of iterations, and inertia weights of the PSO algorithm, as well as the population size and crossover / mutation probability of the GA algorithm, have all been scientifically designed. The two algorithms compute in parallel and exchange optimal solutions. Compared to existing single algorithms or simple control methods, this approach can more accurately calculate the optimal balance point of each key control parameter, improving control accuracy and system response speed.

[0084] 4. Regarding peak-shaving capacity, existing technologies for deep peak shaving in thermal power units can only achieve 40% of rated load, while this invention can reduce it to 30% of rated load, significantly shortening response delay and better adapting to the grid's peak-shaving needs. In terms of energy consumption, it can reduce energy consumption in the pulverizing system, decrease coal consumption, lower equipment maintenance costs, and significantly reduce operating costs. Furthermore, the system's self-learning and adaptive capabilities, as well as fault prediction accuracy, are improved, effectively enhancing operational stability and intelligence, achieving integrated collaborative management across all aspects, and greatly improving the overall operating performance of the thermal power unit.

[0085] When applying the control method for a pulverizing system that considers multi-parameter coordination provided by this invention, it is not necessary to consider... Figure 2 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this invention does not impose any restrictions on it.

[0086] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 2 The provided method is a control method for a powder-making system that considers the coordination of multiple parameters.

[0087] The present invention also provides Figure 3 The schematic diagram of the computer device shown is as follows: Figure 3 As shown, at the hardware level, this computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 2 The provided method is a control method for a powder-making system that considers the coordination of multiple parameters.

[0088] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0089] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A control method for a milling system considering multi-parameter coordination, characterized in that, include: Obtain the operating parameters of the thermal power unit at the current moment; the operating parameters include the parameters of the boiler combustion process, the parameters of the pulverizing system, the storage status information of the external pulverizing silo, and the boiler variable load parameter information; the external pulverizing silo is the pulverized coal storage device of the pulverizing system; The operating parameters are input into the data association model to obtain the initial control commands for the pulverizing system; The data association model is obtained through training based on the fusion of fuzzy inference and neural network algorithms; The initial control command is used as the initial position of the population. With the goal of maximizing the comprehensive evaluation function value, the initial control command is iteratively optimized using a combination of particle swarm optimization and genetic algorithms. The global optimal solution of the last iteration is used as the control command of the milling system. The comprehensive evaluation function is constructed based on the performance of the milling system. The operation of the pulverizing system is controlled by control commands.

2. The method according to claim 1, characterized in that, The training process of the data association model includes: Acquire historical operational data and preprocess the historical operational data to obtain training data; The initial neural network is initially trained using training data, and the connection weights of the initial neural network are continuously adjusted using the backpropagation method. Based on the preset fuzzy rules and membership functions, the training data is fuzzified to obtain fuzzy output results, and the fuzzy output results are used as supervision information to update the connection weights of the initial neural network. The initial neural network after training is identified as the data association model.

3. The method according to claim 2, characterized in that, Based on preset fuzzy rules and membership functions, the training data is fuzzified to obtain fuzzy output results. These fuzzy output results are then used as supervisory information to update the weights of the initial neural network, including: Based on the preset fuzzy rules and membership functions, the training data is fuzzified to obtain fuzzy output results; The fuzzy output result is defuzzified using the centroid method to obtain the adjustment coefficient; The initial neural network is trained using adjusted coefficients and training data, and the connection weights of the initial neural network are updated.

4. The method according to claim 1, characterized in that, The expression for the comprehensive evaluation function F is: ； in, For load change response time, For boiler combustion efficiency, The unit power consumption of the pulverizing system. The comprehensive pollutant emission index, w 1. w 2. w 3. w 4 represents the weight of each sub-item.

5. The method according to claim 1, characterized in that, In the iterative optimization process, the combination of particle swarm optimization and genetic algorithm can be achieved by having the genetic algorithm and particle swarm optimization perform parallel computations and exchange the optimal solution with each other in each optimization process.

6. The method according to claim 1, characterized in that, The method further includes: Perform anomaly detection on operating parameters to obtain anomaly information; The abnormal information is used for fault detection to obtain the fault detection results of the thermal power unit; the fault detection results include the fault mode, fault cause and fault phenomenon.

7. A control system for a powder-making system considering multi-parameter coordination, characterized in that, include: The data acquisition module is used to obtain the operating parameters of the thermal power unit at the current moment. The operating parameters include parameters of the boiler combustion process, parameters of the pulverizing system, storage status information of the external pulverizing silo, and boiler load change parameter information. The external pulverizing silo is a pulverized coal storage device for the pulverizing system. The intelligent computing module is used to input operating parameters into the data association model to obtain the initial control commands for the pulverizing system; The data association model is obtained by training based on the fusion of fuzzy inference and neural network algorithms. The initial control command is used as the initial position of the population. With the goal of maximizing the comprehensive evaluation function value, the initial control command is iteratively optimized by combining particle swarm optimization and genetic algorithms. The global optimal solution of the last iteration is used as the control command of the pulverizing system. The comprehensive evaluation function is constructed based on the performance of the pulverizing system. The execution control module is used to control the operation of the pulverizing system through control commands.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method as described in any one of claims 1 to 6.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 6.

Images