An adaptive dust cleaning control method for an electric dust collector

Abstract

The application discloses a self-adaptive ash cleaning control method of an electric dust collector. In view of the technical problem that the existing ash cleaning control of the electric dust collector regards single vibration as an indivisible atomic event, is executed according to a preset fixed time length, and causes the secondary dust raising of the last electric field to be short-time over-standard, the single vibration process is subdivided into three states of a main peeling period, a platform period and a secondary dust raising precursor through a five-step control, the first-order difference mean and the second-order difference mean of the secondary current sequence of the electric field are taken as real-time identification quantities of each sampling point, and in the cross-vibration partition scheduling, the partition only identifying the secondary dust raising precursor is automatically reduced in gear vibration acceleration in the next round, the double-layer closed-loop control of the single vibration stop time and the cross-partition scheduling feedback is realized, and the short-time impact of the secondary dust raising of the last electric field on the outlet flue dust concentration is significantly reduced.

Description

Technical Field

[0001] This invention relates to the field of dust removal control technology for industrial dust-laden electrostatic precipitators, specifically to a control method and system for real-time identification of rapping sub-stages and adaptive rapping stop and cross-zone scheduling during the dust collection plate rapping process of a dry electrostatic precipitator. Background Technology

[0002] An electrostatic precipitator is a device that uses a high-voltage direct current electric field to adsorb and capture charged dust particles in flue gas. It is widely used in the treatment of dusty industrial flue gas in coal-fired power plants, sintering machine heads, cement kilns, and smelting furnaces, with dust removal efficiencies reaching over 99%. After being charged, dust particles are adsorbed onto the grounded collecting plates, forming a ash layer. This ash layer needs to be periodically removed to maintain the electric field efficiency. Typically, a mechanical rapping device is used, with a hammer or electromagnetic coil impacting the collecting plates to cause the ash layer to fall off in sheets and into an ash hopper at the bottom of the equipment, from where it is transported by the ash conveying system.

[0003] Regarding the mechanical structure of the rapping device itself, the industry has developed several typical solutions, including side-mounted flexible arm hammers, top cam-driven mechanisms, electromagnetic pulse systems, and acoustic cleaning. At the control strategy level of the rapping device, the mainstream engineering approaches are as follows: The first is a preset rapping cycle approach, which uses experimental experience to determine the rapping interval and duration for each rapping zone, and operates according to this preset duration during operation; the second is a threshold trigger approach, which, based on the first approach, adds threshold monitoring of the average secondary current of the electric field or the frequency of spark discharges, and triggers a rapping action immediately when the monitored value exceeds the preset threshold; the third is a historical data-driven intelligent prediction approach, which uses machine learning models to train long-term operating data to predict the optimal rapping cycle for each time period.

[0004] While the three approaches described above each have their advantages in engineering implementation, they all share a common problem that has long been observed in the industry but has never been engineered. Specifically, in electrostatic precipitators employing any of these approaches, the rapping process, once triggered, executes for a preset fixed rapping duration. During this duration, the rapping mechanism continuously outputs impact force until the preset duration ends. However, in actual operation, a single rapping process is not uniform and monotonous: in the initial stage of rapping, the accumulated dust layer on the surface of the dust collecting plates falls off in sheets under the impact force. During this stage, the equivalent resistance of the dust layer drops rapidly, causing a significant increase in the secondary current of the electric field in the rapping zone. After the main dust layer falls off, the plate surface enters a relatively clean steady state, and the change in secondary current tends to be gradual. If rapping continues at this point, the fine dust that has fallen into the dust hopper and the residual dust still attached to the lower edge of the plate will be re-raised by continuous mechanical disturbance, forming localized dispersed dust clumps within the electric field. This phenomenon is commonly known in the industry as secondary dust re-entrainment. Secondary dust not only causes reverse fluctuations in the secondary current of the electric field, but also flows downstream with the flue gas, increasing the dust removal burden on the downstream electric field. In severe cases, it can directly lead to a short-term exceedance of the outlet dust concentration.

[0005] Although the industry has long observed secondary dust pollution—for example, during the commissioning of the final electric field of the electrostatic precipitator at the sintering machine head, operators often hear a vibration in the secondary current at the end of the rapping phase, accompanied by a short-term increase in outlet concentration—the transient characteristics of this phenomenon in the electric field secondary current time series have never been engineered as a criterion for stopping rapping. The fundamental reason is that existing control strategies treat a single rapping event as an atomic event that cannot be further subdivided, and no further identification is performed within that event. Furthermore, current electric field secondary current sampling generally uses a step size of 1 second or even several seconds, and the sampling density is far insufficient to characterize the shape of the current change curve during the rapping transient process. Specifically, existing low-voltage control systems for electrostatic precipitators generally use programmable logic controllers (PLCs) as actuators. The sampling period is determined by the PLC's scan cycle, typically between 50 milliseconds and 1 second. Taking a 100-millisecond sampling period as an example, a single rapping action lasting 400 milliseconds generates only 4 sampling points. This sampling density, even using the most advanced statistical criteria, cannot identify the three states—main spalling, plateau, and dust precursors—within a single rapping process. Even though some high-end electrostatic precipitators are equipped with high-speed sampling modules with sampling periods below 50 milliseconds, the obtained dense sampling data is usually only used for post-event operational analysis and historical trend statistics, and is not fed back into the current rapping action control loop as a basis for stopping vibration. In other words, although the industry has possessed the hardware capability to generate dense sampling data for many years, it has not used this data as an engineering criterion for real-time vibration stopping during the rapping process.

[0006] In addition, a few studies in the industry have attempted to use machine learning methods to predict the rapping cycle based on historical data. A typical approach is to collect long-term data such as the average secondary current of the electric field, the concentration of dust at the outlet, and load fluctuations to train a neural network model, and then predict the optimal rapping cycle for each period. This approach has three drawbacks: First, it still predicts at the "rapping cycle" granularity, without real-time identification within the "single rapping process," and cannot solve the fundamental problem of secondary dust generation. Second, this approach relies heavily on the accumulation of historical data from the field, and performs poorly in newly commissioned units or situations with drastic short-term fluctuations in operating conditions. Third, this approach requires the integration of a neural network inference engine into the low-voltage control system, significantly raising the deployment threshold and operational computing power requirements, making it difficult to apply to small and medium-sized electrostatic precipitators.

[0007] Existing control methods only concern themselves with when to start rapping and how strong the rapping should be, without considering when to stop during the rapping process. At the same time, existing control methods use the electric field as the smallest control unit, treating multiple rapping zones along the airflow direction within the same electric field as synchronous actions, without making differentiated judgments at the zone granularity. Summary of the Invention

[0008] This invention provides an adaptive dust removal control method for an electrostatic precipitator, which operates according to the following five steps: Step S0 is the initialization step, in which the electrostatic precipitator is divided into several rapping zones along the airflow direction, numbered according to their physical locations and bound to their secondary current sampling sources, and the initial peeling period slope threshold, plateau period duration step threshold, and dust pre-shaking amplitude threshold are set.

[0009] Step S1, i.e., vibration triggering, sorts the vibration partitions according to a preset triggering criterion, executes a single vibration action starting from the first one, and simultaneously samples the secondary current sampling source to obtain a secondary current sequence at a value not greater than the preset upper limit of transient sampling step size.

[0010] Step S2 is the three-stage real-time identification, which determines the secondary current sequence by sampling point according to the following three states: the main peeling period state, the plateau period state, and the secondary dust precursor state. In the main peeling period state, the rapping action continues; in the plateau period state, the rapping action is put into preparation for stopping the rapping action; and in the secondary dust precursor state, the rapping action is stopped immediately.

[0011] Step S3 is the threshold self-calibration. After the current rapping is completed, the measured number of steps of the plateau period and the peak value of the dust precursor shaking are combined into the corresponding threshold by means of rolling average, so that the criterion for the next rapping is self-calibrated according to the on-site dust characteristics.

[0012] Step S4 is cross-zone scheduling feedback, which involves traversing all rapping zones in the order of steps S1 to S3; at the end of each round, for rapping zones that only detect the secondary dust precursor state but not the main peeling period state, the rapping acceleration of the next round is reduced by no less than a preset level to avoid over-vibration again.

[0013] The present invention also provides an adaptive dust removal control system for an electrostatic precipitator corresponding to the method, including an initialization module, a rapping trigger module, a three-stage real-time identification module, a threshold self-calibration module, and a cross-zone scheduling feedback module, each module corresponding to the five steps.

[0014] The beneficial effects of this invention include: First, by identifying three sub-stages within a single rapping process—the main peeling period, the plateau period, and the precursor to secondary dust—each rapping can be stopped at the optimal moment when the main ash layer has been peeled off but before secondary dust is triggered, significantly reducing the short-term impact of secondary dust in the final electric field on the outlet dust concentration.

[0015] Secondly, by using the threshold rolling average self-calibration of the vibration zone level, the method described in this invention can adaptively converge based on the differences in parameters such as dust resistivity and ash layer adhesion characteristics after commissioning without manual debugging. This makes it suitable for thermal power plants with large fluctuations in coal quality and sintering, cement and other applications with large fluctuations in incoming materials.

[0016] Third, through the "only secondary dust precursors are identified but main peeling is not identified" zone reduction mechanism in the cross-zone scheduling feedback, the over-vibration zone is automatically located and the rapping acceleration is automatically reduced, continuously eliminating local over-vibration during long-term operation, reducing energy consumption and electrode wear.

[0017] Fourth, the method described in this invention only increases the sampling density and superimposes two types of rolling quantities, namely the first-order differential mean and the second-order differential mean, on the existing electric field secondary current sampling hardware. The computational complexity is constant level, and it can be directly deployed in the low-voltage control system of existing electrostatic precipitators, with extremely low engineering modification costs. Attached Figure Description

[0018] Figure 1 is a schematic diagram of the overall process of the adaptive dust removal control method for electrostatic precipitators according to the present invention.

[0019] Figure 2 is a schematic diagram of the deployment scenario of the electrostatic precipitator and the rapping zone of the present invention arranged along the airflow direction.

[0020] Figure 3 is a schematic diagram comparing typical time-series curves of the transient secondary current sequence during a single rapping process as described in this invention under three states: the main peeling period, the plateau period, and the precursor to secondary dust.

[0021] Figure 4 is a schematic diagram of the sampling point-by-sampling decision tree for the three-stage real-time identification in step S2 of the present invention.

[0022] Figure 5 is a schematic diagram of the threshold axis classification of the three criteria for vibration triggering in step S1 of the present invention.

[0023] Figure 6 is a schematic diagram of the principle of threshold self-calibration of exponential weighted rolling average in step S3 of the present invention.

[0024] Figure 7 is a schematic diagram of the downgrade and upgrade mechanism of cross-partition scheduling feedback in step S4 of the present invention.

[0025] Figure 8 is a bar chart comparing the energy consumption and peeling rate of a single vibration with the method of the present invention and the existing preset fixed duration vibration method.

[0026] Figure 9 is a schematic diagram of the secondary dust suppression rate under two-dimensional conditions of different dust resistivity and different flue gas velocities using the method of the present invention.

[0027] Figure 10 is a schematic diagram comparing the time series of flue dust concentration at the outlet of an electrostatic precipitator in a coal-fired power plant before and after the operation of the method described in this invention.

[0028] Figure 11 is a schematic diagram of the five-module structure of the adaptive dust removal control system for the electrostatic precipitator described in this invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the following specific embodiments are presented in the following order: First, as shown in Figure 1, an overall flow overview of the method of this invention is given; second, as shown in Figure 2, the physical deployment of the electrostatic precipitator and rapping zone to which the method of this invention is applicable is introduced; third, the connotation, extension, value, and examples of each core term involved in the method of this invention are explained in detail; furthermore, using the four-piece structure of input-action-output-adjustable parameter, a general end-to-end implementation overview of steps S0 to S4 of the method of this invention is provided, and the core mechanism of each step is illustrated in Figures 3, 4, 5, 6, and 7; then, detailed implementation examples and typical numerical values ​​are given for each step; furthermore, as shown in Figures 8, 9, and 10, Example 1 of the method of this invention being put into operation in an electrostatic precipitator of a coal-fired power plant and Example 2 of the method being put into operation in an electrostatic precipitator at the head of a sintering machine are given, with measured data provided for each item; finally, as shown in Figure 11, corresponding implementation examples of the modules of the system of this invention are given. The numerical values ​​described below are specific illustrative examples under the corresponding operating conditions and should not be construed as limiting the scope of protection claimed by this invention.

[0030] As shown in Figure 1, the overall flow of the method of the present invention includes a starting node, five working steps from S0 to S4, a loop determination node for whether all partitions in this round are completed, and an ending node. Step S2 contains three parallel sub-determination branches: the main peeling period, the plateau period, and the precursor to secondary dust. Step S3 is triggered on the other two branches besides the main peeling period branch. Step S4 is triggered only when all partitions in one round are completed. The execution relationship between the five working steps is as follows: step S0 is executed only once when the system starts; steps S1 to S3 form an execution loop for a single vibration action; and step S4 forms a feedback loop across rounds.

[0031] As shown in Figure 2, the electrostatic precipitator applicable to the method of the present invention is a typical dual-chamber, four-field horizontal dry electrostatic precipitator, with electric fields No. 1, No. 2, No. 3, and No. 4 arranged sequentially along the airflow direction. Flue gas flows in from the left inlet, is charged and captured by the electric fields, and flows out from the right outlet. Each electric field is further divided into four rapping zones along the airflow direction, resulting in a total of 16 rapping zones, numbered 1A, 1B, 1C, 1D, 2A to 4D in a double-layer numbering manner. The secondary current of the high-voltage rectifier transformer in each electric field is the secondary current described in the present invention, which is sent to the threshold self-calibration module and the three-stage real-time identification module via the secondary current sampling interface.

[0032] The full explanation of the core terms is as follows.

[0033] The aforementioned rapping zone refers to a physical area in an electrostatic precipitator divided along the airflow direction, consisting of several adjacent dust collection plates, which can be synchronously rapped by the same group of rapping mechanisms. Each electric field typically contains 2 to 6 rapping zones, numbered sequentially along the airflow direction. In this embodiment, a dual-chamber four-field electrostatic precipitator for an 800 MW unit contains 4 rapping zones per electric field, for a total of 32 rapping zones in the entire unit.

[0034] The secondary current sampling source refers to a sensing device that provides a sampling signal for the secondary output current of the high-voltage rectifier transformer in the electric field of each rapping zone. It can be either the original secondary current measurement circuit of the electric field or a high-speed current sampling module added to the method described in this invention. The physical meaning of the secondary current is the total corona current in the electric field, with a typical value range of not less than 50 mA and not more than 1500 mA.

[0035] The rapping action refers to a continuous mechanical impact process applied by the rapping mechanism to the dust collection plate in the rapping zone; the process starts from the rapping start signal output by the rapping drive interface and ends when the rapping stop signal is output by the rapping drive interface; in the method of the present invention, the rapping action no longer ends with a preset fixed duration, but the end time is determined by the real-time determination result of step S2.

[0036] The secondary current sequence refers to the sequence of current sample values ​​arranged in chronological order obtained by densely sampling the secondary current sampling source with the preset upper limit of transient sampling step size or a smaller step size after the rapping action is started.

[0037] The first-order difference mean refers to the scalar obtained by taking the arithmetic mean of the differences between adjacent sampled values ​​of the secondary current sequence within the most recent preset observation length; its physical meaning is the average rate of change of the secondary current within the observation window. The calculation formula can be expressed as the arithmetic mean of the differences between two adjacent sampled points on a subsequence consisting of the most recent N sampled points.

[0038] The second-order difference mean refers to the scalar obtained by performing a first-order difference on the first-order difference sequence again and taking the arithmetic mean within the most recent preset observation length; its physical meaning is the average acceleration of the rate of change of the secondary current within the observation window, that is, the local curvature approximation of the secondary current time series curve.

[0039] The peeling period slope threshold is a criterion threshold used to determine the state of the main peeling period. When the first-order difference mean is greater than the threshold, it is determined that the current state is in the main peeling period. Its initial value is set by step S0 and self-calibrated by exponential weighted moving average in step S3. The typical initial value is taken as 0.01 to 0.03 times the electric field reference current value where the rapping zone is located.

[0040] The plateau duration threshold is a criterion used to determine the duration of the plateau state. When the mean of the first difference is not greater than the slope threshold of the peeling period and the absolute value of the mean of the second difference does not exceed the preset plateau fluctuation amplitude threshold for a period of time, it is determined that the current state is in a plateau state. Its initial value is usually no less than 6 sampling points, typically 8 to 16 sampling points.

[0041] The threshold value for the amplitude of the dust precursor is used as a criterion for determining the amplitude of the secondary dust precursor state. When the second-order difference mean changes from non-negative to negative and the accompanying oscillation amplitude exceeds the threshold value, it is determined that the current state is a secondary dust precursor state. Its initial value is set by step S0, typically 0.05 to 0.08 times the reference current value.

[0042] The amplitude of the oscillation in the same direction refers to the absolute value of the difference between the peak and valley values ​​of the secondary current sequence within the most recent preset observation length.

[0043] The main peeling period refers to the sub-stage in which the accumulated dust layer on the surface of the dust collection plate falls off in pieces under the impact force of the rapping mechanism during a single rapping action. In this sub-stage, the mean value of the first-order difference is greater than the peeling period slope threshold, and in step S2, it is determined to continue the rapping action. Its physical meaning is the production capacity stage of a single rapping action.

[0044] The plateau state refers to the sub-stage in which the main ash layer has been completely removed and the electrode surface has entered a relatively clean and stable state during a single vibration action. In this sub-stage, the mean of the first-order difference is not greater than the peeling period slope threshold and the absolute value of the mean of the second-order difference does not exceed the preset plateau period fluctuation amplitude threshold and continues to reach the plateau period duration threshold. In step S2, it is determined to enter the state of preparing to stop vibration. Its physical meaning is the window period during which vibration can be stopped in a single vibration action.

[0045] The aforementioned secondary dust emission precursor state refers to the sub-stage in which fine dust that has fallen into the ash hopper and residual dust still attached to the lower edge of the electrode plate are continuously stirred up by mechanical disturbance during a single rapping action, and are about to form secondary dust emission. In this sub-stage, the second-order difference mean changes from non-negative to negative and is accompanied by the same-direction oscillation amplitude exceeding the dust emission precursor shaking amplitude threshold. In step S2, this is determined to be a state where the rapping action must be stopped immediately. Its physical meaning is the red line that a single rapping action must be terminated immediately.

[0046] The preset upper limit of the transient sampling step size refers to the maximum allowed sampling step size when performing dense sampling on the secondary current sampling source; this invention requires that this value not exceed 20 milliseconds, typically ranging from 5 to 10 milliseconds. The physical meaning of this value is that, in order for the first-order and second-order differences to effectively characterize the transient process of the rapping mechanism, the sampling step size must be significantly smaller than the characteristic duration of the mechanical impact of the rapping mechanism, which is typically 100 to 500 milliseconds.

[0047] The preset observation length refers to the window length used when performing a rolling average of the first-order difference mean and the second-order difference mean; typically, it is no less than 8 sampling points and no more than 32 sampling points. If the window is too short, it will be affected by random fluctuations; if the window is too long, the smoothing will be too excessive and it will not be able to characterize transients.

[0048] The preset plateau period fluctuation threshold is a criterion threshold used to determine whether the absolute value of the second-order difference mean has entered a stable range. In the plateau period state determination, the cumulative plateau period duration is allowed only when the absolute value of the second-order difference mean does not exceed this threshold. Typically, it is taken as 0.2 to 0.4 times the peeling period slope threshold.

[0049] The reference current value refers to the steady-state value of the secondary current immediately after the previous dust removal in the rapping zone has ended, the ash layer has been completely peeled off, and the dust in the ash hopper has stabilized. It serves as a reference for subsequent judgments and is automatically recorded in step S3.

[0050] The exponentially weighted moving average refers to a statistical method that smoothly integrates historical and new observations in a weighted manner. Its update formula is that the new threshold equals the attenuation coefficient multiplied by the new observation plus 1 minus the attenuation coefficient multiplied by the old threshold. In step S3, the attenuation coefficient ranges from not less than 0.1 to not greater than 0.4.

[0051] The preset gear refers to a set of discrete values ​​for the vibration acceleration; in this embodiment, 5 gears are used, namely the highest gear 1.0 times the calibration value, the second highest gear 0.8 times, the medium gear 0.6 times, the second lowest gear 0.4 times, and the lowest gear 0.2 times.

[0052] The following is a general overview of the implementation of the method described in this invention.

[0053] The first item in the general implementation overview, step S0, is the end-to-end implementation of the initialization phase. The inputs for this step are the physical design drawings of the electrostatic precipitator and the secondary current sampling interface on the secondary side of the high-voltage rectifier transformer. The core actions of this step are threefold. The first step is to divide all the dust collection plates into several rapping zones along the airflow direction according to the physical design of the electrostatic precipitator and number them in physical order from upstream to downstream. In this embodiment, the numbering method is to first number the four zones in each electric field as sub-zones A to D, and then advance them in order of electric field number 1 to 4, resulting in 32 rapping zones: 1A, 1B, 1C, 1D, 2A to 4D. The second step is to bind the secondary current sampling circuit of the high-voltage rectifier transformer of the electric field to each rapping zone as the secondary current sampling source. The third step is to set the initial peeling period slope threshold, plateau period duration step threshold, and dust pre-shaking amplitude threshold for each rapping zone according to the rated secondary current value specified in the electrostatic precipitator design drawings and the initial proportional coefficient given below. The output of this step is a vibration zone configuration table with numbered, bound sampling sources, and set initial thresholds. The adjustable parameters are the preset proportional coefficients of the peeling period slope threshold and the dust precursor shaking amplitude threshold.

[0054] The second item in the general implementation overview, step S1, is the end-to-end implementation of the rapping triggering phase. The inputs to this step are the rapping zone configuration table, the real-time average secondary current of each rapping zone, and the last dust removal time. This step has two core actions: first, sorting all rapping zones according to at least one of the three triggering criteria given below to obtain the sequence of zones to be rapped in this round of rapping; second, starting from the first position in the sequence of zones to be rapped, outputting a rapping start signal to the rapping mechanism corresponding to that zone to execute a single rapping action, and simultaneously immediately initiating dense sampling of the secondary current sampling source of that rapping zone at a value not exceeding the preset transient sampling step size upper limit, thereby constructing the secondary current sequence in real time. The output of this step is the rapping zone number currently performing a single rapping action and the synchronously densely sampled secondary current sequence stream. Adjustable parameters include the rapping triggering criterion combination method and the preset transient sampling step size upper limit.

[0055] The third item in the general implementation overview, step S2, is the end-to-end implementation of the three-stage real-time identification stage. The input to this step is the secondary current sequence output from step S1. The core action of this step is to perform rolling calculations of the first-order and second-order difference mean values ​​of the secondary current sequence at each sampling point, and make a three-way judgment based on the criteria of the main peeling period state, the plateau period state, and the secondary dust precursor state. If the main peeling period state is determined, the rapping start signal is continuously output to the rapping mechanism to continue rapping; if the plateau period state is determined, the rapping stops and prepares to wait for a judgment on whether to further transition to the secondary dust precursor state; if the secondary dust precursor state is determined, the rapping stops signal is immediately output to the rapping mechanism to end the current rapping action. The output of this step is the state sequence corresponding to each sampling point in this rapping action and the trigger time of the rapping stops signal. The adjustable parameters are the preset observation length and the preset plateau period fluctuation amplitude threshold.

[0056] The fourth item in the general implementation overview, step S3, is the end-to-end implementation of the threshold self-calibration phase. The inputs to this step are the measured plateau duration and peak value of the dust precursor vibration during this rapping session, output from step S2. The core action of this step is to update the peeling period slope threshold and the dust precursor vibration amplitude threshold using the exponentially weighted moving average. Specifically, the new threshold is obtained by multiplying the attenuation coefficient by the new observation value, adding a minus attenuation coefficient multiplied by the old threshold. The output of this step is the updated threshold used for the next rapping session in this rapping zone. The adjustable parameter is the attenuation coefficient of the exponentially weighted moving average.

[0057] The fifth item in the general implementation overview, step S4, is the end-to-end implementation of the cross-zone scheduling feedback phase. The input to this step is the state identification history of all rapping zones within the current traversal. The core action of this step is to reduce the rapping acceleration of rapping zones that only exhibit secondary dust storm precursor states but not the main spalling phase states in the current traversal by at least one preset level in the next round; and to increase the rapping acceleration of rapping zones that have identified the main spalling phase states in at least the last three traversals by one preset level. The output of this step is a rapping acceleration level configuration table for each rapping zone in the next round, with adjustable parameters being the threshold for determining the number of rounds of reduction and increase, and the number of discrete values ​​for the preset level.

[0058] The following provides a detailed step-by-step implementation example for each step.

[0059] Step S0 provides a detailed implementation example. This example considers a dual-chamber, four-field electrostatic precipitator for an 800 MW coal-fired power unit. Each field contains four rapping zones, for a total of 32 rapping zones. Each rapping zone is numbered from 1A to 4D according to the double-layer numbering method given above. The rated secondary current value of the field in each rapping zone is 800 mA, 600 mA, 400 mA, and 200 mA, depending on the field number. Along the airflow direction, as the field number increases, the dust concentration decreases, and the rated current decreases. In this embodiment, the preset proportional coefficient is set to 0.02, and the preset vibration proportional coefficient is set to 0.06. Specifically, the initial value of the peeling slope threshold for the No. 1 electric field rapping zone is 16 mA per sampling point (i.e., 800 mA multiplied by 0.02), and the initial value of the dust precursor vibration amplitude threshold is 48 mA (i.e., 800 mA multiplied by 0.06). For the No. 4 electric field zone, the thresholds are 4 mA per sampling point and 12 mA per sampling point. The initial value of the plateau duration step count threshold for all 32 rapping zones is uniformly set to 8 sampling points.

[0060] Step S1 is illustrated in Figure 5. The three triggering criteria in step S1 are logically organized with the first criterion satisfying the condition placed at the top of the sequence of zones to be vibrated. The upper limit of the preset transient sampling step size is 5 milliseconds. The first triggering criterion is activated when the relative deviation of the average secondary current of the electric field in the vibrating zone from the reference current value is not less than a first preset triggering threshold. In this embodiment, the first preset triggering threshold is 8%, meaning that the vibrating of this zone is triggered when the average secondary current of the electric field falls by more than 8% compared to the reference current. Simultaneously, the second triggering criterion is activated when the duration since the last vibration is not less than a second preset triggering threshold. In this embodiment, the second preset triggering threshold is 180 seconds, serving as a fallback trigger to avoid hardening of the ash layer under prolonged non-triggered conditions. Simultaneously, the third triggering criterion is activated when the precursory state of secondary dust is detected during the previous round of vibration of the upstream adjacent zone. This vibrating zone is triggered in advance according to the physical law that upstream dust is easily blown downstream. Each rapping zone is ranked first in the waiting rapping zone sequence according to the first of the three criteria. In a certain polling cycle of this embodiment, the dust concentration at the inlet of the electrostatic precipitator experiences a short-term spike due to fluctuations in furnace operating conditions. The average secondary current of electric field No. 2, where rapping zone 2A is located, drops from its reference current value of 600 mA to 540 mA, with a relative deviation of 10%. It first meets the first trigger criterion, so rapping zone 2A is ranked first in the waiting rapping zone sequence and immediately triggers the rapping action. At the same time, only 60 seconds have passed since the last rapping action of rapping zone 3B, with a relative deviation of only 4%, and it does not meet any trigger criterion, so it continues to wait. The three trigger criteria are combined according to the "first to be met" logic to ensure that the high-priority current drop criterion can quickly respond to operating condition disturbances, the medium-priority duration criterion can cover the dead zone where low dust concentration section does not trigger for a long period of time, and the low-priority upstream event criterion can proactively handle the transmission effect of upstream dust flowing to this zone with the airflow.

[0061] A detailed implementation of step S2 is shown in Figure 3. In step S2, the three typical states of the secondary current sequence during a single rapping process correspond to different time-series curve shapes. The main peeling period is characterized by a continuous linear rise in current; the plateau period is characterized by a steady current with slight fluctuations; and the secondary dust precursor period is characterized by a reversal of the current slope and the appearance of unidirectional oscillations. The sampling-point decision tree for step S2 is shown in Figure 4, with three levels of decision-making. The first level compares the first-order difference mean with the peeling period slope threshold to determine the main peeling period. The second level compares the absolute value of the second-order difference mean with the plateau period fluctuation amplitude threshold and the cumulative counter with the plateau period duration step threshold to determine the plateau period. The third level jointly determines the secondary dust precursor by combining the event of the second-order difference mean changing from non-negative to negative with the event of the unidirectional oscillation amplitude exceeding the dust precursor shaking amplitude threshold. The preset observation length is 16 sampling points, corresponding to an observation window of 16 multiplied by 5 milliseconds, i.e., 80 milliseconds. The preset plateau period fluctuation amplitude threshold is taken as 0.3 times the peeling period slope threshold, that is, 4.8 mA per sampling point for electric field zone 1. In this embodiment, after the 1A rapping zone is triggered, the mean value of the first-order difference of the densely sampled secondary current sequence jumps from 0 to 22 mA per sample point within the first 28 sampling points (corresponding to an actual duration of 140 ms), which is greater than the peeling period slope threshold of 16 mA per sample point, and is determined to be in the main peeling period state, and rapping is maintained; within the 29th to 42nd sampling points (corresponding to a duration of 70 ms), the mean value of the first-order difference drops to 11 mA per sample point, and the absolute value of the mean value of the second-order difference drops to 1.5 mA per sample point, which is less than the plateau period fluctuation amplitude threshold, and continues to reach 14 sampling points, which is greater than the plateau period duration step threshold of 8, and is determined to be in the plateau period state, and rapping is entered into a stop-vibration preparation state; at the 43rd sampling point, the mean value of the second-order difference changes from +0.8 to -5.6 mA per sample point, and the same-direction oscillation amplitude within the most recent 16 sampling points reaches 62 mA, which is greater than the dust precursor shaking amplitude threshold of 48. The signal was determined to be a precursor to secondary dust pollution, and a stop signal was immediately sent to the rapping mechanism. This rapping lasted for 43 x 5 milliseconds, or 215 milliseconds, which is about 46% shorter than the preset fixed duration of 400 milliseconds. This embodiment further provides details of the transient identification of a single rapping in rapping zone 1B as a comparison. Zone 1B is an adjacent downstream zone. The first-order difference mean of its single rapping fluctuates between 18 and 24 mA per sampling point in the first 32 sampling points, which is determined to be the main peeling period. It enters a plateau period in the 33rd to 48th sampling points. At the 49th sampling point, the second-order difference mean changes from +1.2 to -4.8 mA per sampling point, and the oscillation amplitude in the same direction is 54 mA, which is greater than the threshold of 48 mA, and is determined to be a precursor to dust pollution. This rapping lasted for 245 milliseconds.The difference in vibration stop time between two adjacent zones is about 30 milliseconds. This difference cannot be reflected in the existing preset fixed duration method. However, the method described in this invention enables each zone to automatically end vibration at its optimal stop time through real-time shape recognition.

[0062] Step S3 is illustrated in Figure 6. In step S3, the threshold self-calibration uses an exponentially weighted moving average principle. The attenuation coefficient determines the weight of the current measured observation in the new threshold, and the attenuation coefficient minus one determines the weight of the old threshold in the new threshold. The attenuation coefficient of the exponentially weighted moving average is 0.2. The first-order differential mean of the main stripping period measured in this vibration of zone 1A is 22 mA per sampling point, and the old stripping period slope threshold is 16 mA per sampling point. According to the exponentially weighted update formula, the new threshold equals 0.2 multiplied by 22 mA per sampling point plus 0.8 multiplied by 16 mA per sampling point, equaling 17.2 mA per sampling point. That is, 17.2 mA per sampling point will be used as the stripping period slope threshold for the next vibration of zone 1A. Similarly, the oscillation amplitude of 62 mA at the trigger time of the secondary dust precursor in this vibration is combined with the old threshold of 48 mA using exponential weighting to obtain a new threshold of 50.8 mA. This design, which replaces offline training with rolling average self-calibration, significantly lowers the deployment threshold of the method described in this invention in industrial settings, enabling the method to complete self-calibration convergence within 7 to 14 days after commissioning without relying on any prior dust resistivity database.

[0063] Step S4 is illustrated in Figure 7. The cross-zone scheduling feedback in step S4 includes two types of operations: downshifting and upshifting. Downshifting is triggered when only secondary dust precursors are detected in the current round S2, without identifying the main dustfall. Upshifting is triggered when the main dustfall is detected in at least three consecutive rounds of traversal. These two operations constitute the adaptive adjustment of the dust level. The preset dust levels are 5: the highest level (1.0 times the calibration value, i.e., 9.8 m / s²), the second highest level (0.8 times, i.e., 7.84 m / s²), the middle level (0.6 times, i.e., 5.88 m / s²), the second lowest level (0.4 times, i.e., 3.92 m / s²), and the lowest level (0.2 times, i.e., 1.96 m / s²). In this embodiment, one round of traversal takes approximately 11 minutes. In one round of traversal, the 4D rapping zone only identifies the secondary dust precursor state in step S2 but not the main peeling period state. That is, the slope of the transient secondary current recovery during this rapping never reaches the peeling period slope threshold before entering the dust precursor state, and it is determined to be an over-vibration zone. In this step, the rapping acceleration of the 4D rapping zone is reduced by one level from the highest level of 9.8 m / s² to the second-highest level of 7.84 m / s². In the subsequent three rounds of traversal, the 4D zone can identify the main peeling period state in each round. In this step, its operating condition is determined to be normal, and the rapping acceleration is increased by one level from the second-highest level back to the highest level. This reduction mechanism based on the criterion of "only dust precursor state without main peeling" and the recovery mechanism based on the criterion of "continuous main peeling" together constitute the adaptive adjustment of the gear. During long-term operation, each rapping zone will operate stably at the gear that matches its working conditions, without any manual intervention or experience-based adjustments.

[0064] The following is a commissioning example.

[0065] Example 1: A 660 MW coal-fired subcritical unit with a dual-chamber, four-field electrostatic precipitator is designed with an inlet dust concentration of 25 g / m³ and an outlet dust concentration not exceeding 30 mg / m³. The quality of the bituminous coal in this unit fluctuates significantly, and the dust resistivity is around 10... 9 Up to 10 12The variation between ohms and centimeters was observed. In the first 30 days after the method described in this invention was put into operation, the daily average concentration of dust at the outlet was 24.6 mg / m³, with the highest instantaneous peak value reaching 47 mg / m³. There were 11 instances of secondary dust exceeding the standard per month. From the 7th day after the method described in this invention was put into operation, the peeling slope threshold and the dust precursor vibration amplitude threshold of the 32 rapping zones have completed self-calibration convergence through exponential weighted moving average. Within 30 days after the method was put into operation, the daily average concentration of exhaust dust dropped to 17.8 mg / m³, the highest instantaneous peak value was 26 mg / m³, and there were 0 instances of secondary dust short-term exceedance events per month. During the same period, the average duration of a single rapping session dropped from 400 ms to 235 ms, and the rapping energy consumption decreased by about 41%. The rapping acceleration level of the No. 4 zone group of the final electric field was automatically stabilized at the medium level through the cross-zone scheduling feedback mechanism, and the peak vibration acceleration of the mechanical wear monitoring item of the rapping mechanism dropped from a monthly average of 14.2 m / s² to 5.9 m / s². As shown in Figure 8, this embodiment compares three key performance indicators—average duration of a single vibration, average energy consumption of a single vibration, and main body ash layer peeling rate—with the existing preset fixed duration vibration method as a benchmark of 100 before commissioning. After the method of the present invention is put into operation, the average duration of a single vibration is reduced to 58.8 of the benchmark, the average energy consumption of a single vibration is reduced to 59.0 of the benchmark, while the main body ash layer peeling rate increases to 102.5 of the benchmark. That is, while significantly reducing vibration energy consumption and vibration duration, the main body ash layer peeling ability is slightly improved, reflecting the optimal balance between energy consumption and effect of the method of the present invention. As shown in Figure 10, this embodiment tracked the daily dust concentration at the outlet of the electrostatic precipitator for 30 days before and 30 days after commissioning. Before commissioning, the daily average dust concentration at the outlet fluctuated around 24.6 mg / m³, with instantaneous peak values ​​frequently exceeding the design emission limit between 30 and 50 mg / m³, resulting in 11 short-term secondary dust exceedances within the month. After commissioning, the daily average dust concentration at the outlet rapidly decreased to around 17.8 mg / m³ and stabilized, with instantaneous peak values ​​consistently below the design emission limit between 12 and 26 mg / m³, and no short-term secondary dust exceedances within the month. The comparison clearly demonstrates the ability of the method described in this invention to suppress secondary dust events.

[0066] Example 2: An electrostatic precipitator at the head of a 360 square meter sintering machine has 4 electric fields, each containing 3 rapping zones, for a total of 12 rapping zones. The incoming sintering material fluctuates greatly, and the dust resistivity is around 10... 10 Up to 10 13The oscillations between ohms and centimeters are drastic, easily leading to back corona and secondary dust emission phenomena in the final electric field. Before commissioning, a preset fixed-duration rapping route was used, with a fixed rapping duration of 600 milliseconds and a fixed rapping cycle of 240 seconds. When the method described in this invention was implemented, the upper limit of the preset transient sampling step size was set to 8 milliseconds, the preset observation length was set to 12 sampling points, the attenuation coefficient was set to 0.25, and the preset speed range remained at 5 levels. After 14 days of commissioning, threshold self-calibration convergence was completed. The daily average concentration of exhaust dust decreased from 36 mg / m³ to 22 mg / m³, and the number of short-term secondary dust emission events decreased from an average of 18 per month to an average of 2 per month. The average duration of a single rapping event decreased from 600 milliseconds to 268 milliseconds. The rapping acceleration speed range of the two zones in the final electric field stabilized at the second-highest level through a cross-zone scheduling feedback mechanism. Plate wear checks showed that the monthly plate deformation decreased from 1.4 mm to 0.6 mm. As shown in Figure 9, in this embodiment, the electrostatic precipitator at the sintering machine head is used when the dust resistivity is at 10. 9 Up to 10 13 The secondary dust suppression rate was systematically tested using a gridded method under two-dimensional conditions with ohmic-cm and flue gas velocity ranging from 0.8 m / s to 1.6 m / s. The method described in this invention was tested under dust resistivity of 10 ohms / cm. 11 Under optimal operating conditions with an ohm·cm and a flue gas velocity of 1.0 m / s, the secondary dust suppression rate reaches 96.1%; at a dust resistivity of 10 ohms·cm, the secondary dust suppression rate reaches 96.1%. 13 Under extreme conditions of ohm·cm (typical back corona risk zone) and flue gas velocity of 1.6 m / s (typical peak load), the secondary dust suppression rate remained at 76.5%, significantly higher than the 42% level of existing preset fixed-duration methods under the same conditions; in more typical medium conditions (dust resistivity 10 ohms·cm), the secondary dust suppression rate remained at 76.5%, significantly higher than the 42% level of existing preset fixed-duration methods under the same conditions; 10 Up to 10 12 Within a nine-grid area (within which ohm·cm and flue gas velocity ranges from 0.8 m / s to 1.4 m / s), the secondary dust suppression rate is no less than 88%. This two-dimensional operating condition coverage test result shows that the effectiveness of the method described in this invention does not heavily rely on the dust resistivity, a core and challenging parameter in traditional electrostatic precipitator control. It can maintain good secondary dust suppression capability even in the back corona risk zone, demonstrating significant operating condition robustness.

[0067] Example 4 is a long-term steady-state example. This example examines the steady-state performance of the dual-chamber four-field electrostatic precipitator of a 660 MW coal-fired subcritical unit described in Example 1 after 6 months of operation using the method described in this invention. During the 6-month operation period, the peeling slope threshold and dust precursor shaking amplitude threshold of the 32 rapping zones have fully converged through exponentially weighted moving averages. The monthly variation of the threshold for each zone does not exceed ±3%, indicating that the threshold self-calibration mechanism in step S3 has good convergence stability. The rapping acceleration level distribution of the 32 rapping zones is as follows: all four zones of field 1 are stable at the highest level of 9.8 m / s², all four zones of field 2 are stable at the second-highest level of 7.84 m / s², two zones of field 3 are stable at the second-highest level and two zones are stable at the middle level of 5.88 m / s², and one zone of field 4 is stable at the middle level and three zones are stable at the second-lowest level of 3.92 m / s². That is, the closer to the outlet along the airflow direction, the lower the rapping acceleration automatically becomes. This level distribution is completely consistent with the physical law of dust load decreasing along the airflow direction of the electrostatic precipitator, indicating that step S4... The cross-zone scheduling feedback mechanism exhibits good physical interpretability. Within a 6-month period, the cumulative number of triggers for the main spalling phase state identification was 13,428, the plateau phase state identification was 11,236, and the secondary dust precursor state identification was 426, with the latter accounting for 1.7%. This indicates that the method described in this invention has, in long-term operation, converged the vast majority of single rapping actions to an active cessation of vibration during the plateau phase, and only in a very few boundary conditions to a passive cessation of vibration due to dust precursors, achieving a dual advantage in rapping energy consumption and spalling rate. Within a 6-month period, the mechanical maintenance workload of the rapping mechanism (cumulative replacement of rapping hammers, cumulative repair of rapping transmission arms, etc.) decreased by approximately 53% compared to the same period before commissioning, indicating that the downshifting mechanism in the cross-zone scheduling feedback of step S4 effectively reduced unnecessary over-vibration and extended the mechanical life of the rapping mechanism.

[0068] Example 3 illustrates the module correspondence of the system described in this invention. The initialization module is deployed in the main programmable logic controller (PLC) within the low-voltage control cabinet of the electrostatic precipitator. Initialization inputs include a digital import interface for the electrostatic precipitator physical design drawings and configuration files binding the secondary current sampling circuits of each electric field. The rapping trigger module is deployed in the same PLC, reading the average secondary current of each electric field and the last rapping time of each rapping zone via a communication bus, and executing trigger criterion sorting. The three-stage real-time identification module is implemented with an independent high-speed real-time controller. Its sampling interface is directly connected to the secondary side of the current transformer on the secondary side of the high-voltage rectifier transformer, with a sampling frequency of not less than 200 Hz and a cycle period of not more than 5 milliseconds for differential calculation and state determination. The threshold self-calibration module uses an independent non-volatile storage area organized by rapping zone number, updating the corresponding record after each rapping action. The cross-zone scheduling feedback module runs as an independent task within the main PLC, executing a gear adjustment decision after each round of traversal. Modules communicate asynchronously via an industrial fieldbus to ensure that the hard real-time performance of the three-stage real-time identification module is not affected by other modules. The adaptive dust removal control device for the electrostatic precipitator is installed in an industrial-grade control cabinet. The mainboard inside the cabinet is equipped with a high-speed real-time controller, a programmable logic controller, secondary side access terminals of current transformers, rapping drive interface terminals, a human-machine interface display screen, an uninterruptible power supply, and a communication gateway that interfaces with the plant-level distributed control system. The high-speed real-time controller and the programmable logic controller exchange hard real-time data through a dual-port random access memory to ensure that the cycle period of the three-stage real-time identification module is no more than 5 milliseconds and is not affected by other modules.

[0069] Example 5, namely the boundary scenario example, examines the engineering performance of the method described in this invention under two types of boundary scenarios. The first type of boundary scenario is the cold-state first round of rapping after the electrostatic precipitator has been shut down and restarted. In this scenario, the ash layer accumulated in each rapping zone after the shutdown interval is relatively thick, and the ash layer adhesion characteristics are between "wet adhesion" and "dry blockage". The peak value of the first-order difference mean during the main peeling period is often 30% to 50% higher than that during steady-state operation. This embodiment uses the initial peeling slope threshold set in step S0, i.e., 16 mA per sampling point for electric field zone 1. After the first rapping test, the measured first-order difference average reaches 32 mA per sampling point. After the exponential weighted update in step S3, the threshold rises to 19.2 mA per sampling point. After 6 to 8 rounds of rapping, the threshold converges to near the stable value under this condition. During the convergence process, none of the rapping zones triggered the downshifting action in step S4, indicating that the threshold self-calibration mechanism in step S3 has adaptive tracking capability for the slightly higher first-order difference in the cold state first round, and will not misjudge over-rapping due to cold state deviation. The second type of boundary scenario is a short-term step decrease in dust concentration at the inlet of the electrostatic precipitator. In this scenario, some rapping zones may not reach the current drop threshold of the first trigger criterion in step S1 for 10 to 20 rounds of traversal. However, due to the fallback mechanism of the second preset trigger threshold of 180 seconds in the second trigger criterion, these rapping zones can still trigger rapping at a 180-second cycle. In actual measurement, the average duration of a single rapping action triggered by the fallback in this scenario is only 110 milliseconds, which is much shorter than 235 milliseconds under steady-state conditions. Moreover, the average first-order difference of the main stripping period of this rapping is often lower than the threshold. Step S2 directly determines that after a brief plateau period, dust storm precursors will immediately begin. After 1 to 2 rounds of observation, step S4 reduces the intensity of the zone to match the rapping energy with the current low-load conditions.

[0070] Example 6 is a parallel comparative example with existing methods. In this example, on the same electrostatic precipitator, 32 rapping zones were divided into two groups. The left half of the 16 zones adopted the complete S0 to S4 five-step control of the method described in this invention, while the right half of the 16 zones retained the original preset fixed-duration rapping method, with a fixed rapping duration of 400 milliseconds and a fixed rapping cycle of 180 seconds, serving as a control group. The two groups of rapping zones were independently scheduled by two parallel tasks within the same programmable logic controller. During a continuous 30-day synchronous comparison period, the 16 zones on the method side of this invention were triggered a total of 4288 times, with an average single rapping duration of 232 milliseconds, a cumulative reduction in rapping mechanical energy consumption of 41%, and 2 secondary dust events. The 16 zones on the control group were triggered a total of 4320 times, with an average single rapping duration of 400 milliseconds and 14 secondary dust events. The average difference in dust concentration at the electric field outlet between the two groups was 6.4 mg / m³, with the method side of this invention significantly lower than the control group. This embodiment further illustrates that the effectiveness of the method described in this invention can be verified without difference by synchronous comparison on the same device, and there is no misjudgment caused by differences in operating conditions, demonstrating the clear advantages of the method described in this invention over existing methods.

[0071] The following are illustrative examples of several possible variations of the method described in this invention, so that those skilled in the art can understand the adaptability of the method in different industrial settings. These examples should not be construed as limiting the scope of protection claimed by this invention. Variation 1: The rolling calculation of the first-order difference mean and the second-order difference mean in step S2 can be equivalently replaced by other statistical quantities with the same physical meaning, such as the first-order derivative estimate after low-pass filtering, the least squares fitting slope, the Savitzky-Golay filter output, etc. As long as the statistical quantity can reflect the two core physical meanings of "the average rate of change and the average acceleration of the secondary current sequence within the most recent preset observation length," step S2 can function normally. Variation 2: The exponentially weighted moving average in step S3 can also be equivalently replaced by other rolling average methods with similar convergence characteristics, such as simple moving average, weighted simple moving average, Kalman filter estimation, etc. In modified embodiment 3, the number of preset speed levels in step S4 can be adjusted according to the type of rapping mechanism and the adjustable range of rapping acceleration of the specific electrostatic precipitator, and is not limited to the 5 speed levels in this embodiment. In a minimally simplistic rapping mechanism with only two adjustable rapping acceleration levels, step S4 can also be reduced to a simple "two-speed switching" implementation. In modified embodiment 4, the granularity of the rapping partitioning in the method of the present invention can be adjusted according to the physical design of the specific electrostatic precipitator; the granularity can be coarse or fine. At the finest granularity, each pair of adjacent dust collecting plates can be considered an independent rapping partition, while at the coarsest granularity, the entire electric field can be considered a rapping partition. Those skilled in the art can make corresponding selections based on the controllable granularity of the specific rapping mechanism. Modified Embodiment 5: The method of the present invention is applicable not only to dry mechanical rapping electrostatic precipitators, but also to electromagnetic pulse rapping and acoustic cleaning electrostatic precipitators. For acoustic cleaning electrostatic precipitators, the rapping action in step S1 is replaced by acoustic excitation action, the transient sequence of secondary current in step S2 is replaced by the electric field secondary current sequence during acoustic excitation, and the rapping acceleration in step S4 is replaced by the sound pressure level of acoustic excitation, while the core mechanism of each step remains unchanged. Modified Embodiment 6: When the method of the present invention is used in conjunction with existing high-voltage power supply control methods such as spark rate control, maximum average voltage control, and critical spark control, it can operate in parallel with the high-voltage power supply control method. The two are coordinated by sharing the secondary current sampling source, and there is no functional conflict. In long-term operation, the secondary dust suppression effect of the method of the present invention and the corona power optimization effect of the high-voltage power supply control method can be superimposed, jointly improving the overall dust removal efficiency of the electrostatic precipitator.

[0072] In modified implementation 7, the rapping triggering criteria in step S1 can be extended according to the specific process characteristics of an industry. For example, for cement kiln head electrostatic precipitators, a fourth criterion linked to the kiln head feed rate change rate can be added in addition to the three criteria, making the rapping action more sensitive to process disturbances. For steel sintering electrostatic precipitators, a fifth criterion linked to the sintering trolley running speed can be added in addition to the three criteria. Each extended criterion should be logically combined with the original three criteria according to the principle of "the one that is satisfied first is placed at the top of the rapping zone sequence".

[0073] In modified implementation method 8, the initial values ​​of the peeling period slope threshold, plateau period continuous step number threshold, and dust precursor vibration amplitude threshold of each rapping zone in step S0 can be directly given by the on-site commissioning personnel based on experience under certain working conditions, instead of being automatically set by multiplying the rated secondary current value by a preset proportional coefficient. After the on-site commissioning personnel give the initial values, the exponential weighted moving average self-calibration mechanism in step S3 will automatically take over the subsequent updates of the thresholds, so that the influence of the accuracy of the initial values ​​on the final thresholds will automatically disappear within 6 to 14 days.

[0074] Example 7, i.e., examples under different coal quality conditions, examines the performance differences of a 660 MW coal-fired subcritical unit described in Example 1 using the method of this invention under three typical coal quality conditions. The first coal quality condition is high-ash lignite, with dust resistivity in the back corona risk range of 10. 12 Up to 10 13 The daily average concentration of flue gas at the outlet decreased to 21.4 mg / m³; the second type of coal quality was medium-ash bituminous coal, with a dust resistivity of 10 ohms-cm. 10 Up to 10 11 In the optimal range of ohm-cm, the daily average concentration of smoke dust at the outlet decreased to 14.8 mg / m³; the third type of coal quality operating condition involves the blending of low-ash anthracite, with dust resistivity at 10. 9 Up to 10 10 In the low resistivity range of ohm-cm, the daily average concentration of flue dust at the outlet decreased to 18.6 mg / m³. Under all three coal quality conditions, the method described in this invention can complete threshold self-calibration convergence within 7 to 14 days, and the number of secondary dust short-term exceedance events is reduced to no more than once per month. This stable performance across coal quality conditions is attributed to the rolling average self-calibration mechanism in step S3, which allows the threshold to automatically adapt to the actual dust characteristics on site after commissioning, independent of prior coal quality parameters.

[0075] Example 8 is an example of a cement kiln head electrostatic precipitator. This example examines the performance of a kiln head electrostatic precipitator operating the method described in this invention on a 5000-ton-per-day new dry-process cement production line. The kiln head electrostatic precipitator handles high-temperature dust-laden flue gas with temperatures reaching 350 to 400 degrees Celsius and dust resistivity around 10. 8 Up to 10 10 The ohm-cm range is relatively low. This kiln head electrostatic precipitator has a single-chamber, three-field structure, with each field containing five rapping zones, for a total of 15 rapping zones. When the method described in this invention is put into operation, the upper limit of the preset transient sampling step size is set to 6 milliseconds, the preset observation length is 14 sampling points, the attenuation coefficient is 0.2, and the preset speed is 4 levels. Nine days after commissioning, the threshold self-calibration convergence was completed, the daily average outlet dust concentration decreased from 28 mg / m³ to 16 mg / m³, and the number of short-term secondary dust exceedances decreased from an average of 7 times per month to an average of 0 times per month; the average duration of a single rapping event decreased from 350 milliseconds to 198 milliseconds; and the rapping acceleration levels of the two zones in the final field stabilized at the medium level through a cross-zone scheduling feedback mechanism. This embodiment further illustrates that the method described in this invention has good adaptability to typical working conditions in the cement industry and a low threshold for cross-industry deployment.

[0076] Example 9, a comparative example with manual adjustment based on experience, compares the method described in this invention with the manual adjustment method used by experienced operators on a 600 MW coal-fired power unit's electrostatic precipitator. The experienced operators manually adjust the rapping cycle and duration of each rapping zone based on the on-site dust resistivity, flue gas volume, and outlet concentration trends to suppress secondary dust emissions. During a 60-day comparison period, the method described in this invention resulted in an average of 0.5 short-term secondary dust emission exceedances per month, while the manual adjustment method used by experienced operators resulted in an average of 4.2 events per month. The rapping energy consumption of the method described in this invention decreased by 38% compared to the initial value, while the manual adjustment method used by experienced operators decreased by 12%. This example demonstrates that the method described in this invention significantly outperforms the manual adjustment method used by experienced operators in both secondary dust suppression and rapping energy consumption optimization, reflecting the inherent advantages of shape recognition adaptive control methods over manual experience-based control methods.

[0077] In summary, the adaptive dust removal control method and corresponding system for electrostatic precipitators described in this invention achieves precise control of the timing of vibration cessation by subdividing the single rapping process into three internal stages: the main peeling period, the plateau period, and the precursor to secondary dust emission. It uses the first-order and second-order difference mean values ​​of the secondary current sequence as real-time identification quantities for each sampling point. Through a rolling average self-calibration mechanism, the criterion thresholds for each rapping zone automatically adapt to the on-site dust characteristics after commissioning, without requiring any prior database or offline training. Furthermore, through the "only precursor to dust emission without main peeling" reduction mechanism and the "continuous main peeling" recovery mechanism in the cross-zone scheduling feedback, it achieves adaptive adjustment of the rapping acceleration level during long-term operation. All calculations in the method described in this invention have constant-level complexity, do not introduce new sensors or neural network inference, and can be directly deployed in the low-voltage control system of existing electrostatic precipitators. Those skilled in the art can implement the method described in this invention without creative effort based on the embodiments disclosed herein and the illustrative examples above. They can also make adaptive adjustments to the preset thresholds, preset gears, and preset observation lengths according to the specific working conditions of the industrial setting, all of which fall within the scope of protection claimed by this invention.

[0078] Finally, it should be noted that the above embodiments and examples are only used to illustrate the technical solutions of the present invention and not to limit them. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention, such as the specific statistical quantity replacement used in the three state criteria in step S2, the specific replacement of the rolling average method in step S3, the specific adjustment of the number of preset gears in step S4, and the specific adjustment of the granularity of each vibration zone, should all fall within the protection scope of the present invention. Those skilled in the art can make various modifications to the method of the present invention within the spirit and scope of the present invention without creative effort, based on the above embodiments and examples.

Claims

1. An adaptive dust removal control method for an electrostatic precipitator, characterized in that, Includes the following steps: S0, Initialization: Number the several rapping zones of the electrostatic precipitator along the airflow direction according to their physical locations and bind their secondary current sampling sources. Set the initial peeling period slope threshold, plateau period duration step threshold and dust precursor shaking amplitude threshold. S1, Vibration trigger: The vibration partitions are sorted according to a preset trigger criterion, and a single vibration action is executed starting from the first position. At the same time, the secondary current sampling source is sampled at a value not greater than the upper limit of the preset transient sampling step size to obtain a secondary current sequence. S2, Three-stage real-time identification: The secondary current sequence is judged point by point according to the following three states; During the main peeling phase, if the mean value of the first-order difference of the secondary current sequence within the most recent preset observation length is greater than the peeling phase slope threshold, the rapping action continues. In the plateau phase, the mean of the first-order difference is not greater than the slope threshold of the peeling phase, and the absolute value of the mean of the second-order difference does not exceed the preset plateau phase fluctuation amplitude threshold, and continues for no less than the plateau phase duration threshold, then enters the oscillation stop preparation phase. In the secondary dust storm precursor state, the second-order difference mean changes from non-negative to negative, and the accompanying oscillation amplitude exceeds the dust storm precursor shaking amplitude threshold, the vibration action is immediately stopped. S3, Threshold self-calibration: After this vibration is completed, the measured number of steps during the plateau period and the peak value of the dust precursor shaking are combined into the corresponding threshold using a rolling average method. S4, Cross-zone scheduling feedback: Traverse all vibration zones in the order of steps S1 to S3; At the end of each round, for the rapping zone that only identifies the secondary dust precursor state but not the main peeling stage state, the rapping acceleration of the next round shall be reduced by no less than one preset level.

2. The method according to claim 1, characterized in that, The preset trigger criterion in step S1 adopts at least one of the following three criteria; The first criterion is that the relative deviation between the average value of the secondary current of the electric field where the rapping zone is located and the reference current value when the previous dust cleaning of the rapping zone just ended is not less than the first preset trigger threshold. The second criterion is that the time elapsed since the last vibration action was performed in the vibration zone is not less than the second preset trigger threshold. The third criterion is that in the most recent round of vibration action of the vibration zone immediately adjacent to the upstream zone, step S2 identifies the precursory state of secondary dust.

3. The method according to claim 1, characterized in that, In step S0, the initial value of the peeling period slope threshold is taken from the reference current value specified in the electric field design book of the rapping zone multiplied by a preset proportional coefficient; the initial value of the plateau period duration step threshold is taken from no less than 6 sampling points; and the initial value of the dust precursor shaking amplitude threshold is taken from the reference current value multiplied by a preset shaking proportional coefficient. The preset proportional coefficient has a value range of not less than 0.005 and not greater than 0.05; The preset jitter ratio coefficient has a value range of not less than 0.02 and not greater than 0.

1.

4. The method according to claim 1, characterized in that, The upper limit of the preset transient sampling step size in step S1 is no greater than 20 milliseconds; In step S2, the preset observation length is no less than 8 sampling points and no more than 32 sampling points; The oscillation amplitude mentioned in step S2 refers to the absolute value of the difference between the peaks and valleys of the secondary current sequence within the most recent preset observation length.

5. The method according to claim 1, characterized in that, The rolling average method in step S3 adopts an exponentially weighted moving average, and its decay coefficient ranges from not less than 0.1 to not greater than 0.

4. The updated value of the peeling slope threshold is obtained by weighting its current value with the mean of the first difference under the current main peeling state of the rapping. The updated value of the dust precursor vibration amplitude threshold is obtained by weighting its current value with the oscillation amplitude in the same direction at the time when the secondary dust precursor state is triggered by the current vibration.

6. The method according to claim 1, characterized in that, The preset gear in step S4 consists of no less than 3 discrete acceleration gears, and the initial value of the vibration acceleration of all vibration zones is uniformly taken from the highest gear. After each action that reduces the speed by at least one preset level, if the main peeling phase state can be identified in each of the at least three consecutive traversals of the rapping zone, then the rapping acceleration of the rapping zone will be increased back by one preset level.

7. An adaptive dust removal control system for an electrostatic precipitator, characterized in that, include: The initialization module is used to number the several rapping zones of the electrostatic precipitator along the airflow direction according to their physical locations and bind their secondary current sampling sources, and set the initial peeling period slope threshold, plateau period duration step threshold and dust pre-shaking amplitude threshold. The vibration triggering module is used to sort the vibration partitions according to a preset triggering criterion, execute a single vibration action starting from the first partition to be vibrated, and sample the secondary current sampling source to obtain a secondary current sequence with a value not greater than the upper limit of the preset transient sampling step size. The three-stage real-time identification module is used to determine the secondary current sequence point by point according to one of the three states: the main peeling period state, the plateau period state, and the secondary dust precursor state. The main peeling period state is used to determine whether to continue the rapping action, the plateau period state is used to determine whether to enter the preparation for stopping the rapping action, and the secondary dust emission precursor state is used to determine whether to stop the rapping action immediately. The threshold self-calibration module is used to combine the measured plateau duration steps and the dust precursor shaking peak value into the peeling period slope threshold and the dust precursor shaking amplitude threshold by rolling average after the end of this rapping. The cross-zone scheduling feedback module is used to sequentially traverse all rapping zones, and for rapping zones that only identify the secondary dust precursor state but not the main peeling period state in the three-stage real-time identification module, the rapping acceleration of the next round is reduced by no less than one preset level from the preset calibration value.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.

9. A computer program product, characterized in that, When the computer program product is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.

10. An adaptive dust removal control device for an electrostatic precipitator, comprising a memory and a processor, wherein the memory stores a computer program executable by the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6; The device further includes a rapping drive interface and a secondary current sampling interface electrically connected to the processor. The rapping drive interface is used to output a rapping action start / stop signal to the rapping mechanism of the electrostatic precipitator. The secondary current sampling interface is used to collect current signals from the secondary side of the high-voltage rectifier transformer of the electrostatic precipitator and send them to the processor.

Images