Accurate regulation and control method for working voltage of multi-electric-field corrugated plate type ESP (electronic stability program)

Abstract

The invention discloses a multi-electric-field corrugated plate type ESP working voltage accurate regulation and control method, and relates to the technical field of electrostatic precipitation, and the method comprises the specific steps: constructing a working condition-dust removal efficiency database containing training and testing data; the ATPSO-SVM prediction model is trained to predict the optimal voltage; working condition parameters are collected and processed in real time; predicting and outputting the optimal voltage meeting the constraint; regulation and control are executed through a PLC, the actual dust removal efficiency is optimized in a closed-loop mode, and efficient and stable operation is ensured; according to the method, an ATPSO-SVM prediction model is established, and a Gaussian radial kernel function and an improved active target point particle swarm optimization algorithm are combined, so that the relation between the working condition and the voltage is accurately fitted, and the problem of traditional regulation response lag is solved; meanwhile, exclusive regulation and control logic is designed, a complete regulation and control process is constructed, the structural advantages of the equipment are brought into full play, collaborative optimization of dust removal efficiency and energy consumption is achieved, and the stability and practicability of equipment operation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrostatic precipitation, in particular to a method for precisely regulating the working voltage of a multi-electric-field corrugated plate ESP. BACKGROUND

[0002] In the field of industrial flue gas purification, electrostatic precipitator (ESP) is a key equipment for controlling PM2.5 and other fine particulate matter emissions, and is widely used in coal-fired power plants, metal smelting and other high-pollution industries. With the continuous tightening of environmental protection standards, the combination design of multi-electric-field series structure and corrugated plate dust collection electrode plate has become the mainstream. Multi-electric-field can improve the particle removal rate through staged collection, and corrugated electrode plate can increase the dust collection area and optimize the flow field. The combination of the two can significantly improve the fine particle collection effect. However, in actual operation, the working conditions such as flue gas flow rate and particle concentration often fluctuate dynamically, which puts higher requirements on the working voltage regulation of multi-electric-field corrugated plate ESP. It is necessary to reduce energy consumption while ensuring dust removal efficiency, but the existing regulation technology cannot fully adapt to the structural characteristics and working condition requirements of such equipment.

[0003] The energy-saving optimization method and system for the electrostatic precipitator disclosed in the patent with the authorization announcement number CN116060212B can realize collaborative control by globally deploying the power of each high-voltage electric field, thereby improving the energy utilization rate. However, there are still some deficiencies. This method does not design a dedicated regulation logic for the structural characteristics of the corrugated plate ESP, and does not consider the influence of the unique structure of the corrugated electrode plate on the electric field distribution, so it cannot fully utilize the low-voltage high-capture-efficiency advantage of the corrugated plate itself. Moreover, the regulation process relies on pre-set operating modes and fixed control parameters, and lacks predictive regulation capability based on real-time working conditions. In the commonly seen fluctuation range of industrial working conditions, voltage adjustment lag may occur, which may lead to large fluctuations in dust removal efficiency or unnecessary energy waste due to excessive voltage regulation, making it difficult to meet the actual demand for precise voltage regulation of multi-electric-field corrugated plate ESP. SUMMARY

[0004] The present application aims to overcome the deficiencies of the prior art and provides a method for precisely regulating the working voltage of a multi-electric-field corrugated plate ESP. This method constructs a working condition-dust removal efficiency database and trains an ATPSO-SVM prediction model based on the Gaussian radial kernel function and the improved active target point particle swarm optimization algorithm to realize precise prediction and regulation of the working voltage of the multi-electric-field corrugated plate ESP. This method effectively solves the problem that traditional regulation methods cannot adapt to the nonlinear characteristics and response lag of the equipment, can respond to changes in working conditions in real time, and output an adaptive voltage regulation scheme. At the same time, combined with a closed-loop optimization mechanism, it ensures that the dust removal efficiency meets the standard stably, realizes the collaborative optimization of dust removal efficiency and energy consumption, and improves the stability and practicality of equipment operation.

[0005] The application provides the following technical scheme to solve the above technical problems: a multi-electric field corrugated plate ESP working voltage precise regulation and control method, and the specific steps of the method are as follows: Database construction: determine the multi-electric field corrugated plate ESP core structure parameters, select a flow rate sensor, a PM2.5 concentration monitor and a high-voltage power supply, set an industrial typical working condition range, continuously monitor the dust removal efficiency under each working condition combination and take the average value to form a working condition-dust removal efficiency database containing training data and test data; Training of the prediction model: take the database training data as the input, build a model architecture with the flue gas flow rate and the target dust removal efficiency as the input variables and the optimal working voltage of each electric field as the output variable, adopt a Gaussian radial kernel function to build a support vector machine basic model, introduce an active target point particle swarm optimization algorithm and improve the speed update formula, take the average relative error between the predicted value and the measured value as the fitness function, iteratively optimize the support vector machine parameters, and complete the training of the ATPSO-SVM prediction model; Real-time acquisition of working condition parameters: start the sensor, collect the ESP inlet flue gas flow rate and outlet actual dust removal efficiency at a preset frequency, pretreat the collected data and obtain a preset target dust removal efficiency, and form a real-time working condition parameter set; Prediction of the optimal working voltage: input the pretreated current flue gas flow rate and target dust removal efficiency into the trained ATPSO-SVM prediction model, and the model calculates and outputs the optimal working voltage values of the three electric fields, and the output satisfies the preset constraint condition; Execution of regulation and control and closed-loop optimization: the programmable logic controller receives the optimal working voltage, drives the high-voltage power supply to adjust the voltages of the electric fields in a preset order, monitors the actual dust removal efficiency at a preset frequency after the adjustment, calculates the deviation from the target efficiency, re-inputs the current working condition parameters into the model for iterative regulation and control if the deviation exceeds the preset threshold, and automatically adjusts the voltages of the electric fields if an emergency condition is triggered.

[0006] Further, in the database construction step, the core structure parameters of the multi-electric field corrugated plate ESP are as follows: the corrugated plate height of a single electric field is 10 mm, the corrugated plate bottom distance is 50 mm, the discharge electrode wire radius is 1 mm, and the distance between adjacent electrode wires is 40 mm; the accuracy of the flow rate sensor is ±0.01 m / s, the accuracy of the PM2.5 concentration monitor is ±1 μg / m 3 , and the adjustment accuracy of the high-voltage power supply is ±0.1 kV.

[0007] Further, in the constructing database step, the industrial typical working condition range is: flue gas flow rate 0.6-1.2 m / s, flow rate gradient 0.1 m / s; working voltage 20-30 kV, voltage gradient 0.5 kV; the particle characteristics involved in the working condition is PM2.5, particle size range 0.1-2.5 μm and Rosin-Rammler distribution is met, wherein the median particle size is 1.25 μm and the distribution parameter is 1.64; the training data is 20 groups and the test data is 7 groups.

[0008] Further, in the training prediction model step, the velocity update formula of the improved active target point particle swarm optimization algorithm is: ; Wherein: is the velocity of the i-th particle in the d-th dimension in the k+1-th iteration; is the inertia weight of the algorithm; is the velocity of the i-th particle in the d-th dimension in the k-th iteration; , , are all learning factors of the algorithm; , , are all random numbers with values in the interval [0, 1]; is the individual optimal position of the i-th particle in the d-th dimension in the k-th iteration; is the current position of the i-th particle in the d-th dimension in the k-th iteration; is the current position of the i-th particle in the d-th dimension in the k-th iteration; is the current position of the i-th particle in the d-th dimension in the k-th iteration; is the current position of the i-th particle in the d-th dimension in the k-th iteration; is the active target point of the i-th particle in the d-th dimension in the k-th iteration.

[0009] Further, in the training prediction model step, the active target point is generated by worst point mapping, and the mapping formula is: ; Wherein: is the active target point obtained after mapping, and the value is equal to the active target point ; is the individual optimal position of the i-th particle in the d-th dimension in the k-th iteration process; is the group optimal position of the entire particle swarm in the d-th dimension in the k-th iteration process, and the group optimal position refers to the position with the optimal fitness function value of the entire particle swarm from the start of the iteration to the k-th iteration process; is a mapping coefficient, used to adjust the amplitude of the worst point mapping to balance the diversity and convergence of particle search; This represents the worst-case scenario in the k-th iteration. The worst-case scenario refers to the optimal position of the i-th particle in that iteration. The optimal position of the particle swarm The fitness function is the point with the worst fitness function value in the set consisting of the active target points and the initial values. The fitness function is constructed based on the average relative error between the model predictions and the measured values.

[0010] Furthermore, in the step of training the prediction model, the expression for the Gaussian radial kernel function is: ; in: The result is the calculation result of the Gaussian radial kernel function; The real-time input operating parameters are preprocessed in the real-time acquisition of operating parameters, including the current flue gas velocity and the target dust removal efficiency. The historical operating condition parameters corresponding to the training data in the database are used in the step of constructing the operating condition-dust removal efficiency database; g is the kernel function parameter; To input operating parameters in real time Compared with historical operating parameters The Euclidean distance.

[0011] Furthermore, in the training and prediction model step, the support vector machine base model adopts an ε-insensitive loss function, and the corresponding objective function is: ; And it satisfies the following constraints: ; in: This is the normal vector of the hyperplane in the support vector machine, used to determine the orientation of the hyperplane in the high-dimensional feature space; The bias of the hyperplane in the support vector machine is used to adjust the position of the hyperplane in the high-dimensional feature space; Let be the positive relaxation factor corresponding to the i-th training data. is the negative relaxation factor corresponding to the i-th training data, and together they are used to tolerate biases in the training data that exceed the ε-insensitive loss range; C is the penalty factor of the support vector machine model, used to balance the model's tolerance for training data bias and the generalization ability of the hyperplane; n is the total number of training data samples in the database during the step of constructing the working condition-dust removal efficiency database. In the step of constructing the working condition-dust removal efficiency database, the actual dust removal efficiency corresponding to the i-th group of training data in the database is required. In the step of constructing the operating condition-dust removal efficiency database, the operating condition parameters of the i-th group of training data in the database are... The mapping results in the high-dimensional feature space are used to transform low-dimensional operating parameters into high-dimensional features to adapt to linear classification. ε is the tolerance error of the ε-insensitive loss function, used to set the acceptable range of deviation between the predicted and actual values ​​of the model; i is the index of the training data, ranging from 1 to n.

[0012] Furthermore, in the real-time acquisition of operating parameters step, the preset frequency is 1 time / second; the preprocessing includes removing abnormal data where flue gas velocity fluctuations exceed ±0.05m / s, taking a 3-fold moving average of the data stream after removing abnormal data, and standardizing the data stream after the moving average to the [0,1] interval.

[0013] Furthermore, in the step of predicting the optimal operating voltage, the preset constraints are that the dust removal efficiency predicted by the model is ≥ the preset target dust removal efficiency - 0.5%, and the optimal operating voltage of each electric field output is ≤ 95% of the operating voltage of the traditional parallel plate ESP under the same working conditions; the response time of the model to calculate and output the optimal operating voltage is ≤ 0.3s.

[0014] Furthermore, in the execution control and closed-loop optimization steps, the preset order is to prioritize adjusting the first electric field voltage of the multi-field corrugated plate ESP, and then simultaneously adjust the second and third electric field voltages; the programmable logic controller drives the high-voltage power supply through the 485 communication protocol, and the voltage adjustment response time is ≤0.5s; the preset frequency for monitoring the actual dust removal efficiency after adjustment is 2 times / second, the preset threshold for the deviation is 0.5%, and the number of iterative adjustments is ≤3 times; the emergency conditions are a sudden increase in flue gas velocity ≥0.2m / s or a sudden increase in PM2.5 concentration ≥100μg / m³. 3 The automatic adjustment method is to increase the voltage of each electric field by 5%.

[0015] Compared with existing technologies, this method for precise control of the operating voltage of a multi-field corrugated plate ESP has the following advantages: I. This invention establishes an ATPSO-SVM prediction model with flue gas velocity and target dust removal efficiency as input variables and optimal operating voltages for each electric field as output variables. It combines this with a Gaussian radial kernel function to construct a support vector machine basic model and introduces an improved active target point particle swarm optimization algorithm to optimize model parameters. This effectively solves the technical problems of traditional multi-field ESP voltage control methods, such as difficulty in adapting to the nonlinear dust removal characteristics of equipment and lag in control response. The improved algorithm avoids getting trapped in local optima during parameter optimization by mapping active target points to worst-case points. The Gaussian radial kernel function accurately fits the complex relationship between operating parameters and voltage, enabling the model to respond to changes in operating conditions in real time and output an appropriate voltage control scheme. This significantly improves the operational stability of the equipment under dynamic operating conditions and ensures that the dust removal efficiency is always maintained within the target range.

[0016] II. This invention designs a dedicated control logic tailored to the structural characteristics of multi-field corrugated plate ESPs, constructing a complete control process of "real-time acquisition of operating parameters - model prediction of optimal voltage - execution of control and closed-loop optimization." This fully leverages the structural advantages of multi-field series connection and corrugated plates, while simultaneously achieving synergistic optimization of dust removal efficiency and energy consumption. During the control process, the influence of the corrugated plates on the electric field distribution can be specifically adapted, avoiding voltage redundancy caused by traditional control methods that do not consider the equipment structure. The closed-loop optimization mechanism can monitor dust removal efficiency deviations in real time and iteratively adjust them, ensuring the reliability of dust removal performance, reducing unnecessary energy waste, and quickly responding to sudden fluctuations in operating conditions.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 Flowchart of a method for precise control of the operating voltage of a multi-field corrugated plate ESP. Figure 2 Flowchart for building the operating condition-dust removal efficiency database; Figure 3 Flowchart for ATPSO-SVM model training and voltage prediction. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] Example 1: This embodiment takes the flue gas purification system of a 300MW coal-fired power plant as the application scenario. The system uses a multi-field corrugated plate ESP as the core dust removal equipment to treat the PM2.5-containing flue gas generated by the unit's combustion of pulverized coal. It needs to achieve the synergistic goal of efficient PM2.5 capture and equipment energy consumption optimization.

[0022] like Figure 1As shown, the overall control method of the present invention includes five core steps: constructing a database, training a prediction model, collecting operating parameters in real time, predicting the optimal operating voltage, and performing control and closed-loop optimization. The specific implementation process of this embodiment is described in detail below according to the steps of the present invention: Build the database: like Figure 2 As shown, the core structural parameters of this multi-field corrugated plate ESP were first determined. The device contains three sets of series-connected electric fields. The dust collection plates are corrugated plate designs, and the discharge electrode lines are uniformly arranged along the vertical direction of the corrugated plates to ensure that the electric field distribution is compatible with the corrugated plate structure. A flow velocity sensor, a PM2.5 concentration monitor, and a high-voltage power supply were selected. The flow velocity sensor is installed at the center of the main flue section at the ESP inlet to collect the inlet flow velocity of the flue gas. The PM2.5 concentration monitor is installed at the ESP outlet flue to monitor the concentration of fine particulate matter at the outlet to calculate the dust removal efficiency. The high-voltage power supply is equipped with an independent adjustable module for each electric field, supporting continuous adjustment of the operating voltage.

[0023] Based on the actual operating characteristics of the coal-fired unit, a typical industrial operating condition range was set, covering three core operating states: 50% rated load, 75% rated load, and 100% rated load. Each load state further covers combinations of different coal calorific values ​​and excess air coefficients to ensure coverage of common unit operating scenarios. For each operating condition combination, the unit was first allowed to run stably for one hour to stabilize the operating parameters. Then, the dust removal efficiency was continuously monitored for two hours, with flow velocity, PM2.5 concentration, and corresponding electric field voltage data recorded every 10 seconds. The arithmetic mean of all recorded data under the same operating condition was calculated, and abnormal data caused by transient sensor interference were removed to ensure data reliability.

[0024] All the sorted operating condition data and corresponding dust removal efficiency and electric field voltage data are classified and divided into training data and test data according to the proportion. The training data is used for subsequent model parameter optimization, and the test data is used to verify the model's generalization ability. At the same time, the operating condition type is labeled for each group of data to facilitate the analysis of operating condition correlation during subsequent model training, and finally a complete operating condition-dust removal efficiency database is formed.

[0025] Training the prediction model: like Figure 3 As shown, using the training data in the working condition-dust removal efficiency database constructed in the first step as the input source, an ATPSO-SVM prediction model architecture is built on an industrial control computer. The model input is set to two core parameters: flue gas velocity and target dust removal efficiency. The output corresponds to the optimal working voltage of three sets of electric fields, ensuring that the model directly matches the control requirements.

[0026] In the model building phase, a Gaussian radial kernel function is used to construct the basic support vector machine model. Leveraging the kernel function's ability to map low-dimensional operating parameters to high dimensions, the nonlinear relationship between flue gas velocity, target dust removal efficiency, and electric field voltage is fitted. Particularly addressing the local electric field intensity differences caused by corrugated plates, the kernel function is adjusted to enhance the model's adaptability to structural characteristics. An active target point particle swarm optimization algorithm is introduced, and its velocity update formula is improved to increase the diversity of particle search and avoid the problem of traditional particle swarm algorithms easily getting trapped in local optima. Specifically, active target points are added during algorithm iteration to guide particles towards the global optimum. Simultaneously, the average relative error between the model's predicted dust removal efficiency and the measured dust removal efficiency is used as the fitness function to quantify the model's prediction accuracy.

[0027] During iterative optimization, an ε-insensitive loss function is used to control the model's tolerance to errors, preventing overfitting of the training data. The initial number of iterations is set to 100. After each iteration, the fitness function value is calculated. If the change in the fitness function value is less than a preset threshold after 10 consecutive iterations, the model is considered converged. During this period, test data is used periodically to verify the model's performance. If the prediction error of the test data exceeds the allowable range, the initial value of the learning factor for the active target point particle swarm optimization algorithm is adjusted, and the iteration is restarted until the prediction errors of both the training and test data meet the control accuracy requirements. This completes the training of the ATPSO-SVM prediction model, which is then embedded into the control system of the industrial control computer.

[0028] Real-time acquisition of operating parameters: During normal operation of the coal-fired unit, the flow rate sensor and PM2.5 concentration monitor are started according to the preset procedure. After the equipment has preheated and entered a stable acquisition state, the inlet flue gas velocity and outlet PM2.5 concentration data of the ESP are continuously acquired at a frequency of 1 time / second. The data is transmitted to the database of the industrial control computer in real time.

[0029] The collected raw data undergoes preprocessing as follows: First, outlier removal is performed. Based on historical operating data of the unit, a reasonable fluctuation range for flow velocity is set, and outlier data caused by flue turbulence or momentary sensor malfunctions are identified and removed. Second, data smoothing is performed. The flow velocity and PM2.5 concentration data streams after outlier removal are processed using a moving average method. By averaging multiple consecutive data points, the impact of short-term fluctuations on parameter stability is reduced, avoiding control errors caused by momentary disturbances. Third, data standardization is performed. The flow velocity and PM2.5 concentration data after moving average are converted to the 0-1 range to eliminate the interference of different parameter magnitudes on the model input and ensure the consistency of model calculations.

[0030] Simultaneously, based on the emission standards and unit operation scheduling instructions issued by the local environmental protection department, the preset target dust removal efficiency is calculated. The current flue gas velocity after pretreatment, the actual dust removal efficiency at the outlet, and the preset target dust removal efficiency are integrated to form a set of real-time operating parameters that includes real-time operating conditions and target requirements, which are then pushed to the model input interface of the control system in real time.

[0031] Predicting the optimal operating voltage: The current flue gas velocity and the preset target dust removal efficiency, which are obtained from the real-time operating parameters formed in the third step, are used as the core input parameters and imported into the ATPSO-SVM prediction model trained in the second step. The model is then started to perform calculations through the computing module of the industrial control computer.

[0032] During the model calculation process, the built-in operating condition-voltage mapping relationship is first read. Combined with the input real-time flow rate and target efficiency, the voltage matching requirements of the three electric fields under the current operating condition are automatically analyzed. After the calculation is completed, the optimal operating voltage values ​​of the three electric fields are output, and the constraint conditions are judged at the same time: First, it checks whether the dust removal efficiency predicted by the model reaches the lower limit of the preset target dust removal efficiency to ensure that environmental protection emissions are met; second, it compares the output voltage with the voltage data of the traditional parallel plate ESP under the same operating condition, and controls the output voltage not to exceed the corresponding ratio of the traditional voltage to achieve energy consumption optimization; in addition, the model calculation response time is monitored to ensure that the time from parameter input to voltage output does not exceed the preset upper limit to avoid control lag due to calculation delay.

[0033] If the output voltage meets all constraints, the voltage data is marked as valid and pushed to the next control execution module; if the constraints are not met, the model automatically adjusts the calculation logic and re-outputs the optimal voltage value until the constraints are met.

[0034] Execution control and closed-loop optimization: The control execution module receives the effective voltage data output in the fourth step through the programmable logic controller (PLC). The PLC establishes a stable data connection with the high-voltage power supply of each electric field through the 485 communication protocol and performs operations according to the preset voltage regulation sequence: First, it sends a voltage regulation command to the high-voltage power supply of the first electric field and monitors the voltage feedback signal in real time. After the voltage of the first electric field stabilizes to the target value and the voltage fluctuation amplitude is less than the allowable range, it then sends regulation commands to the high-voltage power supply of the second and third electric fields simultaneously. This avoids grid load fluctuations and inter-field interference caused by simultaneous voltage regulation of multiple electric fields and ensures a smooth voltage regulation process.

[0035] After voltage adjustment, the PM2.5 concentration monitor continues to collect the outlet dust removal efficiency at a frequency of 1 time / second. The programmable logic controller (PLC) calculates the deviation between the actual dust removal efficiency and the preset target dust removal efficiency in real time. If the deviation is less than the preset threshold, the current voltage is determined to be suitable for the operating condition, and the equipment maintains the current voltage operation. If the deviation is greater than the preset threshold, the PLC automatically pushes the real-time operating parameters of the current moment back to the ATPSO-SVM prediction model. The model iteratively calculates the new optimal voltage value. After receiving the new voltage, the PLC repeats the voltage adjustment operation, and the number of iterative adjustments does not exceed 3 times to avoid infinite loops that could lead to unstable equipment operation.

[0036] If an emergency condition is triggered during operation, such as a sudden increase in PM2.5 concentration at the outlet exceeding the warning value, the programmable logic controller (PLC) immediately activates the emergency mechanism. It skips the model prediction stage and directly sends a voltage boost command to the high-voltage power supply of each electric field, rapidly increasing the voltage by a preset amount to suppress the risk of pollutant emissions exceeding the standard. At the same time, it sends an emergency alarm signal to the central control room to notify maintenance personnel to investigate the cause of the abnormal operating conditions. After the operating conditions return to stability, the PLC automatically switches back to the normal control mode and re-predicts the voltage through the model to ensure long-term stable operation of the equipment.

[0037] In summary, this embodiment applies the method for precise control of the operating voltage of a multi-field corrugated plate ESP to a 300MW coal-fired unit flue gas purification system. By constructing a working condition-dust removal efficiency database step by step, training the ATPSO-SVM prediction model, collecting and processing working condition parameters in real time, predicting the optimal operating voltage, and executing closed-loop control, the method fully adapts to the structural characteristics of the equipment's three sets of series electric fields and corrugated plates. During the process, the model accuracy is ensured by relying on the Gaussian radial kernel function and an improved active target point particle swarm optimization algorithm. Combined with data preprocessing, constraint condition determination, and emergency adjustment mechanisms, the method achieves stable dust removal efficiency under dynamic working conditions, effectively optimizes equipment energy consumption, and avoids interference and control lag issues caused by multi-field voltage regulation. This verifies the feasibility and practicality of the method in actual industrial scenarios.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for precise control of the operating voltage of a multi-field corrugated plate ESP, characterized in that, The specific steps of this method are as follows: Database construction: Determine the core structural parameters of the multi-field corrugated plate ESP, select flow velocity sensor, PM2.5 concentration monitor and high voltage power supply, set the typical industrial working condition range, continuously monitor the dust removal efficiency under each working condition combination and take the average value to form a working condition-dust removal efficiency database containing training data and test data. Training the prediction model: Using the training data in the database as input, a model architecture is built with flue gas velocity and target dust removal efficiency as input variables and the optimal working voltage of each electric field as output variables. A Gaussian radial kernel function is used to construct the basic support vector machine model. The active target point particle swarm optimization algorithm is introduced and its velocity update formula is improved. The average relative error between the predicted value and the measured value is used as the fitness function to iteratively optimize the support vector machine parameters and complete the training of the ATPSO-SVM prediction model. Real-time acquisition of operating parameters: The sensor is activated to collect the inlet flue gas velocity and the actual dust removal efficiency at the outlet of the ESP at a preset frequency. The collected data is preprocessed and the preset target dust removal efficiency is obtained to form a set of real-time operating parameters. Predicting the optimal operating voltage: Input the pre-processed current flue gas velocity and target dust removal efficiency into the trained ATPSO-SVM prediction model. The model calculates and outputs the optimal operating voltage values ​​of three sets of electric fields, and the output meets the preset constraints. Execution control and closed-loop optimization: The optimal operating voltage is received by the programmable logic controller, which drives the high-voltage power supply to adjust the voltage of each electric field in a preset order. After adjustment, the actual dust removal efficiency is monitored at a preset frequency, and the deviation from the target efficiency is calculated. If the deviation exceeds the preset threshold, the current operating parameters are re-input into the model for iterative control. If an emergency condition is triggered, the voltage of each electric field is automatically adjusted.

2. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the database construction step, the core structural parameters of the multi-field corrugated plate ESP are as follows: it includes three sets of series electric fields; the height of the corrugated plate in a single electric field is 10 mm, the corrugation bottom distance is 50 mm, the discharge electrode line radius is 1 mm, and the spacing between adjacent electrode lines is 40 mm; the accuracy of the flow velocity sensor is ±0.01 m / s, and the accuracy of the PM2.5 concentration monitor is ±1 μg / m³. 3 The regulation accuracy of the high-voltage power supply is ±0.1kV.

3. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the database construction step, the typical industrial operating conditions are as follows: flue gas velocity 0.6~1.2m / s, velocity gradient 0.1m / s; operating voltage 20~30kV, voltage gradient 0.5kV; the particle characteristics involved in the operating conditions are PM2.5, with a particle size range of 0.1~2.5μm and satisfying the Rosin-Rammler distribution, where the median particle size is 1.25μm and the distribution parameter is 1.64; the training data consists of 20 sets, and the test data consists of 7 sets.

4. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the training and prediction model step, the improved velocity update formula for the active target point particle swarm optimization algorithm is as follows: ； in: The velocity of the i-th particle in the d-th dimension during the (k+1)-th iteration; The inertia weights of the algorithm; Let be the velocity of the i-th particle in the d-th dimension during the k-th iteration; , , All of these are learning factors of the algorithm. , , All are random numbers with values ​​in the interval [0,1]. Let be the optimal position of the i-th particle in the d-th dimension during the k-th iteration; For the first The current position of the i-th particle in the d-th dimension in the next iteration; For the first The optimal position of the particle swarm in the d-th dimension during the next iteration; Let be the active target point of the i-th particle in the d-th dimension during the k-th iteration.

5. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 4, characterized in that, In the training and prediction model steps, active target points Generated through worst-case mapping, the mapping formula is: ； in: The active target point obtained after mapping has a value equal to the active target point. ; Let be the optimal position of the i-th particle in the d-dimensional space during the k-th iteration; This represents the optimal position of the entire particle swarm in the d-dimensional space during the k-th iteration. These are the mapping coefficients; This represents the worst point in the k-th iteration.

6. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the training and prediction model step, the expression for the Gaussian radial kernel function is: ； in: The result is the calculation result of the Gaussian radial kernel function; This refers to the preprocessed real-time input operating condition parameters in the real-time acquisition of operating condition parameters step; The historical operating condition parameters corresponding to the training data in the database are used in the step of constructing the operating condition-dust removal efficiency database; g is the kernel function parameter; To input operating parameters in real time Compared with historical operating parameters The Euclidean distance.

7. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the training and prediction model step, the support vector machine base model uses an ε-insensitive loss function, and the corresponding objective function is: ； And it satisfies the following constraints: ； in: The normal vector of the hyperplane of the support vector machine; For the bias of the hyperplane of the support vector machine; Let be the positive relaxation factor corresponding to the i-th training data. is the negative relaxation factor corresponding to the i-th group of training data; C is the penalty factor of the support vector machine model; n is the total number of training data samples in the database during the step of constructing the working condition-dust removal efficiency database; In the step of constructing the working condition-dust removal efficiency database, the actual dust removal efficiency corresponding to the i-th group of training data in the database is required. In the step of constructing the operating condition-dust removal efficiency database, the operating condition parameters of the i-th group of training data in the database are... The mapping result in the high-dimensional feature space; ε represents the tolerance error of the ε-insensitive loss function; i is the index of the training data.

8. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the real-time acquisition of operating parameters, the preset frequency is 1 time / second; the preprocessing includes removing abnormal data where flue gas velocity fluctuations exceed ±0.05m / s, taking a 3-fold moving average of the data stream after removing abnormal data, and standardizing the data stream after the moving average to the [0,1] interval.

9. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the step of predicting the optimal operating voltage, the preset constraints are that the dust removal efficiency predicted by the model is ≥ the preset target dust removal efficiency - 0.5%, and the optimal operating voltage of each electric field output is ≤ 95% of the operating voltage of the traditional parallel plate ESP under the same working conditions; the response time of the model to calculate and output the optimal operating voltage is ≤ 0.3s.

10. The method for precise control of the working voltage of a multi-field corrugated plate ESP according to claim 1, characterized in that, In the execution control and closed-loop optimization steps, the preset order is to first adjust the voltage of the first electric field of the multi-field corrugated plate ESP, and then simultaneously adjust the voltages of the second and third electric fields; the programmable logic controller drives the high-voltage power supply through the 485 communication protocol, and the voltage adjustment response time is ≤0.5s; the preset frequency for monitoring the actual dust removal efficiency after adjustment is 2 times / second, the preset threshold for the deviation is 0.5%, and the number of iterative adjustments is ≤3 times; the emergency conditions are a sudden increase in flue gas velocity ≥0.2m / s or a sudden increase in PM2.5 concentration ≥100μg / m³. 3 The automatic adjustment method is to increase the voltage of each electric field by 5%.

Images