A DCS electric dust removal system optimization control method and device

Abstract

The application provides a DCS electric dust removal system optimization control method and device, relates to the field of data processing, and comprises a data acquisition module, an edge calculation module, a concentration prediction unit, a parameter optimization unit, a safety check unit, a communication issuing module and a Web terminal display module.The edge calculation module can quickly output a dust emission concentration prediction value according to real-time working condition characteristics.The parameter optimization unit adopts an improved particle swarm algorithm to perform second-level optimization on the high-voltage power supply set value, can ensure energy minimization on the premise that the predicted dust emission concentration does not exceed the standard, and avoids stealing sparks between electric fields through a frequency consistency penalty term.The optimization result is directly issued by a DCS communication module after safety check, closed-loop control is realized, and thus stable operation of the electric dust removal system can be realized without increasing an online analyzer.

Description

Technical Field

[0001] This invention relates to data processing technology, and more particularly to an optimization control method and apparatus for a DCS electrostatic precipitator system. Background Technology

[0002] As the core terminal treatment process for particulate matter control in coal-fired flue gas, the operating efficiency of electrostatic precipitators directly determines whether the dust concentration at the chimney outlet can stably meet the national ultra-low emission standards. Its energy consumption level is also a core component of the power plant power control system. The intelligent optimization control technology of electrostatic precipitators has become a key technology for the thermal power industry to balance environmental protection compliance, energy conservation and carbon reduction.

[0003] Currently, existing electrostatic precipitator control schemes mainly rely on online analyzers to measure dust emission concentrations and then adjust parameters based on the measured values. In such schemes, static optimization boundaries are generally used when optimizing parameters, which cannot adaptively adjust based on the deviation between the real-time predicted emission concentration and the target value. When the boiler's operating conditions fluctuate drastically, this static optimization method can easily lead to a serious mismatch between the optimization space and the actual control requirements: if the optimization boundary is too wide, it will significantly increase the computational load of the algorithm, prolong the optimization time, and even generate dangerous parameters that reach the safe operating limits of the equipment; if the optimization boundary is too narrow, it will significantly compress the optimization space, causing the algorithm to easily get trapped in local optima and fail to minimize energy consumption while ensuring stable emission compliance.

[0004] Therefore, finding a set of optimal high-voltage power supply operating parameters in a short period of time, so as to minimize energy consumption while ensuring that the predicted dust emission concentration does not exceed the standard, has become an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a DCS electrostatic precipitator system optimization control method and device, which can find a set of optimal high-voltage power supply operating parameters in a short time, so as to ensure energy consumption is minimized while ensuring that the predicted dust emission concentration does not exceed the standard.

[0006] A first aspect of the present invention provides an optimization control method for a DCS electrostatic precipitator system, comprising:

[0007] The operating data of the electrostatic precipitator is acquired based on a preset acquisition cycle, and the operating data is preprocessed to obtain standard operating data.

[0008] The dust emission concentration is predicted based on the standard operating data, and the dynamic optimization boundary is determined based on the concentration difference between the dust emission concentration and the target emission concentration.

[0009] The optimal set of control parameters is obtained by performing global optimization within the dynamic optimization boundary based on the particle swarm optimization algorithm.

[0010] The optimal control parameter set is verified by the safety verification unit. Once the verification is successful, the parameter set is sent to the high-voltage control cabinet for execution.

[0011] Optionally, in one possible implementation of the first aspect, the step of predicting the dust emission concentration based on the standard operating data and determining the dynamic optimization boundary based on the concentration difference between the dust emission concentration and the target emission concentration includes:

[0012] Using the control parameter values ​​corresponding to the current standard operating data as a reference point, the direction of parameter adjustment is determined according to the sign of the concentration difference, and the search range width is determined according to the absolute value of the concentration difference.

[0013] Centered on the reference point, the search range is expanded within the width of the search range along the parameter adjustment direction and intersected with the physical limit parameters of the electrostatic precipitator to generate the dynamic optimization boundary.

[0014] The dynamic optimization boundary is recalculated and updated every preset collection cycle based on the concentration difference at the current moment.

[0015] Optionally, in one possible implementation of the first aspect, the particle swarm optimization algorithm performs global optimization within the dynamic optimization boundary to obtain the optimal set of control parameters, including:

[0016] The frequency coordination term is determined based on the degree of dispersion between the operating frequencies of different power supply units within the same electric field, and a fitness function is constructed based on the absolute value of the frequency coordination term and the concentration difference.

[0017] The particle swarm is initialized within the dynamic optimization boundary, and the particles are iteratively calculated based on the fitness function to update the individual extreme value and the population extreme value of each particle.

[0018] When the preset iteration termination condition is met, the parameter combination corresponding to the population extreme value is output to obtain the optimal control parameter set.

[0019] Optionally, in one possible implementation of the first aspect, determining the frequency coordination term based on the degree of dispersion between the operating frequencies of different power supply units within the same electric field, and constructing a fitness function based on the absolute value of the frequency coordination term and the concentration difference, includes:

[0020] Based on the secondary voltage setpoint, secondary current setpoint, and high frequency power supply frequency setpoint of each electric field high voltage power supply, a set of control parameters to be optimized is obtained;

[0021] The operating frequencies of different power supply units within the same electric field during historical operating periods are obtained, and the average historical frequency difference between each power supply unit is calculated based on the operating frequencies.

[0022] Calculate the current frequency difference based on the frequency set value corresponding to different power supply units in the same electric field from any set of control parameters.

[0023] The degree of deviation between the current frequency difference and the average historical frequency difference is used as a frequency coordination term, and the absolute value of the concentration difference is weighted and summed with the frequency coordination term to construct a fitness function.

[0024] Optionally, in one possible implementation of the first aspect, the initialization of the particle swarm within the dynamic optimization boundary and the iterative calculation of the particles based on the fitness function to update the individual extreme value and the population extreme value of each particle includes:

[0025] Within the dynamic optimization boundary, a particle swarm of a preset population size is randomly generated, wherein the position vector of each particle in the particle swarm corresponds to a set of control parameters.

[0026] For each iteration, the particle's position vector is substituted into the fitness function, and the particle's fitness value is calculated.

[0027] If the fitness value is better than the particle's historical best value, then the particle's current position vector is updated to the particle's individual extreme value;

[0028] The optimal value of the population is determined based on the comparison results between the fitness values ​​of each particle.

[0029] If the fitness value is better than the population optimum, then the particle's current position vector is updated to the population optimum.

[0030] Based on the individual extreme value, the population extreme value, and the current position vector, the optimization speed of each particle is updated. According to the updated optimization speed and the current position vector, the position vector of each particle is updated, and the updated position vector is constrained within the dynamic optimization boundary.

[0031] Optionally, in one possible implementation of the first aspect, determining the population optimum based on the comparison results between the fitness values ​​of each particle includes:

[0032] During the initialization phase, the initial fitness value of all particles is calculated, and the smallest fitness value is taken as the initial population optimum.

[0033] In each iteration round, after each particle completes its individual extreme value update, the fitness values ​​corresponding to the individual extreme values ​​of all particles are compared, and the minimum fitness value is determined as the optimal fitness value for the current iteration round.

[0034] If the optimal fitness value of the current iteration is better than the optimal value of the population, then the optimal value of the population is updated to the optimal fitness value of the current iteration.

[0035] Optionally, in one possible implementation of the first aspect, the step of outputting the parameter combination corresponding to the population extreme value when a preset iteration termination condition is met, to obtain the optimal control parameter set, includes:

[0036] If the number of iterations reaches the maximum number of iterations, the iteration is terminated, and the set of control parameters corresponding to the current population extreme value is output as the optimal set of control parameters.

[0037] If the number of iterations does not reach the maximum number of iterations, or if the optimal value of the population is less than the preset fitness threshold, the iteration is terminated, and the set of control parameters corresponding to the current extreme value of the population is output as the optimal set of control parameters.

[0038] If the optimal value of the population is greater than or equal to the fitness threshold, then continue to the next iteration.

[0039] Optionally, in one possible implementation of the first aspect, the step of performing a security verification on the optimal control parameter set based on the security verification unit, and then sending the verification result to the high-voltage control cabinet for execution, includes:

[0040] Each control parameter in the optimal control parameter set is checked for exceeding the limit. If any control parameter exceeds the dynamic optimization boundary, it is marked as an exceeding the limit error, and the distribution is terminated.

[0041] A mutation check is performed on the optimal control parameter set. The optimal control parameter set is compared with the control parameter set issued in the previous cycle. If the adjustment range of any control parameter exceeds the range threshold, it is marked as a mutation error and the issuance is terminated.

[0042] Gradient verification is performed on the optimal control parameter set. If the secondary voltage difference between adjacent electric fields exceeds the voltage difference threshold, it is marked as a gradient error and the transmission is terminated.

[0043] If the over-limit check, mutation check, and gradient check all pass, the optimal control parameter set will be sent to the high-voltage control cabinet for execution.

[0044] If any check fails, the control parameter set issued in the previous cycle remains unchanged, and a safety alarm is triggered.

[0045] Optionally, in one possible implementation of the first aspect, it also includes:

[0046] When the security verification unit receives a cut-out command, it stops issuing the optimal control parameter set and switches to traditional spark tracking mode to continue operating until it receives a cut-in command and resumes issuing the optimal control parameter set.

[0047] A second aspect of the present invention provides an optimization control device for a DCS electrostatic precipitator system, comprising:

[0048] The data acquisition module is used to acquire the operating data of the electrostatic precipitator based on a preset acquisition cycle, and to preprocess the operating data to obtain standard operating data;

[0049] The concentration prediction module is used to predict the dust emission concentration based on the standard operating data, and to determine the dynamic optimization boundary based on the concentration difference between the dust emission concentration and the target emission concentration.

[0050] The parameter optimization module is used to perform global optimization within the dynamic optimization boundary based on the particle swarm optimization algorithm to obtain the optimal set of control parameters.

[0051] The verification and distribution module is used to perform a safety verification on the optimal control parameter set according to the safety verification unit, and after the verification is passed, it is distributed to the high-voltage control cabinet for execution.

[0052] The beneficial effects of this invention are as follows:

[0053] 1. This invention can collect and standardize the operating data of existing DCS systems, and accurately predict the dust emission concentration based on the collected operating data. It can reduce the energy consumption of electrostatic precipitators while ensuring that the predicted dust emission concentration does not exceed the standard.

[0054] 2. This invention can dynamically define the optimization boundary by the concentration difference between the dust emission prediction result and the target value. Within the dynamic optimization boundary, the particle swarm optimization algorithm can complete the global optimization of multi-dimensional control parameters in a short time, and can perform more refined optimization of the control parameters of the electrostatic precipitator, thereby quickly determining the best combination of control parameters and improving dust removal efficiency.

[0055] 3. This invention can achieve safety constraints on optimization parameters through a multi-level safety verification mechanism. The optimal parameters that pass the verification can be directly sent to the high-voltage control cabinet for execution without manual input. This enables real-time closed-loop operation of optimization control and the field DCS system, improving the safety of equipment operation. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the overall structure of an optimization control device in an electrostatic precipitator DCS system provided by an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the internal software flow of the edge computing module of an electrostatic precipitator system optimization control device provided in an embodiment of the present invention;

[0058] Figure 3This is a schematic flowchart of an optimization control method for a DCS electrostatic precipitator system provided in an embodiment;

[0059] Figure 4 This is a flowchart illustrating a particle swarm optimization algorithm provided in an embodiment of the present invention;

[0060] Figure 5 This is a schematic diagram of the structure of a DCS electrostatic precipitator optimization control device provided in an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0063] See Figure 1 This is a schematic diagram of the overall structure of an optimization control device in an electrostatic precipitator DCS system according to an embodiment of the present invention. Figure 1 The method shown can be implemented by software and / or hardware devices. For example... Figure 1 As shown, the device can read real-time operating data of the electrostatic precipitator from the main ring network switch of the power plant's DCS through a data acquisition module with redundant dual network ports. After preprocessing, one path is transmitted to the edge computing module with a built-in adaptive PSO optimization algorithm via USB / SFTP, and the other path is synchronized to the web-based display module via WebSocket, realizing real-time visualization of the operating status. After the edge computing module completes the second-level emission concentration prediction and global optimization of the high-voltage power supply control parameters, it outputs the optimal control parameter set to the safety verification unit. After passing the triple forced safety verification of over-limit, sudden change, and gradient, the parameters that have passed the verification can be written into the existing high-voltage control cabinet of the power plant for execution by the redundant RS-485 communication module with opto-isolation of 2.5kV or above through the industrial general Modbus-RTU protocol. Finally, a control link of data acquisition, emission prediction, parameter optimization, safety verification, and closed-loop execution is formed, which can take into account the goals of ultra-low emission compliance of the electrostatic precipitator system, equipment operation safety, and energy saving and consumption reduction.

[0064] See Figure 2is a schematic diagram of the internal software process of the edge computing module of the optimal control device for the electrostatic precipitation system provided by the embodiment of the present invention, as Figure 2 shown in Figure 2 fully demonstrates the full-cycle closed-loop operation logic of the present invention: after the process starts, it first reads the multi-dimensional real-time working condition data of the electrostatic precipitator and the boiler and completes standardized preprocessing, then the corresponding dust emission concentration can be predicted according to the standard working condition data, then the global optimization of the high-voltage power supply control parameters is completed through the adaptive PSO algorithm based on the predicted value, after outputting the optimal control quantity, triple safety checks of over-limit, sudden change and gradient are performed. If the check is passed, the optimal control quantity is issued to the high-voltage control cabinet for execution via the Modbus-RTU protocol. If the check fails, the issuance is frozen and the parameter value of the stable operation in the previous cycle is maintained. Finally, the emission predicted value and energy saving rate of the current cycle are synchronously updated to the Web end, then the process returns to the data collection link to enter the next cycle, realizing second-level closed-loop intelligent optimal control under any working condition.

[0065] refer to Figure 3 , which is a schematic flowchart of the optimal control method for the DCS electrostatic precipitation system provided by the embodiment of the present invention, Figure 3 the execution subject of the method shown may be a software and / or hardware device. The execution subject of the present application may include but is not limited to at least one of the following: user equipment, network equipment, etc. The user equipment may include but is not limited to a computer, a smart phone, a Personal Digital Assistant (PDA for short) and the above-mentioned electronic devices. The network equipment may include but is not limited to a single network server, a server group composed of multiple network servers, or a cloud composed of a large number of computers or network servers based on cloud computing, wherein cloud computing is a type of distributed computing, which is a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose restrictions on this. The method includes steps S1 to S4, which are specifically as follows:

[0066] S1: Acquire the operation data of the electrostatic precipitation device based on a preset acquisition cycle, and preprocess the operation data to obtain standard operation data.

[0067] The preset acquisition cycle refers to a preset fixed data refresh cycle, which can be 1s for example. The electrostatic precipitation device can be a double-chamber multi-electric-field high-voltage electrostatic precipitation device supporting a coal-fired unit. The operation data refers to multi-dimensional real-time data covering the full electric field operation state of the electrostatic precipitator and the boiler working condition, which can specifically include core parameters such as the primary voltage or current and the secondary voltage or current of each electric field. The standard operation data refers to the standardized working condition data after preprocessing.

[0068] It is understandable that the raw operating data collected by the power plant's DCS system may have issues such as asynchronous timing, abnormal jumps caused by instrument malfunctions, and missing data. Directly using this data for subsequent predictions may lead to severely distorted emission prediction results, causing the parameter optimization process to deviate from the actual operating conditions, ultimately resulting in excessive emissions or wasted energy. At the same time, electrostatic precipitators need to adapt to the rapid fluctuations in the boiler's operating state. Scattered data collection without a fixed cycle may not guarantee the timing consistency of second-level closed-loop control, easily leading to control actions lagging behind changes in operating conditions. Therefore, by collecting and preprocessing data with a fixed collection cycle, a stable and accurate data source can be provided to the control device to ensure the accuracy and stability of the closed-loop control process.

[0069] Specifically, the data acquisition module can be connected to the ring network switch of the power plant's DCS system simultaneously through two independent Ethernet ports. According to the preset acquisition cycle, such as 1 second, it can continuously read the multi-dimensional real-time operating data of the electrostatic precipitator. Simultaneously, preprocessing operations are performed on the acquired raw operating data. For example, abnormal data during power-on or power-off, instrument failure periods, and jump data exceeding the physical limits of the equipment can be removed first. Then, the effective values ​​of the previous acquisition cycle are used to fill the missing data points. Finally, the parameters of different refresh cycles are time-series aligned to obtain standard operating data with unified timing and complete data.

[0070] S2, predict the dust emission concentration based on the standard operating data, and determine the dynamic optimization boundary based on the concentration difference between the dust emission concentration and the target emission concentration.

[0071] Among them, dust emission concentration refers to the dust mass concentration at the outlet chimney of the electrostatic precipitator; target emission concentration refers to the preset standard dust concentration value, which is set to simultaneously consider ultra-low emission compliance and energy conservation and consumption reduction requirements, such as 3 mg / Nm³; concentration difference refers to the absolute difference between the predicted dust emission concentration value and the preset target emission concentration; and dynamic optimization boundary refers to the parameter range for searching the optimal high-pressure control parameters by adjusting in real time according to emission prediction deviation and on-site operating conditions.

[0072] Existing electrostatic precipitator (ESP) control solutions in the thermal power industry generally rely on measured data from online emission analyzers for parameter adjustments. However, mainstream analyzers have a long measurement lag, making it impossible to achieve proactive emission concentration control. This can easily lead to situations where emissions exceed standards and require remedial measures. To avoid the risk of exceeding emission standards, operators often adopt a conservative strategy of excessive voltage boosting, resulting in high energy consumption in the ESP system. This solution can predict dust emission concentrations based on standard operating data and dynamically define a safe optimization boundary based on the prediction deviation, thereby achieving refined optimization of the control parameters of the ESP equipment.

[0073] After obtaining the standard operating data, the operating data within the historical time period can be extracted from the standard operating data and identified as historical operating data. The historical dust emission concentration under the historical operating data can be retrieved, and a correspondence database between historical operating data and historical dust emission concentration can be established. Then, the current standard operating data can be compared with the historical operating data in the established correspondence database. For example, the similarity between the current standard operating data and each historical operating data can be calculated using existing methods based on Euclidean distance. Then, historical operating data with a similarity greater than or equal to the similarity threshold can be selected. The measured dust emission concentration values ​​corresponding to the selected historical operating data are then weighted and averaged according to the similarity from high to low. The higher the similarity, the greater the corresponding weight. The calculated weighted average value can then be used as the predicted dust emission concentration under the current standard operating data, i.e., the dust emission concentration.

[0074] After obtaining the predicted dust emission concentration, the concentration difference between the predicted dust emission concentration and the target emission concentration can be calculated. Using the current real-time operating control parameters as a reference point, the direction of parameter adjustment is determined according to the sign of the concentration difference, and the search range width is determined according to the absolute value. After forming the initial search interval, the intersection with the physical limit parameters of the equipment is taken to generate a dynamic optimization boundary. This boundary can be updated in real time with the latest equipment operating data every preset collection cycle. Finally, the predicted value and the dynamic boundary are synchronously transmitted to the parameter optimization module.

[0075] Based on the above embodiments, the dynamic optimization boundary can be determined in the following way:

[0076] S21, using the control parameter value corresponding to the current standard operating data as a reference point, determine the parameter adjustment direction according to the sign of the concentration difference, and determine the search range width according to the absolute value of the concentration difference.

[0077] Among them, the control parameter value refers to the specific parameter value of the electrostatic precipitator high-voltage power supply that is currently running in the DCS system in real time; the parameter adjustment direction refers to the parameter optimization direction determined by the positive or negative value of the emission concentration difference; and the search range width refers to the size of the upper and lower fluctuation range of parameter optimization based on the current control parameter value.

[0078] After obtaining the standard operating data, the corresponding control parameter values ​​can be read and set as the benchmark for parameter optimization. This ensures that the optimization process is based on the current stable operating parameters, avoiding significant jumps. Then, the absolute concentration difference between the predicted dust emission concentration and the preset target emission concentration can be calculated. The direction of parameter adjustment is determined by the sign of the concentration difference. For example, when the concentration difference is positive, it means the predicted emission concentration is higher than the target emission concentration, posing a risk of exceeding environmental standards. Therefore, the parameter adjustment direction can be determined as increasing the concentration upwards, which means improving dust removal efficiency, i.e., increasing the secondary voltage / electricity. The direction of adjusting the flow and optimizing the high-frequency frequency to enhance the dust removal effect is as follows: When the concentration difference is negative, it means that the predicted emission concentration is lower than the target emission concentration, and there is room for energy saving. Therefore, the parameter adjustment direction is determined to be downward, which is the direction of energy saving and consumption reduction. That is, the adjustment is made towards reducing the secondary voltage / current and optimizing the high-frequency frequency to reduce energy consumption. Then, the search range width is determined according to the absolute value of the concentration difference. The larger the absolute value, the more serious the emission deviation, and a wider search range can be set to achieve rapid correction of deviation. The smaller the absolute value, the closer the emission is to the target value, and a narrower search range can be set to achieve refined energy saving optimization.

[0079] S22, with the reference point as the center, expand within the search range width along the parameter adjustment direction, and intersect with the physical limit parameters of the electrostatic precipitator to generate the dynamic optimization boundary.

[0080] Among them, physical limit parameters refer to the upper and lower limits of the safe operating parameters that are rated and cannot be exceeded when the high-voltage power supply hardware equipment of the electrostatic precipitator is manufactured. For example, it may include the rated operating range of secondary voltage 25-72kV and the rated operating range of high frequency power supply frequency 4500-7500Hz.

[0081] Specifically, the control parameter value can be used as a reference point, and the search range can be expanded vertically and vertically within the set search range width along the determined parameter adjustment direction to form an initial parameter search interval. Then, the physical limit parameters of the electrostatic precipitator high-voltage equipment can be retrieved, including the rated operating range of secondary voltage 25-72kV and high frequency 4500-7500Hz. The upper and lower limits of the initial search interval are compared with the upper and lower limits of the corresponding physical limit parameters one by one, and the overlapping interval between them is taken as the final search interval of that dimension parameter. For example, if the upper limit of the secondary voltage in the initial interval is 75kV, and the upper limit of the physical limit is 72kV, then 72kV is taken as the final upper limit of the parameter. By integrating the final search intervals of the control parameters, a complete dynamic optimization boundary can be formed.

[0082] S23, based on the concentration difference at the current moment, recalculate and update the dynamic optimization boundary every preset collection cycle.

[0083] Understandably, during normal operation of a coal-fired unit, operating data such as boiler load, coal quality, flue gas volume, and flue gas temperature are constantly fluctuating in real time. The corresponding predicted dust emission concentration will also change synchronously with the operating data. If the dynamic optimization boundary remains fixed after it is generated, it may quickly become disconnected from the real-time operating conditions, causing subsequent parameter optimization results to be unable to adapt to the current operating state, resulting in problems such as excessive emissions or energy waste. At the same time, the closed-loop control cycle should be synchronized with the preset acquisition cycle set in step S1. If the optimization boundary is fixed, it may not be able to achieve synchronization with data acquisition and emission prediction. Therefore, the optimization boundary can be continuously updated according to the latest emission concentration difference based on the preset acquisition cycle to ensure that the boundary is always synchronized with the real-time operating conditions and emission deviations.

[0084] Specifically, a preset data acquisition cycle can be used as the fixed update cycle. At the beginning of each cycle, the latest standard operating data and dust emission concentration prediction value are first obtained. The absolute concentration difference between the predicted dust emission concentration value and the target emission concentration is recalculated. Based on the latest concentration difference, the parameter adjustment direction and search range width are redefined according to the process in step S21. The current real-time control parameter reference point is updated synchronously. Then, according to the process in step S22, based on the updated reference point, adjustment direction, and search range width, and combined with the physical limit parameters of the equipment, a dynamic optimization boundary adapted to the current operating conditions is regenerated. The newly generated dynamic optimization boundary replaces the old boundary of the previous cycle and is synchronously updated to the parameter optimization module to ensure that subsequent parameter optimization is always performed according to the latest boundary constraints. This process is repeated until the entire electrostatic precipitator control process terminates, always ensuring that the dynamic optimization boundary is synchronously adapted to the real-time operating conditions and emission deviations.

[0085] S3. Based on the particle swarm optimization algorithm, global optimization is performed within the dynamic optimization boundary to obtain the optimal set of control parameters.

[0086] Among them, the particle swarm optimization algorithm refers to a global optimization algorithm that makes multi-dimensional customized improvements to the traditional particle swarm optimization algorithm for scenarios involving high-dimensional coupled control parameters, multiple electric field coupling, and cooperative constraints of the same electric field frequency in electrostatic precipitators. This algorithm simulates the group cooperation and information sharing mechanism of birds foraging, treating each set of control parameters to be optimized as a bird in the search space, i.e., a particle, and the optimization objectives, such as emission compliance, equipment stability, and optimal energy consumption, as the food source, i.e., the global optimal solution. Using the dual-objective fitness function constructed in this scheme as the evaluation criterion, under the rigid constraints of the dynamic optimization boundary, the global optimization of high-dimensional parameters can be completed in a short time through iterative optimization of particles, individual memory, and group information interaction. The optimal control parameter set refers to the combination of high-voltage power supply control parameters obtained through global optimization that takes into account emission compliance, equipment operation stability, and optimal energy consumption.

[0087] With the widespread adoption of high-frequency power supplies, the dimensions of electrostatic precipitator control parameters have also changed to higher dimensions. The coupling effect of multiple electric fields makes it impossible for traditional manual tuning and single closed-loop control to find the global optimal solution, which is prone to local optimization and sparking problems between electric fields. Traditional intelligent optimization algorithms have slow convergence speed, cannot meet the second-level closed-loop requirements, and are prone to getting trapped in local optima. Therefore, particle swarm optimization algorithm can be used to achieve second-level global optimization of high-dimensional control parameters and output the optimal control parameters that can simultaneously meet the requirements of compliance, stability and energy saving.

[0088] Specifically, the following steps are included:

[0089] S31, determine the frequency coordination term based on the degree of dispersion between the operating frequencies of different power supply units in the same electric field, and construct a fitness function based on the absolute value of the frequency coordination term and the concentration difference.

[0090] Among them, the operating frequency refers to the real-time operating frequency of different high-frequency power supply units in the same electric field of the electrostatic precipitator; the dispersion degree refers to the degree of deviation between the operating frequencies of different power supply units in the same electric field; the greater the dispersion degree, the easier it is to have sparking phenomena, resulting in frequent electric field flashover; the frequency coordination term refers to a numerical term that can represent the degree of deviation of the frequency of different power supply units in the same electric field from the historical steady-state mean; and the fitness function refers to the evaluation standard for judging the quality of the combination of control parameters in the particle swarm optimization algorithm, which achieves a balance between multiple objectives such as emission compliance, equipment stability, and optimal energy consumption through calculation.

[0091] Understandably, with the widespread adoption of high-frequency power supplies, the adjustable control parameters of electrostatic precipitators have jumped to 24 dimensions. There are obvious space charge shielding and secondary dust coupling effects between various electric fields. Traditional single-objective optimization algorithms only use the lowest emission concentration as the evaluation criterion, which can easily lead to excessive frequency dispersion of different power supply units in the same electric field, causing sparking phenomena, resulting in frequent flashover and downsizing of the electric field, and reducing the overall dust removal efficiency. This solution can construct a fitness function with dual objectives, including frequency coordination terms and emission deviation, to ensure that the emission concentration is close to the target value and to constrain the frequency dispersion to ensure equipment stability.

[0092] Specifically, based on the historical steady-state operation data of the power plant, the coordination pattern of the operating frequencies of different power supply units within the same electric field can be analyzed to determine the reasonable range of frequency dispersion. Then, based on this range, the calculation method of the frequency coordination term can be defined, the deviation between the current frequency dispersion and the reasonable range can be calculated, and then combined with the absolute value of the obtained concentration difference, a fitness function can be constructed. It is clear that the smaller the function value, the better the combination of control parameters. This ensures that dust emissions are close to the target value while constraining the frequency dispersion to avoid sparking.

[0093] In some embodiments, step S31 may be implemented as follows:

[0094] S311, based on the secondary voltage setting value, secondary current setting value and high frequency power supply frequency setting value of each electric field high voltage power supply, obtains the set of control parameters to be optimized.

[0095] Among them, the set of control parameters to be optimized refers to the combination of core controllable parameters that have a direct impact on the dust removal efficiency, equipment stability and energy consumption of electrostatic precipitators, and is the specific optimization object of the particle swarm algorithm.

[0096] Specifically, by sorting out the controllable operating parameters of the high-voltage power supply of the electrostatic precipitator, parameters that directly affect dust removal efficiency, equipment stability, and energy consumption can be screened out. First, the secondary voltage setting value and secondary current setting value of the high-voltage power supply in each electric field can determine the dust removal intensity of the electric field. Second, the frequency setting value of the high-frequency power supply can affect the discharge stability of the electric field and is a related parameter for the spark problem. Then, according to the electric field zone and power supply unit classification, a complete set of control parameters to be optimized can be formed.

[0097] S312, obtain the operating frequency of different power supply units in the same electric field during the historical operating period, and calculate the average historical frequency difference between each power supply unit based on the operating frequency.

[0098] Among them, the historical frequency difference mean refers to the arithmetic mean of the operating frequency difference of different power supply units in the same electric field during the historical steady-state operation period.

[0099] Since the function of the frequency coordination term is to suppress the phenomenon of sparking, and the essence of sparking is the frequency difference between different power supply units in the current electric field, which deviates from the reasonable range for stable operation of the equipment, a benchmark value that fits the characteristics of the equipment on site is needed to accurately determine the degree of deviation of the current frequency difference. Therefore, the most recent historical time period of the electrostatic precipitator can be retrieved from the power plant's DCS historical database, such as the operating data of the last 3 months. Effective steady-state operating condition data with stable boiler load, compliant emissions, and no frequent flashovers can be selected. The real-time operating frequency of two power supply units in the same electric field can be extracted according to the electric field zone and power supply unit classification. The absolute frequency difference is calculated in real time to form a historical frequency difference dataset. After removing abnormal jump data, the arithmetic mean is calculated to obtain the historical frequency difference mean between each power supply unit in each electric field.

[0100] S313 calculates the current frequency difference based on the frequency set values ​​corresponding to different power supply units in the same electric field within any set of control parameters.

[0101] The current frequency difference refers to the absolute difference between the frequency setpoints of different high-frequency power supply units within the same electric field in any set of control parameters to be evaluated.

[0102] Understandably, the essence of the frequency coordination term is the degree of deviation between the current frequency difference and the historical steady-state benchmark. If the current frequency difference corresponding to any set of control parameters cannot be obtained, the quantitative calculation of the degree of deviation cannot be completed, the frequency coordination term cannot be generated, and the fitness function cannot achieve a complete evaluation of the parameter combination. At the same time, the control parameter combination corresponding to each particle in the particle swarm optimization algorithm will change dynamically in each iteration, and its corresponding frequency setpoint and frequency difference will also change synchronously. Only by calculating the current frequency difference for each set of parameter combinations in real time can we accurately determine whether there is a risk of sparking for that set of parameters, ensuring that the evaluation result of the fitness function is true and effective, and can accurately reflect the parameter set. To optimize the overall performance of the particle swarm optimization algorithm, for any control parameter set corresponding to a particle, the high-frequency power supply frequency setpoint is first selected from the parameter set. Then, according to the electric field partitioning order and the correspondence between the power supply unit and the frequency setpoint, the electric field is matched with the frequency setpoint. Next, the electric field is grouped, and the high-frequency power supply frequency setpoints corresponding to different power supply units within the same electric field are paired to form frequency pairing data corresponding to each electric field. For each pair of frequency setpoints, the absolute difference between the two frequency values ​​is calculated, which is the current frequency difference corresponding to that electric field. This calculation process is repeated to complete the calculation of the current frequency difference for all electric fields, ultimately obtaining the current frequency difference data that corresponds one-to-one with that set of control parameters.

[0103] S314, the degree of deviation between the current frequency difference and the average historical frequency difference is taken as a frequency coordination term, and the absolute value of the concentration difference is weighted and summed with the frequency coordination term to construct a fitness function.

[0104] The degree of deviation refers to the magnitude of the deviation between the current frequency difference and the historical average frequency difference.

[0105] Specifically, for each calculated current frequency difference, it can be compared with the historical average frequency difference of the corresponding electric field to calculate the frequency deviation coefficient of each electric field. Specifically, the ratio of the difference between the current frequency difference and the historical average frequency difference to the historical average frequency difference can be calculated to obtain the frequency deviation coefficient of each electric field. The arithmetic mean of the deviation coefficients of each electric field can be calculated to obtain the final frequency coordination term. The larger this value is, the worse the frequency coordination state under the current parameter combination, the higher the risk of sparking, and the greater the corresponding penalty. Then, the absolute value of the difference between the predicted dust emission concentration and the preset target emission concentration is extracted as the emission deviation term. According to the core requirements of on-site operation, corresponding weight coefficients are set for the emission deviation term and the frequency coordination term. The two are weighted and summed to construct a complete fitness function. The evaluation rules of the fitness function can be clearly defined: the smaller the function value, the better the comprehensive performance of the corresponding control parameter combination. It can ensure that the dust emission concentration is close to the preset target value, meet environmental compliance requirements, and constrain the frequency dispersion through the frequency coordination term to avoid sparking, ensure stable equipment operation, and achieve optimal system energy consumption.

[0106] S32, initialize the particle swarm within the dynamic optimization boundary, and perform iterative calculations on the particles based on the fitness function to update the individual extreme value and population extreme value of each particle.

[0107] Among them, initializing the particle swarm refers to randomly generating a preset number of particle sets within the dynamic optimization boundary. Each particle corresponds to a set of control parameter combinations. Individual extreme value refers to the optimal control parameter combination found by a single particle during the iteration process, and population extreme value refers to the globally optimal control parameter combination found by the entire particle swarm during the iteration process.

[0108] See Figure 4 This is a flowchart illustrating a particle swarm optimization algorithm provided in an embodiment of the present invention, as shown below. Figure 4The diagram illustrates the complete operational logic of the adaptive particle swarm optimization algorithm for high-dimensional parameter optimization of electrostatic precipitators (ESPs) according to this invention: After the process starts, multiple particle swarms corresponding to multi-dimensional control parameters are initialized. The fitness value of each particle is calculated using the dual-objective fitness function: fitness = |predicted emissions - target emissions| + λ·penalty term, where λ is the dual-objective optimization weight balance coefficient, a preset adjustable constant used to balance the weight ratio between the ultra-low emission compliance target and the safe and energy-saving operation target of the ESP system. The individual optimal solution pBest and the global optimal solution gBest of the particle are updated synchronously. Then, based on the current parameter value and combined with the physical limits of the equipment, the optimization boundary is dynamically contracted. An inertial weight reduction strategy is used to update the particle's velocity and position. A premature convergence judgment mechanism is used to avoid the risk of local convergence. Finally, it is determined whether the dual convergence conditions of error being less than the preset error threshold or iteration equaling the maximum number of iterations are met. If not, the fitness value is recalculated to enter the next iteration; if met, the global optimal control quantity is output, completing the entire optimization process.

[0109] Specifically, the generated dynamic optimization boundary can be used as a constraint to randomly generate a particle swarm of a preset size. Each particle corresponds to a set of initial control parameter combinations, ensuring that the initial parameters are all within a safe range. Then, iterative calculation is started, and the dust removal effect and energy saving effect of the control parameter combinations corresponding to each particle are evaluated according to the fitness function. Furthermore, based on the results of the iterative calculation, the individual extreme values ​​and population extreme values ​​of each particle are updated, that is, the control parameter combinations corresponding to each particle and the global control parameter combinations are updated, until the optimal control parameter combination is determined when the convergence condition is met.

[0110] In some embodiments, step S32 may be implemented as follows:

[0111] S321, a particle swarm of a preset population size is randomly generated within the dynamic optimization boundary, wherein the position vector of each particle in the particle swarm corresponds to a set of control parameters.

[0112] Among them, the preset population size refers to the total number of particles in the particle swarm, the position vector refers to the coordinate representation of the particle in the control parameter space, each dimension of the vector corresponds to an electrostatic precipitator control parameter, and the entire vector corresponds to a complete set of control parameters.

[0113] Understandably, the optimization performance and convergence speed of the particle swarm optimization algorithm directly depend on the rationality and safety of the initial particle distribution. If the initial particles are concentrated in a local region of the parameter space, the algorithm will lose its global search capability, easily getting trapped in local optima and failing to find the globally optimal parameter combination. If the position vectors of the initial particles exceed the dynamic optimization boundary, the initial parameters will exceed the safe operating range of the equipment, posing a safety risk of flashover and tripping of high-voltage equipment. At the same time, the particle swarm size also affects the algorithm's performance. Too small a swarm size will result in an incomplete search of the parameter space, possibly failing to cover all potential optimal solutions, while too large a swarm size may lead to a large computational load. This approach cannot meet the real-time requirements of second-level closed-loop control in electrostatic precipitator systems. Therefore, by combining the dimensional characteristics of the control parameter combination with the real-time requirements of the electrostatic precipitator system, a preset population size can be determined through on-site debugging, for example, 50. Then, the generated dynamic optimization boundary can be retrieved to obtain the safe upper and lower limits of each parameter in the control parameter combination, including the rated operating range of secondary voltage, secondary current, and high-frequency frequency. Within the safe upper and lower limits of each parameter, values ​​conforming to a uniform distribution of the preset population size are randomly generated to form a corresponding number of multidimensional position vectors. Each position vector corresponds to a complete set of initial control parameters for the electrostatic precipitator, constituting the initial particle swarm.

[0114] S322, For each iteration, substitute the particle's position vector into the fitness function and calculate the particle's fitness value.

[0115] Among them, the iteration round refers to the complete process of the particle swarm algorithm completing one particle fitness calculation, updating the individual extreme value and the population extreme value, and adjusting the particle position. The fitness value is the value calculated by the constructed fitness function, which is used to determine the overall quality of the control parameter combination corresponding to the particle. The smaller the value, the better the overall performance of the parameter combination.

[0116] Specifically, after each iteration starts following the initialization of the particle swarm, all particles in the swarm are traversed first. The current position vector of each particle is extracted sequentially, which is the combination of electrostatic precipitator control parameters corresponding to that particle. The position vector is then decomposed into the high-frequency power supply frequency setpoint and emission-related parameters according to the input requirements of the constructed fitness function. These parameters are used to calculate the frequency coordination term and emission deviation term, respectively. The decomposed parameters are substituted into the fitness function. After the calculation is completed, the fitness value of the particle in the current iteration can be obtained. The above calculation process is repeated to complete the calculation of the fitness values ​​of all particles in the particle swarm, and the comprehensive performance evaluation result of all particles in the current iteration is obtained, which is the fitness value of each particle.

[0117] S323, if the fitness value is better than the historical best value of the particle, then update the current position vector of the particle to the individual extreme value of the particle.

[0118] Among them, the historical best value refers to the minimum fitness value of a single particle calculated by the fitness function in all previous iterations, which corresponds to the optimal combination of control parameters found by the particle in the current iteration. The current position vector refers to the real-time coordinates of the particle in the control parameter space in the current iteration, which corresponds to a set of the latest electrostatic precipitator control parameter combinations.

[0119] Specifically, during the initialization phase, an independent individual extreme value storage unit can be established for each particle. The initial position vector and the corresponding initial fitness value of the particle are stored in this storage unit as the initial individual extreme value and historical best value of the particle. After the fitness values ​​of all particles are calculated in each iteration, for each particle, the historical best value in its individual extreme value storage unit is retrieved and compared with the current fitness value calculated in the current iteration. If the current fitness value is less than the historical best value, it means that the control parameter combination found by the particle in this iteration has better overall performance than the results of all its previous iterations. In this case, the historical best value in the storage unit is updated to the current fitness value, and the individual extreme value is updated to the current position vector of the particle, thus completing the update of the individual extreme value. If the current fitness value is greater than or equal to the historical best value, it means that the parameter combination in this iteration is not better than the historical best result. In this case, the historical data in the individual extreme value storage unit remains unchanged and is not updated. The above process is repeated to complete the update of the individual extreme values ​​of all particles in the particle swarm, ensuring that each particle continues to retain the optimal control parameter combination found in its iterations so far.

[0120] S324 determines the optimal value of the population based on the comparison results between the fitness values ​​of each particle.

[0121] The optimal value of the population refers to the minimum fitness value corresponding to all combinations of control parameters found by the entire particle swarm in all iterations to date.

[0122] The core advantage of the particle swarm optimization algorithm lies in collective intelligence and information sharing. If there is a lack of unified guidance for the optimal value of the population, and each particle only searches for its own individual extreme value, it may cause the particles to be scattered in different regions of the parameter space, unable to share the information of the high-quality solutions found, resulting in extremely low optimization efficiency and inability to quickly converge to the global optimal solution. Therefore, the optimal value of the population can be determined based on the comparison results between the fitness values ​​of each particle.

[0123] Based on the above embodiments, step S324 can be implemented in the following ways:

[0124] S3241, during the initialization phase, calculate the initial fitness value of all particles, and take the smallest fitness value as the initial population optimum.

[0125] The initial fitness value refers to the fitness value calculated by substituting the initial position vector of each particle into the fitness function after the particle swarm is initialized.

[0126] Specifically, after the particle swarm initialization is completed and before the first iteration calculation is started, all particles in the particle swarm can be traversed first, and the initial position vector of each particle can be extracted in turn. The vector is then substituted into the constructed fitness function to calculate the initial fitness value of each particle. The calculated initial fitness values ​​are compared one by one, and the smallest initial fitness value is selected by sorting by size. This smallest initial fitness value is then determined as the initial optimal value of the population.

[0127] S3242, In each iteration round, after each particle completes its individual extreme value update, the fitness values ​​corresponding to the individual extreme values ​​of all particles are compared, and the minimum fitness value is determined as the optimal fitness value for the current iteration round.

[0128] The optimal fitness value refers to the minimum fitness value among all the individual extreme values ​​of all particles in the current iteration, representing the optimal evaluation result found by the entire particle swarm in the current iteration.

[0129] Since the particle position vector is updated in each iteration, and the corresponding individual extreme values ​​are also updated synchronously, directly selecting the optimal value of the current iteration based on the individual extreme values ​​of the previous iteration will lead to a lag in the selection results. This may fail to reflect the true optimization results of the current iteration, resulting in untimely updates of the population's optimal value and deviations in the algorithm's optimization direction. Therefore, after all the individual extreme values ​​of all particles are updated in each iteration, the individual extreme value storage units of all particles in the particle swarm can be traversed, and the fitness value corresponding to the latest updated individual extreme value of each particle can be extracted one by one. The extracted latest fitness values ​​are then compared and sorted, and the fitness value with the smallest value is selected. This smallest fitness value is determined as the optimal fitness value of the current iteration, and the particle position vector corresponding to this fitness value is recorded synchronously, which is the optimal combination of control parameters found in the current iteration.

[0130] S3243, if the optimal fitness value of the current iteration is better than the optimal value of the population, then update the optimal value of the population to the optimal fitness value of the current iteration.

[0131] After selecting the optimal fitness value for the current iteration, the population optimum can be retrieved. The optimal fitness value for the current iteration is compared with the population optimum to determine whether a better combination of control parameters has been found in the current iteration. If the optimal fitness value for the current iteration is less than the population optimum, it means that the combination of control parameters found in the current iteration has better overall performance than all results so far in the algorithm iterations. In this case, the population optimum is replaced with the optimal fitness value for the current iteration. If the optimal fitness value for the current iteration is greater than or equal to the population optimum, it means that no better global optimum solution than the historical ones has been found in the current iteration. In this case, the population optimum is kept unchanged, and no update operation is performed.

[0132] S325, if the fitness value is better than the population optimum, then update the particle's current position vector to the population optimum.

[0133] After completing the update determination of the population optimum, the particle position vector corresponding to the current population optimum can be confirmed. If it is determined that the population optimum needs to be updated, the population extreme value is updated to the position vector corresponding to the current optimum value to ensure that the two correspond one-to-one. If it is not necessary to update the population optimum, the historical population extreme value is kept unchanged, and finally the population extreme value is always the current global optimum combination of control parameters.

[0134] S326. Based on the individual extreme value, the population extreme value and the current position vector, update the optimization speed of each particle. According to the updated optimization speed and the current position vector, update the position vector of each particle, and the updated position vector is constrained within the dynamic optimization boundary.

[0135] Among them, the optimization speed refers to the direction of movement and adjustment range of the particle in the control parameter space.

[0136] The core of the particle swarm optimization algorithm is to achieve continuous search and optimization of the parameter space through the dynamic adjustment of particle positions. Without a scientific optimization speed and position update rule, the particles will be in a state of random motion without any rules, which may not be able to effectively approach the global optimum. They may either get stuck in local optima or fail to converge. At the same time, the speed update rule needs to balance global exploration and local development. In the early stage, particles need to be encouraged to explore the parameter space on a large scale to avoid getting stuck in local optima. In the later stage, particles need to be guided to finely search for optimization in high-quality regions to improve convergence accuracy.

[0137] Specifically, an optimization speed update formula can be set, as follows:

[0138]

[0139] in: The adaptive inertia weights decrease linearly, with significant weights in the early stages to encourage global particle exploration, and smaller weights in the later stages to guide particles to refine local optimization. The optimization speed of the particle in the previous iteration, , As a learning factor, The focus is on guiding particles toward their historical best. The focus is on guiding particles toward the global optimum of the population. , The value is a random number between 0 and 1 to increase the randomness of the search. pBest is the individual extreme value of the particle, and gBest is the population extreme value. This is the position vector of the particle in the previous iteration.

[0140] Based on this formula, the optimization speed of each particle in the current iteration is calculated, the adjustment direction and magnitude of the particle are determined, and then the updated current position vector of the particle can be calculated. Finally, the updated position vector is dynamically optimized and its boundary is checked. If the parameter value of a certain dimension exceeds the upper or lower limit of the dynamic optimization boundary, the parameter value of that dimension is adjusted to the corresponding upper or lower limit of the boundary to ensure that the updated particle position vector is always within a safe range. The above process is repeated to complete the optimization speed and position vector update of all particles, and to push the particle swarm toward the global optimum.

[0141] S33, when the preset iteration termination condition is met, output the parameter combination corresponding to the population extreme value to obtain the optimal control parameter set.

[0142] Specifically, while infinite iteration of the particle swarm optimization algorithm may improve optimization accuracy, it leads to excessive computation time, failing to meet the real-time requirements of second-level closed-loop control and being unable to adapt to rapidly fluctuating operating conditions. Therefore, iterative termination conditions can be preset, including a maximum number of iterations and a fitness threshold. The maximum number of iterations is set to meet the second-level closed-loop requirements, while the fitness threshold corresponds to the comprehensive requirements of emission compliance and stable operation. During the iteration process, real-time monitoring is conducted. If the maximum number of iterations is reached, or the fitness value corresponding to the population extreme value meets the preset threshold, the iteration is terminated, and the combination of control parameters corresponding to the current population extreme value is extracted and defined as the optimal control parameter set.

[0143] In some embodiments, step S33 may be implemented in the following ways:

[0144] S331, if the number of iterations reaches the maximum number of iterations, then the iteration is terminated, and the set of control parameters corresponding to the current population extreme value is output as the optimal set of control parameters.

[0145] Among them, the number of iterations refers to the number of complete iteration rounds that the particle swarm algorithm has completed from the start to the current moment, and the maximum number of iterations refers to the preset upper limit of the number of iteration rounds at which the algorithm terminates.

[0146] Specifically, at the start of the iteration process, the iteration counter is initialized with an initial value of 0. After each complete iteration, including fitness calculation, extreme value update, and particle position adjustment, the iteration counter is incremented by 1. After each iteration, the current value of the iteration counter is read and it is determined whether it is equal to the preset maximum number of iterations. If the iteration counter value is equal to the maximum number of iterations, the algorithm iteration is terminated immediately, and no further iteration calculations are performed. The control parameter combination corresponding to the current population extreme value is extracted as the optimal control parameter set and output to the subsequent safety verification and distribution stage. If the iteration counter value is less than the maximum number of iterations, the subsequent iteration judgment process continues to ensure that the optimization is completed within the specified number of rounds.

[0147] S332, if the number of iterations does not reach the maximum number of iterations and the optimal value of the population is less than the preset fitness threshold, the iteration is terminated, and the control parameter set corresponding to the current population extreme value is output as the optimal control parameter set.

[0148] Among them, the fitness threshold refers to the preset critical fitness value that determines whether the optimization accuracy meets the standard, and it is the condition for early termination of the iteration.

[0149] After each iteration completes the update of the population optimum, the current iteration counter value is read first to determine if it is less than the preset maximum number of iterations, ensuring that no termination condition is triggered. Then, the current population optimum value is read to determine if it is less than the preset fitness threshold. If the number of iterations has not reached the maximum number of iterations, but the population optimum value is less than the fitness threshold, it is also determined that a globally optimal solution that meets the requirements has been found, and the algorithm iteration is terminated immediately without executing the remaining iteration rounds. After the iteration terminates, the combination of control parameters corresponding to the current population extreme value is extracted and output as the optimal control parameter set to the subsequent safety verification and distribution stage. This can minimize the optimization time while ensuring the optimization accuracy.

[0150] S333, if the optimal value of the population is greater than or equal to the fitness threshold, then continue to the next iteration.

[0151] Specifically, if the current iteration count is less than the preset maximum iteration count, and the current optimal value of the population is greater than or equal to the preset fitness threshold, meaning there are still remaining iteration rounds and computation time, and the currently found parameter combination does not meet the accuracy requirements, then it is determined that the current optimization accuracy has not met the standard, and there is still room and time to continue optimization, so the next iteration is executed.

[0152] Through the above implementation method, global optimization of multi-dimensional control parameters can be completed within seconds, thereby avoiding the problem of low dust removal efficiency caused by spark competition between multiple electric fields and achieving coordinated optimization of all electric field operating parameters.

[0153] S4. The optimal control parameter set is verified by the safety verification unit. After the verification is passed, the parameter set is sent to the high-voltage control cabinet for execution.

[0154] Among them, the safety verification unit refers to the functional unit that has built-in multi-level interlock verification rules and realizes full-dimensional safety verification of control parameters, and the high-voltage control cabinet refers to the PLC / microcontroller control cabinet installed on site for controlling the operation of the high-voltage power supply of each electric field of the electrostatic precipitator.

[0155] Existing electrostatic precipitator optimization schemes generally suffer from a disconnect between algorithms and equipment safety constraints, easily outputting dangerous parameters that exceed limits or undergo sudden changes, leading to flashover and tripping of high-voltage equipment. Furthermore, most schemes can only output recommended values, requiring manual input, resulting in significant control lag and human error, and failing to achieve real-time closed-loop control. Therefore, a safety verification unit can perform triple safety verification on the optimal control parameter set obtained through optimization. If all three verifications pass, the optimal control parameter set can be written into the PLC / microcontroller of the electrostatic precipitator high-voltage control cabinet via a redundant RS-485 bus with opto-isolation of 2.5kV or higher, using the Modbus-RTU protocol for execution, thus completing the control closed loop.

[0156] Based on the above embodiments, step S4 can be implemented in the following ways:

[0157] S41, perform limit check on each control parameter in the optimal control parameter set. If any control parameter exceeds the dynamic optimization boundary, it is marked as a limit error and the issuance is terminated.

[0158] Among them, the limit violation check refers to the first safety check step of verifying the compliance of each parameter in the optimal control parameter set with upper and lower limits, which is a rigid constraint of the generated dynamic optimization boundary. The limit violation error refers to the abnormal state in which the value of any control parameter in the optimal control parameter set exceeds the upper or lower limit of the corresponding dimension of the dynamic optimization boundary. This means that the parameter has exceeded the safe operating range of the equipment and there is a risk of equipment damage after it is issued.

[0159] Specifically, the system can retrieve the latest updated dynamic optimization boundary that matches the current operating conditions, extract the upper and lower safety limits corresponding to each dimension parameter in the optimal control parameter set, and then receive the output optimal control parameter set. Each control parameter is paired with the upper and lower limits of the dynamic optimization boundary according to its dimension. For each paired parameter set, the system checks whether the current parameter value is between the corresponding dimension's upper and lower safety limits. If the parameter value is greater than or equal to the lower safety limit and less than or equal to the upper safety limit, the parameter passes the limit violation check. If the parameter value is less than the lower safety limit or greater than the upper safety limit, the parameter fails the check and is marked as a limit violation error. The dimension, value, and corresponding boundary upper and lower limits of the abnormal parameter are recorded synchronously. After completing the dimension-by-dimensional verification of all control parameters, if all parameters pass the verification, the limit violation check is considered passed, and the system proceeds to the subsequent mutation verification stage. If any parameter is marked as a limit violation error, the entire safety verification process is immediately terminated, triggering a limit violation anomaly and terminating the parameter distribution process.

[0160] S42, perform mutation verification on the optimal control parameter set, compare the optimal control parameter set with the control parameter set issued in the previous cycle, if the adjustment range of any control parameter exceeds the range threshold, mark it as a mutation error and terminate the issuance.

[0161] Among them, mutation verification refers to the stability verification of the single adjustment range of the optimal control parameter set. It is the second safety verification link to avoid fluctuations in the operating status of the equipment caused by large parameter jumps. The adjustment range refers to the relative change ratio between a single dimension parameter in the current optimal control parameter set and the same dimension parameter successfully issued in the previous cycle. The amplitude threshold is a pre-set critical judgment value used to limit the maximum relative adjustment range of the single adjustment of the electrostatic precipitator high voltage power supply control parameter. Mutation error refers to the abnormal state in which the adjustment range of any parameter in the optimal control parameter set exceeds the preset amplitude threshold, which means that the single adjustment range of the parameter is too large and may cause drastic fluctuations in the electric field operation status after issuance.

[0162] Specifically, the set of valid control parameters successfully issued in the previous cycle and running stably on-site can be retrieved from the parameter latch unit as the benchmark for stability verification. Then, the current optimal control parameter set and the valid parameter set of the previous cycle are paired one by one according to the dimensions of the control parameters. For each pair of parameters, the adjustment range corresponding to each parameter can be calculated based on the difference between the current parameter value and the parameter value of the previous cycle and the ratio of the quality of the parameter value of the previous cycle. The preset amplitude threshold is retrieved, and the calculated adjustment range of each parameter is compared with the amplitude threshold. If the adjustment range is less than or equal to the amplitude threshold, the parameter is determined to have passed the mutation verification. If the adjustment range is greater than the amplitude threshold, the parameter is immediately determined to have failed the verification and is marked as a mutation error. After completing the dimension-by-dimensional verification of all control parameter combinations, if all parameters pass the verification, the mutation verification is determined to have passed, and the subsequent gradient verification stage is entered. If any parameter is marked as a mutation error, the entire safety verification process is immediately terminated, triggering a mutation anomaly.

[0163] S43, perform gradient verification on the optimal control parameter set. If the secondary voltage difference between adjacent electric fields exceeds the voltage difference threshold, it is marked as a gradient error and the transmission is terminated.

[0164] Gradient verification refers to the dust removal efficiency adaptability verification of the difference between the set values ​​of the secondary voltages of adjacent electric fields arranged along the flue gas flow direction. The secondary voltage difference refers to the absolute difference between the set values ​​of the secondary voltages of two adjacent electric fields along the flue gas flow direction. The voltage difference threshold refers to the preset maximum allowable absolute difference between the secondary voltages of adjacent electric fields. Gradient error refers to the abnormal state in which the absolute difference between the secondary voltages of adjacent electric fields exceeds the preset voltage difference threshold, which indicates that the dust removal gradient of the electric field is unreasonable and there is a risk of secondary dust generation and reduced dust removal efficiency.

[0165] Specifically, the secondary voltage setpoints corresponding to each electric field can be extracted from the optimal control parameter set. For example, according to the flow direction of flue gas in the electrostatic precipitator, if there are 8 electric fields, they can be sorted in the order of electric field 1 (the first electric field at the inlet) to electric field 8 (the last stage electric field at the outlet), forming an ordered electric field voltage sequence. Then, based on the sorting result, the absolute difference between the secondary voltage setpoints of two adjacent electric fields can be calculated group by group, i.e., electric field 1 and electric field 2, electric field 2 and electric field 3... electric field 7 and electric field 8, for a total of 7 groups of secondary voltage differences between adjacent electric fields. The preset voltage difference threshold is then retrieved. This threshold is adjusted based on the dust removal mechanism of the horizontal electrostatic precipitator and the on-site operating conditions. The optimal value, determined through trial and error, balances dust removal efficiency with the requirement to prevent secondary dust emission. The secondary voltage difference of each group of adjacent electric fields is compared to a voltage difference threshold. If the absolute difference is less than or equal to the voltage difference threshold, the group of adjacent electric fields passes the gradient verification. If the absolute difference is greater than the voltage difference threshold, the verification fails and is marked as a gradient error. Simultaneously, the adjacent electric field number, voltage difference, and threshold of the abnormality are recorded. After verifying all seven groups of adjacent electric fields, if all groups pass the verification, the gradient verification is considered successful, completing the entire triple safety verification process. If any group is marked as a gradient error, the entire safety verification process is immediately terminated, triggering a safety fallback strategy for gradient anomalies.

[0166] S44. If the over-limit check, mutation check and gradient check all pass, the optimal control parameter set is sent to the high-voltage control cabinet for execution.

[0167] Specifically, after the triple checks of over-limit verification, mutation verification, and gradient verification all pass, the optimal control parameter set is first distributed to the high-voltage control cabinets corresponding to each electric field according to the address allocation rules of the high-voltage control cabinets. A corresponding Modbus-RTU protocol register address is matched for each parameter. Then, the redundant RS-485 communication link of the verification distribution module is activated. This link has two independent physical buses with opto-isolation of over 2.5kV, which can withstand strong electromagnetic interference in the field. In the event of a single bus failure, it can automatically switch to the other bus in a short time to ensure uninterrupted communication. Industrial communication is adopted. Using the Modbus-RTU protocol, the split parameters are written bit by bit into the holding registers of the corresponding high-voltage control cabinet's PLC / microcontroller according to the register address. After the parameters are written, a register readback operation is immediately performed. The readback parameter values ​​are compared bit by bit with the issued optimal control parameter set to confirm that all parameters are written accurately and without transmission errors. After the parameters are confirmed to be correct, a status message indicating that the parameter issuance was successful is sent to the Web display module. At the same time, the parameter set issued this time can be stored in the parameter latch unit as the benchmark for the next cycle of mutation verification, thus completing the closed-loop issuance and execution of the entire optimization parameters.

[0168] S45 If any verification fails, the control parameter set issued in the previous cycle remains unchanged, and a safety alarm is triggered.

[0169] If a parameter verification fails and the corresponding error type is marked at any stage of the over-limit verification, mutation verification, or gradient verification, the safety fallback procedure of this step is immediately triggered: First, immediately freeze the entire process of issuing the current optimal control parameter set, prohibiting any abnormal parameters from being written to the communication link and the high-voltage control cabinet. Then, retrieve the valid control parameter set that was successfully issued and is running stably on-site in the previous cycle, stored in the parameter latch unit, and keep issuing this parameter set to the high-voltage control cabinet. Latch the current operating status of the electrostatic precipitator to ensure that there are no jumps or loss of control in the equipment's operating status. At the same time, according to the marked error type, including over-limit errors... Errors, mutation errors, and gradient errors trigger corresponding alarms to the web-based display module via the WebSocket protocol. The alarm information clearly indicates the anomaly type, the number / value of the anomaly parameter, the verification threshold, and other key information, enabling operators to quickly locate and troubleshoot the problem. Simultaneously, all information about this anomaly event, including the timestamp, the set of anomaly parameters, the verification results, and the current operating data, is recorded to facilitate subsequent anomaly tracing and analysis. Until the anomaly is resolved, the system maintains stable operation of the valid parameters from the previous cycle until the operators resolve the anomaly or receive a new set of optimal control parameters that has passed all verifications.

[0170] Through the above implementation methods, the deep integration and real-time closed-loop operation of optimization control and on-site DCS system can be achieved, thereby improving dust removal efficiency.

[0171] Based on the above steps, this solution also includes the following embodiments:

[0172] A1, when the security verification unit receives the cut-out command, it stops issuing the optimal control parameter set and switches to the traditional spark tracking mode to continue running until it receives the cut-in command and resumes issuing the optimal control parameter set.

[0173] The safety verification unit can monitor command signals from the web-based display module and the local control panel in real time, while simultaneously monitoring the device's own hardware and communication operating status. Upon receiving a switch-out command from the operator, or detecting serious anomalies such as hardware failure or communication interruption, it immediately executes the mode switching process: stopping the iterative optimization and distribution of the optimal control parameter set, locking the current device operating parameters, and using these parameters as a transition benchmark, seamlessly switching to the power plant's original traditional spark tracking control mode in the next control cycle. There are no parameter jumps or operational shocks during the switching process. Simultaneously, the current operating mode can be updated in real time on the web-based display module and locked. The parameter distribution channel is optimized, retaining only data acquisition and status monitoring functions. During operation in traditional spark tracking mode, the device continuously maintains data acquisition and status monitoring functions, while simultaneously listening for switchover commands in real time. When a switchover command is received from the operator, the device first completes the latest emission prediction and parameter optimization based on the current real-time operating data. The generated optimal control parameter set undergoes triple safety verification. After all verifications pass, the device smoothly transitions to the optimized mode in the next control cycle, using the current traditional mode operating parameters as a benchmark. The optimal control parameter set is gradually distributed, restoring the complete optimization control closed loop. The switching process is free from operating condition shocks and emission fluctuations, thus completing the mode switchover.

[0174] See Figure 5 This is a schematic diagram of the structure of a DCS electrostatic precipitator optimization control device provided in an embodiment of the present invention. The data processing system of the DCS electrostatic precipitator optimization control device includes:

[0175] The data acquisition module is used to acquire the operating data of the electrostatic precipitator based on a preset acquisition cycle, and to preprocess the operating data to obtain standard operating data;

[0176] The concentration prediction module is used to predict the dust emission concentration based on the standard operating data, and to determine the dynamic optimization boundary based on the concentration difference between the dust emission concentration and the target emission concentration.

[0177] The parameter optimization module is used to perform global optimization within the dynamic optimization boundary based on the particle swarm optimization algorithm to obtain the optimal set of control parameters.

[0178] The verification and distribution module is used to perform a safety verification on the optimal control parameter set according to the safety verification unit, and after the verification is passed, it is distributed to the high-voltage control cabinet for execution.

[0179] Figure 5 The apparatus of the illustrated embodiment can be used to perform corresponding actions. Figure 1 The steps in the method embodiments shown are implemented in a similar manner and have similar technical effects, and will not be repeated here.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimal control of a DCS electrostatic precipitator system, characterized in that, include: The operating data of the electrostatic precipitator is acquired based on a preset acquisition cycle, and the operating data is preprocessed to obtain standard operating data. The dust emission concentration is predicted based on the standard operating data, and the dynamic optimization boundary is determined based on the concentration difference between the dust emission concentration and the target emission concentration. The optimal set of control parameters is obtained by performing global optimization within the dynamic optimization boundary based on the particle swarm optimization algorithm. The optimal control parameter set is verified by the safety verification unit. Once the verification is successful, the parameter set is sent to the high-voltage control cabinet for execution.

2. The method according to claim 1, characterized in that, The step of predicting dust emission concentration based on the standard operating data and determining the dynamic optimization boundary based on the concentration difference between the dust emission concentration and the target emission concentration includes: Using the control parameter values ​​corresponding to the current standard operating data as a reference point, the direction of parameter adjustment is determined according to the sign of the concentration difference, and the search range width is determined according to the absolute value of the concentration difference. Centered on the reference point, the search range is expanded within the width of the search range along the parameter adjustment direction and intersected with the physical limit parameters of the electrostatic precipitator to generate the dynamic optimization boundary. The dynamic optimization boundary is recalculated and updated every preset collection cycle based on the concentration difference at the current moment.

3. The method according to claim 2, characterized in that, The particle swarm optimization algorithm performs global optimization within the dynamic optimization boundary to obtain the optimal set of control parameters, including: The frequency coordination term is determined based on the degree of dispersion between the operating frequencies of different power supply units within the same electric field, and a fitness function is constructed based on the absolute value of the frequency coordination term and the concentration difference. The particle swarm is initialized within the dynamic optimization boundary, and the particles are iteratively calculated based on the fitness function to update the individual extreme value and the population extreme value of each particle. When the preset iteration termination condition is met, the parameter combination corresponding to the population extreme value is output to obtain the optimal control parameter set.

4. The method according to claim 3, characterized in that, The process of determining a frequency coordination term based on the degree of dispersion between the operating frequencies of different power supply units within the same electric field, and constructing a fitness function based on the absolute value of the frequency coordination term and the concentration difference, includes: Based on the secondary voltage setpoint, secondary current setpoint, and high frequency power supply frequency setpoint of each electric field high voltage power supply, a set of control parameters to be optimized is obtained; The operating frequencies of different power supply units within the same electric field during historical operating periods are obtained, and the average historical frequency difference between each power supply unit is calculated based on the operating frequencies. Calculate the current frequency difference based on the frequency set value corresponding to different power supply units in the same electric field from any set of control parameters. The degree of deviation between the current frequency difference and the average historical frequency difference is used as a frequency coordination term, and the absolute value of the concentration difference is weighted and summed with the frequency coordination term to construct a fitness function.

5. The method according to claim 4, characterized in that, The process of initializing the particle swarm within the dynamic optimization boundary and iteratively calculating the particles based on the fitness function to update the individual extreme value and the population extreme value of each particle includes: Within the dynamic optimization boundary, a particle swarm of a preset population size is randomly generated, wherein the position vector of each particle in the particle swarm corresponds to a set of control parameters. For each iteration, the particle's position vector is substituted into the fitness function, and the particle's fitness value is calculated. If the fitness value is better than the particle's historical best value, then the particle's current position vector is updated to the particle's individual extreme value; The optimal value of the population is determined based on the comparison results between the fitness values ​​of each particle. If the fitness value is better than the population optimum, then the particle's current position vector is updated to the population optimum. Based on the individual extreme value, the population extreme value, and the current position vector, the optimization speed of each particle is updated. According to the updated optimization speed and the current position vector, the position vector of each particle is updated, and the updated position vector is constrained within the dynamic optimization boundary.

6. The method according to claim 5, characterized in that, The determination of the optimal population value based on the comparison results between the fitness values ​​of each particle includes: During the initialization phase, the initial fitness value of all particles is calculated, and the smallest fitness value is taken as the initial population optimum. In each iteration round, after each particle completes its individual extreme value update, the fitness values ​​corresponding to the individual extreme values ​​of all particles are compared, and the minimum fitness value is determined as the optimal fitness value for the current iteration round. If the optimal fitness value of the current iteration is better than the optimal value of the population, then the optimal value of the population is updated to the optimal fitness value of the current iteration.

7. The method according to claim 3, characterized in that, When the preset iteration termination condition is met, the parameter combination corresponding to the population extreme value is output to obtain the optimal control parameter set, including: If the number of iterations reaches the maximum number of iterations, the iteration is terminated, and the set of control parameters corresponding to the current population extreme value is output as the optimal set of control parameters. If the number of iterations does not reach the maximum number of iterations, or if the optimal value of the population is less than the preset fitness threshold, the iteration is terminated, and the set of control parameters corresponding to the current extreme value of the population is output as the optimal set of control parameters. If the optimal value of the population is greater than or equal to the fitness threshold, then continue to the next iteration.

8. The method according to claim 1, characterized in that, The step of performing a security verification on the optimal control parameter set based on the security verification unit, and then issuing the verification to the high-voltage control cabinet for execution after passing the verification, includes: Each control parameter in the optimal control parameter set is checked for exceeding the limit. If any control parameter exceeds the dynamic optimization boundary, it is marked as an exceeding the limit error, and the distribution is terminated. A mutation check is performed on the optimal control parameter set. The optimal control parameter set is compared with the control parameter set issued in the previous cycle. If the adjustment range of any control parameter exceeds the range threshold, it is marked as a mutation error and the issuance is terminated. Gradient verification is performed on the optimal control parameter set. If the secondary voltage difference between adjacent electric fields exceeds the voltage difference threshold, it is marked as a gradient error and the transmission is terminated. If the over-limit check, mutation check, and gradient check all pass, the optimal control parameter set will be sent to the high-voltage control cabinet for execution. If any check fails, the control parameter set issued in the previous cycle remains unchanged, and a safety alarm is triggered.

9. The method according to claim 8, characterized in that, Also includes: When the security verification unit receives a cut-out command, it stops issuing the optimal control parameter set and switches to traditional spark tracking mode to continue operating until it receives a cut-in command and resumes issuing the optimal control parameter set.

10. A DCS electrostatic precipitator system optimization control device, characterized in that, include: The data acquisition module is used to acquire the operating data of the electrostatic precipitator based on a preset acquisition cycle, and to preprocess the operating data to obtain standard operating data; The concentration prediction module is used to predict the dust emission concentration based on the standard operating data, and to determine the dynamic optimization boundary based on the concentration difference between the dust emission concentration and the target emission concentration. The parameter optimization module is used to perform global optimization within the dynamic optimization boundary based on the particle swarm optimization algorithm to obtain the optimal set of control parameters. The verification and distribution module is used to perform a safety verification on the optimal control parameter set according to the safety verification unit, and after the verification is passed, it is distributed to the high-voltage control cabinet for execution.

Images