Search

CN-121978959-A - Digital twinning-based operation control method of special PLC (programmable logic controller) integrated machine for boiler

CN121978959ACN 121978959 ACN121978959 ACN 121978959ACN-121978959-A

Abstract

The invention discloses a special PLC all-in-one machine operation control method based on digital twinning, which comprises the steps of collecting boiler control input data and measurement output data and executing time alignment processing to form an input sequence and an output sequence, executing cleaning processing by a digital twinning system to generate standardized operation data, constructing an input Hankel matrix and an output Hankel matrix, calculating and extracting a time delay parameter set, carrying out line offset processing on the Hankel matrix and packaging the Hankel matrix into DeePC data packets, intercepting an initialization input segment and an initialization output segment, executing improved DeePC algorithm solution and introducing a phase maintaining mechanism into constraint to generate a control input quantity, and carrying out amplitude limiting on the control input quantity and returning the input output data. The invention improves the control stability and engineering applicability of the boiler under the multivariable coupling working condition.

Inventors

  • WANG XIANGFENG
  • CHEN JIAN

Assignees

  • 浙江迪恩智控科技有限公司

Dates

Publication Date
20260505
Application Date
20260309

Claims (8)

  1. 1. The operation control method of the special PLC all-in-one machine for the boiler based on digital twinning is characterized by comprising the following steps of: collecting boiler control input data and measurement output data, and executing time alignment processing to form an input sequence and an output sequence; Step two, the digital twin system receives the input sequence and the output sequence and executes cleaning treatment to generate standardized operation data; Thirdly, constructing an input Hankel matrix and an output Hankel matrix based on the standardized operation data, and executing initialization partitioning and prediction partitioning; performing cross-correlation calculation on preset control variables and measurement variables based on historical data for constructing the Hankel matrix, extracting a time delay parameter set, performing line offset processing on the Hankel matrix based on the time delay parameter set, and packaging the data into DeePC data packets; Step five, the PLC all-in-one machine receives the DeePC data packet, and intercepts an initialization input fragment and an initialization output fragment from a cache; Step six, based on the DeePC data packet, the initialization input segment and the initialization output segment, executing improved DeePC algorithm solution, introducing phase maintaining constraint based on the time delay parameter set into improved DeePC algorithm constraint, and generating control input quantity; and seventhly, executing amplitude limiting processing on the control input quantity and sending corresponding input and output data to the digital twin system.
  2. 2. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system as claimed in claim 1, wherein the first step is specifically as follows: The PLC all-in-one machine collects boiler control input data and measurement output data in the same sampling period, wherein the control input data comprises a fuel given value, a primary air given value, a secondary air given value and a water supply given value, and the measurement output data comprises main steam pressure, main steam temperature, steam drum water level, flue gas oxygen content and nitrogen oxide concentration; In order to control input data and measurement output data to be written with sampling time stamps respectively, the sampling time stamps are generated by an internal clock source of the PLC all-in-one machine, and the internal clock source is kept consistent with a boiler field unified time reference in a network time protocol synchronization or IEEE1588 precision time protocol synchronization mode; Performing time alignment processing on the control input data and the measurement output data, wherein the time alignment processing comprises the steps of constructing a target time sequence by taking a sampling period as a target time reference, and determining the time position corresponding to each sampling point in the target time sequence according to a sampling time stamp; when the sampling time stamp is offset from the target time sequence, resampling the offset sampling points by adopting a linear interpolation mode of two adjacent sampling points, wherein the linear interpolation takes the target time in the target time sequence as an independent variable, and an interpolation result is constructed by using the values of the two adjacent sampling points; the control input data for completing the time alignment processing is arranged in time sequence to form an input sequence, and the measurement output data for completing the time alignment processing is arranged in time sequence to form an output sequence; And writing a validity flag for each sampling point in the input sequence and the output sequence, wherein the validity flag comprises an original sampling flag and an interpolation resampling flag, and the validity flag and the corresponding sampling point number value form a marked data pair.
  3. 3. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system as claimed in claim 1, wherein the second step is specifically as follows: the digital twin system receives an input sequence and an output sequence from the PLC all-in-one machine through an industrial communication interface and writes the input sequence and the output sequence into an original data buffer area according to a sampling time sequence; constructing a sliding time window in the original data buffer area aiming at each channel data, calculating indexes of a mean value, a standard deviation and a change rate of the data in the sliding time window, and marking sampling points, the deviation mean value of which exceeds a preset multiple standard deviation and the change rate of which exceeds a threshold value, as abnormal points; Performing replacement processing on the sampling points marked as abnormal points, wherein the replacement mode comprises replacing abnormal point values by linear interpolation results of the effective sampling points adjacent to each other before and after replacement, and writing an abnormal replacement mark after replacement; Performing interval positioning on the continuous missing sampling interval, generating a missing interval mark according to the length of the missing interval and a preset allowable length threshold, and complementing the missing interval meeting the allowable length threshold by adopting an adjacent effective sampling point interpolation mode; Executing low-pass digital filtering processing on the channel data after the completion and abnormal replacement, wherein the low-pass digital filtering adopts a finite impulse response filter or an infinite impulse response filter, and configures cutoff frequency parameters according to the dynamic response characteristics of each measured quantity of the boiler; Writing the input sequence and the output sequence after the exception handling, the missing complement and the filtering handling into a historical data storage area, and writing an exception replacement mark, a missing complement mark and a filtering handling mark for each sampling point to form standardized operation data.
  4. 4. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system as claimed in claim 1, wherein the third step is as follows: The digital twin system reads continuous sampling points from the historical data storage area in time sequence to form a historical data section, and the historical data section simultaneously comprises control input channel data and measurement output channel data; Performing length verification on the historical data segment according to the preset initialization time domain length and the predicted time domain length, and entering the next processing when the historical data segment length meets the sum of the initialization time domain length and the predicted time domain length; Sequentially intercepting a plurality of data subsections with the same length on a historical data section in a sliding window mode, arranging the data subsections in time sequence to form an input Hankel matrix and an output Hankel matrix, wherein the input Hankel matrix is formed by controlling input channel data, and the output Hankel matrix is formed by measuring output channel data; Performing line segmentation processing on the input Hankel matrix and the output Hankel matrix according to the initialized time domain length to form an initialized input matrix block and an initialized output matrix block, and performing line segmentation processing on the input Hankel matrix and the output Hankel matrix according to the predicted time domain length to form a predicted input matrix block and a predicted output matrix block.
  5. 5. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system as claimed in claim 1, wherein the fourth step is specifically as follows: The digital twin system selects a preset control input channel and a measurement output channel in a historical data segment for constructing an input Hankel matrix and an output Hankel matrix to form a variable pair set, wherein the variable pair set at least comprises fuel given and main steam temperature, fuel given and main steam pressure, water supply given and steam drum water level, primary air given and flue gas oxygen content; for each variable pair, extracting a corresponding time sequence under a unified time index, and respectively constructing a multi-scale resampling sequence under an original sampling scale, a double sampling scale and a quadruple sampling scale; Respectively executing delay scanning processing under each sampling scale, wherein the delay scanning processing comprises translating one time sequence one by one within a preset maximum delay range, calculating dot product correlation values with the other time sequence, and forming a delay response curve corresponding to the scale; screening delay intervals with continuous multiple delay position correlation values exceeding a preset consistency threshold value in a delay response curve of each scale, and marking the delay intervals as candidate phase sections; Performing intersection operation on candidate phase sections obtained under different sampling scales, extracting delay positions which occur simultaneously in a plurality of scales as stable phase delays, and recording the stable phase delays as integer sampling steps; Based on the stable phase delay execution frequency statistics respectively calculated by a plurality of continuous historical time windows, selecting delay values appearing in the historical time windows with the preset proportion as final time delay parameters; Performing row offset processing on the input Hankel matrix and the output Hankel matrix according to the final time delay parameter, wherein the row offset processing comprises performing forward shift or backward shift rearrangement on matrix rows belonging to the same variable pair according to the corresponding delay step number, and performing deletion processing on generated vacancy rows; And combining and packaging the input Hankel matrix, the output Hankel matrix, constraint parameters, weight parameters and version identifiers after the line offset processing to obtain DeePC data packets.
  6. 6. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system as claimed in claim 1, wherein the fifth step is specifically as follows: the PLC all-in-one machine receives DeePC data packets through an industrial communication interface, performs consistency verification processing on version identifiers and verification fields in the data packets, and writes the data packets passing the verification into a control parameter buffer area; Reading the initialized time domain length parameter in the control parameter buffer area, and searching the corresponding number of nearest sampling points in the operation data buffer area according to the time sequence; intercepting control input data corresponding to the initialized time domain length from the operation data buffer area according to the control input channel in sequence to form an initialized input segment; Sequentially intercepting measurement output data corresponding to the initialized time domain length from the operation data buffer area according to the measurement output channels to form initialized output fragments; And respectively writing a time index number and a channel identifier for the initialization input segment and the initialization output segment.
  7. 7. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system according to claim 1, wherein the step six is specifically as follows: The improved DeePC algorithm comprises an initialization consistency constraint construction module, a predicted track generation module, an objective function assembly module, a basic constraint matrix construction module, a phase maintenance constraint construction module and an optimization solving module; the initialization consistency constraint construction module constructs linear equality constraint terms in an input Hankel matrix initialization block and an output Hankel matrix initialization block based on an initialization input segment and an initialization output segment, and writes the linear equality constraint terms into a constraint matrix cache region; The prediction track generation module constructs a prediction time domain control input variable and a prediction time domain measurement output variable based on the input Hankel matrix prediction block and the output Hankel matrix prediction block, and writes the prediction time domain control input variable and the prediction time domain measurement output variable into a decision variable buffer area; the objective function assembly module constructs a quadratic objective function matrix according to the deviation weight parameter between the predicted time domain measurement output variable and the reference track and the amplitude weight parameter of the predicted time domain control input variable, and writes the quadratic objective function matrix into the interface buffer area of the solver; The basic constraint matrix construction module constructs an amplitude constraint matrix according to the upper and lower limit parameters of the control input, the upper and lower limit parameters of the change rate of the control input and the upper and lower limit parameters of the measurement output, and forms a first constraint set together with the initialized consistency constraint matrix; The phase retention constraint construction module generates a time index mapping table according to the time delay parameter set, and the time index mapping table records a stride corresponding relation in a prediction time domain for each pair of preset variables; Constructing a double-boundary inequality matrix aiming at each group of cross-time differential vectors, wherein the double-boundary inequality matrix defines that the differential components of the measurement output variables at the corresponding prediction step positions are in a preset numerical interval, and the double-boundary inequality matrix is combined into a first constraint set to form a second constraint set; The optimization solving module calls a PLC integrated machine embedded quadratic programming solver to execute numerical value solving processing on the objective function matrix and the second constraint set, and a prediction time domain control input sequence is generated; and selecting a control input quantity corresponding to the first sampling period from the predicted time domain control input sequence according to time sequence, and writing the control input quantity into an execution buffer area.
  8. 8. The method for controlling the operation of the special PLC all-in-one machine for the boiler based on the digital twin system according to claim 1, wherein the step seven is specifically as follows: The PLC all-in-one machine reads a control input quantity corresponding to the first time index in an execution buffer zone and forms a limiting parameter set according to a preset control input amplitude upper limit, an amplitude lower limit and a change rate upper limit parameter; Performing channel-by-channel comparison of the read control input quantity and the amplitude limiting parameter set, updating to an amplitude upper limit when the control input quantity exceeds a corresponding amplitude upper limit, and updating to an amplitude lower limit when the control input quantity is lower than a corresponding amplitude lower limit; calculating the change rate of the control input quantity after the amplitude update and the control input quantity output in the previous control period, and updating the component with the change rate exceeding the upper limit of the preset change rate as a change rate boundary value; Writing the control input quantity subjected to the constraint processing of the amplitude and the change rate into an execution register area, and sending the control input quantity to a boiler execution mechanism through a field bus; synchronously acquiring the control input data and the measurement output data after execution, and writing the corresponding time stamp, the channel identifier and the period identifier into a return data structure; the backhaul data structure is sent to the digital twinning system via the industrial communication interface.

Description

Digital twinning-based operation control method of special PLC (programmable logic controller) integrated machine for boiler Technical Field The invention relates to the technical field of self-adaptive control systems and machine learning, in particular to a digital twin-based operation control method of a special PLC (programmable logic controller) integrated machine for a boiler. Background With the wide application of large-scale coal-fired boilers, biomass boilers and industrial waste heat boilers in power production and process industries, the safety, economy and environmental protection of the boiler operation process are increasingly in high demands on the performance of a control system. In the existing engineering, a distributed control system or a programmable logic controller is generally adopted to carry out closed-loop adjustment on key variables such as fuel given, air supply quantity, water supply quantity, main steam pressure, main steam temperature, steam drum water level, smoke components and the like, and an empirical rule control method, a cascade control method or a model prediction control method based on a mechanism model is adopted to adapt to load change and working condition disturbance. However, in an actual operation scene, a boiler object generally has the characteristics of strong nonlinearity, large inertia, multi-loop strong coupling, long time delay and the like, and is also influenced by factors such as fuel quality fluctuation, dust accumulation and slagging, soot blowing operation and the like, so that a control mode relying on a fixed mechanism model or a simplified identification model is difficult to maintain prediction precision for a long time, control parameters need to be frequently set, and engineering debugging cost is high. In recent years, with the improvement of the industrial internet and data acquisition capability, a data-driven control method for constructing a prediction model based on historical operation data is paid attention to gradually, part of researches begin to introduce a data-enabled prediction control method into the field of industrial processes, directly solve an optimization problem by constructing a historical data matrix to generate a control sequence, and attempt to combine a digital twin platform to perform operation state mapping and parameter updating. However, the prior proposal focuses on fitting the future input-output relation by using historical data, generally only applies constraint on control variables at the amplitude or change rate level, lacks depiction on the dynamic propagation time sequence relation existing between the internal combustion-heat transfer-steam-water systems of the boiler, and under the working condition of multiple variables and long time delay, the response sequence among different measurement quantities easily drifts in the prediction process, so that the prediction track is inconsistent with the real physical process in time structure, thereby influencing the stability and repeatability of the control effect. In addition, in the prior art, a unified time reference is often adopted to directly splice multi-channel signals when a historical data matrix is constructed, and a structural processing means aiming at inherent propagation delay of each variable is lacked, so that the implicit dynamic rhythms in the historical data cannot be effectively utilized in the process of optimizing and solving. Therefore, how to provide a digital twin-based operation control method for a special PLC (programmable logic controller) integrated machine for a boiler is a problem to be solved by a person skilled in the art. Disclosure of Invention According to the method, the digital twin and the special PLC all-in-one machine for the boiler are used for operating cooperatively, an improved DeePC control algorithm with phase retention constraint is introduced, multi-scale time delay mining and Hankel line index rearrangement are combined, a multi-variable dynamic propagation structure of the boiler is embedded into a predictive control solving process, the predictive track time consistency and control stability are improved under a complex working condition, the phase drift risk is reduced, and the control effect of high safety and engineering applicability for industrial boiler operation is achieved. According to the embodiment of the invention, the operation control method of the special PLC all-in-one machine for the boiler based on digital twinning comprises the following steps: collecting boiler control input data and measurement output data, and executing time alignment processing to form an input sequence and an output sequence; Step two, the digital twin system receives the input sequence and the output sequence and executes cleaning treatment to generate standardized operation data; Thirdly, constructing an input Hankel matrix and an output Hankel matrix based on the standardized operation data, and executing initi