Waste gas treatment equipment control system based on industrial internet

Abstract

The invention relates to a waste gas treatment equipment control system based on industrial internet, which belongs to the technical field of industrial automation control, and comprises a distributed state prediction module used for acquiring a posterior state estimation vector, a current control input vector and a neighborhood influence vector at a previous moment, according to the method, transition from passive response to active prediction is achieved, the inherent contradiction between the control time efficiency and the information integrity is solved, through the cooperation of prediction and self-organization, the real-time performance of the waste gas treatment equipment is improved, and the real-time performance of the waste gas treatment equipment is improved. The response capability of the system to sudden disturbance is improved to a sub-second level from several seconds and even tens of seconds, the challenge of an industrial turbulence field can be dealt with, the system entropy is introduced to serve as a quantitative evaluation standard of overall performance, and a top feedback correction closed loop is constructed, so that the system performance is improved. And the high-speed response characteristic of distributed control and the high-precision advantage of centralized control are ingeniously integrated.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically to a control system for waste gas treatment equipment based on the Industrial Internet. Background Technology

[0002] In the waste gas treatment process of chemical industrial parks, especially in chemical reaction scenarios involving turbulence, key system parameters such as temperature and component concentration exhibit strong nonlinear and spatiotemporal coupling characteristics. Current control systems generally adopt a linear causal architecture of observation-decision-execution. The inherent defect of this architecture is that, in order to ensure the integrity of the information on which the decision is based, the data acquisition and analysis process often requires a long time. However, effectively suppressing instantaneous disturbances in the turbulent field requires the control system to intervene in a very short time. This fundamental contradiction between the long-cycle requirement for information acquisition and the short-term requirement for control intervention leads to the common problems of response lag, control overshoot, and insufficient accuracy when facing sudden disturbances. The core issue is that this architecture cannot make fast and effective non-causal decisions under conditions of incomplete information. Summary of the Invention

[0003] The purpose of this invention is to provide a waste gas treatment equipment control system based on the Industrial Internet to solve the problems mentioned in the background art.

[0004] The technical solution of the present invention includes: a distributed state prediction module, used to obtain the posterior state estimation vector, the current control input vector and the neighborhood influence vector of the previous moment, and to predict the future state of the waste gas treatment equipment using a spatiotemporal coupling prediction model to obtain the prior predicted state vector; The self-organizing control module is used to calculate the attractive force vector and the repulsive force vector in the artificial potential field based on the prior predicted state vector and the preset global target state vector, fuse the two to generate a total virtual force vector, and generate self-organizing control instructions based on the total virtual force vector. The system entropy assessment module is used to collect the actual state values ​​of the waste gas treatment equipment, calculate the state deviation based on the actual state values ​​and the global target state vector, and determine the treatment efficiency entropy. The self-organizing parameter correction module is used to respond to the processing performance entropy, compare the processing performance entropy with a preset entropy threshold, and generate a self-organizing parameter correction signal for adjusting the operating parameters of the self-organizing control module based on the comparison result.

[0005] Preferably, the step of generating the total virtual force vector specifically includes: The attraction vector is determined based on the difference between the global target state vector and the prior predicted state vector, combined with the attraction gain coefficient. Based on the prior predicted state vector and the predicted state vector of the disturbance source, the repulsive force vector is determined; The attraction vector and the repulsion vector are combined to generate the total virtual force vector.

[0006] Preferably, the step of determining the repulsive force vector specifically includes: Calculate the state-space distance between the prior predicted state vector and the state vector of the disturbance source; The state space distance is compared with a preset disturbance impact threshold; When the state space distance is less than or equal to the disturbance influence threshold, the repulsive force vector is calculated based on the state space distance and the repulsive force gain coefficient. When the state space distance is greater than the disturbance influence threshold, the repulsive force vector is determined to be a zero vector.

[0007] Preferably, the step of obtaining the prior predicted state vector specifically includes: The posterior state estimation vector is transformed by the state transition matrix to obtain the time evolution term; The control input vector is transformed by the control input matrix to obtain the control action term; The spatial propagation term is obtained by transforming the neighborhood influence vector using a spatiotemporal coupling weight matrix. The prior predicted state vector is obtained by fusing the time evolution term, the control action term, and the spatial propagation term.

[0008] Preferably, the parameters of the state transition matrix and the control input matrix are derived from computational fluid dynamics simulations of the waste gas treatment equipment; the parameters of the spatiotemporal coupling weight matrix are derived from online learning optimization.

[0009] Preferably, the step of determining the processing efficiency entropy specifically includes: Obtain the actual state values ​​collected by all sensors and compare them with the global target state vector to calculate the state deviation of each sensor. The range of the state deviation is divided into preset deviation intervals, and the number of sensors falling into each deviation interval is counted to determine the probability distribution of the state deviation. Based on the probability distribution of the state deviation, the processing efficiency entropy is calculated using the Shannon entropy formula.

[0010] Preferably, the step of generating the self-organizing parameter correction signal specifically includes: When the processing efficiency entropy is detected to exceed the entropy threshold, a correction mechanism is activated to generate the self-organizing parameter correction signal; When the processing efficiency entropy does not exceed the entropy threshold, the current operating parameters are maintained and the self-organizing parameter correction signal is not generated.

[0011] Preferably, the parameters of the attractive force gain coefficient and the repulsive force gain coefficient are derived from experimental tuning; the value of the disturbance influence threshold is calculated based on the process safety boundary, the system's maximum response capability, and the prediction time margin.

[0012] This invention provides an improved control system for waste gas treatment equipment based on the Industrial Internet, which, compared with existing technologies, has the following improvements and advantages: First, through the synergy of prediction and self-organization, the system's response capability to sudden disturbances is improved from several seconds or even tens of seconds to the sub-second level, which is sufficient to meet the challenges of industrial turbulent fields; Secondly, by introducing system entropy as a quantitative evaluation standard for global performance and constructing a top-level feedback correction closed loop, the high-speed response characteristics of distributed control and the high-precision advantages of centralized control are cleverly integrated. Third, the decentralized control architecture is not sensitive to single points of failure, and adding or removing device nodes does not require reconfiguring the entire system, demonstrating superior robustness and engineering scalability. Attached Figure Description

[0013] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the waste gas treatment equipment control system based on the Industrial Internet of Things of this invention. Figure 2 This is a flowchart of the steps for generating the total virtual force vector of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0015] Example 1 Please see Figure 1 This invention provides a technical solution for a waste gas treatment equipment control system based on the Industrial Internet, including: a distributed state prediction module, used to obtain the posterior state estimation vector, the current control input vector and the neighborhood influence vector of the previous moment, and to predict the future state of the waste gas treatment equipment using a spatiotemporal coupling prediction model to obtain the prior predicted state vector; The self-organizing control module is used to solve the attractive force vector and the repulsive force vector in the artificial potential field based on the prior predicted state vector and the preset global target state vector, fuse the two to generate the total virtual force vector, and generate self-organizing control commands based on the total virtual force vector. The system entropy assessment module is used to collect the actual state values ​​of the waste gas treatment equipment, calculate the state deviation based on the actual state values ​​and the global target state vector, and determine the treatment efficiency entropy. The self-organizing parameter correction module is used to respond to the processing performance entropy by comparing the processing performance entropy with a preset entropy threshold, and generating a self-organizing parameter correction signal based on the comparison result to adjust the operating parameters of the self-organizing control module.

[0016] The core design of this invention lies in constructing a predictive-self-organizing-emergent non-causal control architecture to address the turbulence and nonlinear disturbances prevalent in chemical processes. Current control systems, due to their linear causal model of observation-decision-execution, inherently contradict the long-cycle requirements of information acquisition with the short-term requirements of control intervention, leading to response lag and insufficient control accuracy. The four-module linkage system proposed in this invention, comprising a distributed state prediction module, a self-organizing control module, a system entropy assessment module, and a self-organizing parameter correction module, aims to fundamentally overcome this technical predicament. These four modules are functionally progressive and tightly coupled in data flow, jointly constructing a complete adaptive closed-loop system capable of achieving proactive response and precise control.

[0017] Example 2 The specific steps for obtaining the prior predicted state vector are as follows: The time evolution term is obtained by transforming the posterior state estimation vector using the state transition matrix. The control input vector is transformed by the control input matrix to obtain the control action term; The spatial propagation term is obtained by transforming the neighborhood influence vector using a spatiotemporal coupling weight matrix. By integrating the time evolution term, the control action term, and the spatial propagation term, the prior predicted state vector is obtained.

[0018] The parameters of the state transition matrix and control input matrix are derived from computational fluid dynamics simulations of the waste gas treatment equipment; the parameters of the spatiotemporal coupling weight matrix are derived from online learning optimization.

[0019] In this embodiment, the core function of the distributed state prediction module is to provide high-quality decision-making basis with time lead for subsequent control links. This is achieved through an innovative spatiotemporal coupling prediction model, mathematically expressed as:

[0020] The theoretical basis of this model is a deep extension of the Kalman filter prediction equation in control theory. The technical motivation lies in the profound insight that standard prediction models only consider the temporal evolution of the nodes themselves, while ignoring the physical reality of disturbances propagating along the spatial dimension in industrial turbulent fields. To this end, this invention innovatively introduces a spatiotemporal coupling term to explicitly characterize this spatial correlation effect mathematically, making the prediction results complete in both temporal extrapolation and spatial propagation, thereby achieving accurate prediction of future states. This design changes the decision-making basis from observation of the past to prediction of the future, giving the system a valuable lead time to deal with sudden disturbances. Where is the prior predicted state vector at time k, which is the final output of this module and represents the most likely state of the system in the next control cycle. Its physical dimensions are, for example, temperature (°C) or concentration (ppm); is the prior predicted state vector at time k; is the posterior state estimation vector at time k-1, which is the optimal result of the previous iteration and serves as the benchmark for this prediction; is the known control input vector at time k; is the dimensionless state transition matrix, describing the natural evolution of the system driven by internal physical laws without external intervention. The parameters are obtained by discretizing the exhaust gas treatment equipment after computational fluid dynamics simulation; is the known control input vector at time k, such as heating power or valve opening; is the control input matrix, and the physical dimensions (e.g., °C / W) ensure that the dimensions of the product terms are consistent with the state vector. Consistent, the parameters can also be determined through CFD simulation or system identification methods; is the neighborhood influence vector defined in this invention, representing the deviation between the state of other neighboring nodes around the current node and the target value at time k-1, used to quantify the perturbation information from the spatial dimension; is the dimensionless spatiotemporal coupling weight matrix, which is the core adjustable parameter of this invention, defining the contribution weight of the neighborhood influence to the prediction result of the current node. Its parameters are not fixed values, but are continuously adaptively optimized according to the final control accuracy through an online learning mechanism based on reinforcement learning; the subscript represents the current time step, while represents the previous time step.

[0021] Example 3, as Figure 2 As shown; The specific steps for generating the total virtual force vector are as follows: The attraction vector is determined based on the difference between the global target state vector and the prior predicted state vector, combined with the attraction gain coefficient. The repulsive force vector is determined based on the prior predicted state vector and the predicted state vector of the disturbance source. By merging the attraction vector and the repulsion vector, a total virtual force vector is generated.

[0022] The specific steps for determining the repulsive force vector are as follows: Calculate the state-space distance between the prior predicted state vector and the state vector of the disturbance source; The state space distance is compared with a preset disturbance impact threshold; When the state space distance is less than or equal to the disturbance influence threshold, the repulsive force vector is calculated based on the state space distance and the repulsive force gain coefficient. When the state space distance is greater than the disturbance influence threshold, the repulsive force vector is determined to be the zero vector.

[0023] In this embodiment, the self-organizing control module receives the prior predicted state vector output by the prediction module and transforms it into specific control actions. The core technology is the creative application of the artificial potential field method from the field of robot path planning to industrial process control, cleverly decomposing a complex global optimization problem into a simple local mechanical problem for each control node. Each node acts as a control agent, making autonomous decisions based on local forces in a virtual potential field, achieving sub-second-level anticipatory response. This force is composed of attractive and repulsive forces, and the calculation logic is as follows: Total virtual force: Attraction: Repulsive force: at that time: then: The structure and parameter definitions of this set of formulas are as follows: where is the total virtual force vector experienced by the i-th agent, and its direction and magnitude together determine the final control output. The direction and intensity of adjustment; is the attractive force vector, whose physical dimensions are the same as the state vector. Consistent, its direction points towards the global target state. The drive system converges towards the set value; It is a repulsive force vector, and its physical dimensions are also the same as those of the state vector. Consistent, direction divergence from the predicted first Disturbance source This is used to proactively avoid potential process deviations; The attraction gain coefficient is a dimensionless, adjustable parameter obtained through experimental tuning, used to balance convergence speed and system overshoot; its force direction is determined by the state deviation vector, and its direction is away from the disturbance source. ; This is the repulsive force gain coefficient, whose physical dimension is the cube of the state vector dimension. For example, if the state is temperature °C, then... The dimension of is (°C). 3 The aim is to ensure that the results obtained after calculating the entire repulsive force formula are accurate. Vector and state vector The dimensions are kept consistent; its values ​​are based on experimental tuning; This refers to the prior predicted state vector of the i-th control node at time k, which is the output of the prediction module. ; For intelligent agents Predicted state and disturbance source The distance in the state space has the same dimensions as same; To define the threshold for the impact of disturbances, a boundary is defined at which the repulsive force takes effect. Its value is a dynamic safety margin calculated based on the process safety boundary, the system's maximum response capability, and the prediction time margin, ensuring that the system has sufficient capability to push the state back before it enters the dangerous region. It can be defined as the forecast time margin Inside, the system operates at its maximum response rate. The maximum state deviation that can be eliminated, combined with the safety boundary. Applying constraints can be represented as ; Subscript The index representing the current controlling agent, and This represents the index of potential disturbance sources; this mechanism enables each control node to quickly and autonomously find a balance point that both approaches the target and avoids risks, ultimately resulting in a global self-organizing and collaborative control effect. Acquisition method: It can be defined as the real-time measurement and short-term extrapolation prediction obtained by setting up an independent sensor network at key disturbance locations, such as the air inlet, or the theoretical disturbance state calculated based on upstream operating parameters according to the process mechanism model.

[0024] Example 4 The specific steps for determining the processing efficiency entropy are as follows: Acquire the actual state values ​​collected by all sensors and compare them with the global target state vector to calculate the state deviation of each sensor. The range of state deviation values ​​is divided into preset deviation intervals, the number of sensors falling into each deviation interval is counted, and the probability distribution of state deviation is determined. Based on the probability distribution of state deviation, the processing efficiency entropy is calculated using the Shannon entropy formula.

[0025] The entire range of the state deviation Δx is divided into M non-overlapping deviation intervals B_j (j=1, 2, ...,M); In this embodiment, the technical objective of the system entropy assessment module is to transform macroscopic process indicators, such as the waste gas treatment qualification rate or control accuracy ±5%, into a physical quantity that can be calculated online and quantified. This invention introduces the processing efficiency entropy, derived from the Shannon entropy concept in information theory, and its calculation formula is as follows:

[0026] in, To handle performance entropy, it is a dimensionless scalar. The smaller its value, the more uniform and orderly the overall state of the system, that is, the better the control effect. This represents the total number of sensors in the waste gas treatment equipment. For the first The deviation between the actual state collected by each sensor and the target state; This represents the ratio of the number of sensors whose state deviation falls within the j-th deviation interval B_j to the total number of sensors N; j is the index of the deviation interval, and M is the total number of intervals. In engineering implementation, this probability distribution is approximated by dividing the deviation range into several preset deviation intervals and then counting the frequency of the number of sensors falling within each interval relative to the total number. The index represents the sensor; the essence of this design lies in providing a macroscopic, global perspective to examine the overall effect emerging from the lower-level self-organizing control, providing a solid quantitative foundation for achieving higher-level feedback correction. The division of the deviation interval can adopt the equal-width division method, that is, determine the range based on the maximum and minimum values ​​of the historical data of the state deviation and divide it equally; or adopt the equal-frequency division method to ensure that the number of samples falling into each interval is approximately equal.

[0027] Example 5 The specific steps for generating the self-organizing parameter correction signal are as follows: When the processing efficiency entropy is detected to exceed the entropy threshold, a correction mechanism is activated to generate a self-organizing parameter correction signal. When the processing efficiency entropy does not exceed the entropy threshold, the current operating parameters are maintained and no self-organizing parameter correction signal is generated.

[0028] The parameters of the attractive force gain coefficient and the repulsive force gain coefficient are derived from experimental tuning; the value of the disturbance influence threshold is calculated based on the process safety boundary, the maximum response capability of the system, and the prediction time margin.

[0029] In this embodiment, the self-organizing parameter correction module constitutes the highest-level feedback correction closed loop of the present invention; the core operating logic is to calculate the processing efficiency entropy in real time. With a preset entropy threshold Perform continuous comparisons; the entropy threshold here. The setting logic is crucial; it is not set arbitrarily, but determined through in-depth statistical analysis of historical operating data. This threshold ensures that when... At that time, the deviation of more than 99.7% of the measuring points in the equipment could be stably within the allowable range of ±5% of the process requirements, thus ensuring... It becomes a direct representation of the final process parameters within the control system; This threshold is the processing efficiency entropy calculated based on a large amount of historical data under stable operating conditions. The statistical distribution can be calculated by taking its mean plus three standard deviations. The value of This corresponds to a statistically significant confidence interval of approximately 99.7%, ensuring a strong correlation between this threshold and process indicators. The module's operating mechanism has clear conditional logic: once the processing performance entropy is detected... There are more than the entropy threshold If the trend is as described, the correction module will generate a self-organizing parameter correction signal; conversely, if... If the value remains below the threshold, no signal is generated, and the system maintains its current operating parameters. This correction signal does not directly interfere with any specific valve or heater; instead, it is returned to the self-organizing control module to affect its core operating rules, namely the attraction gain coefficient. and repulsive force gain coefficient Parameters are fine-tuned; this indirect intervention method of adjusting rules rather than adjusting behavior can guide the system's self-organizing behavior back to a high-precision and high-stability operating track, while preserving the self-organizing characteristics of the underlying control to the greatest extent. It is the existence of this top-level correction loop that ultimately ensures that the system can achieve high-precision stable control of ±5% while achieving sub-second high-speed response, thus achieving a unity of speed and precision. The specific adjustment logic for this correction signal can employ a proportional-integral control strategy based on error; the entropy deviation... As input, the gain coefficient is calculated by the PI controller. and Adjustment amount and For example, when the entropy value continues to exceed the limit, the attraction gain can be gradually reduced. To reduce the system's convergence speed and avoid overshoot, while increasing the repulsive force gain. To enhance the system's ability to avoid disturbances; the adjustment formula can be exemplified as follows:

[0030] in These are the proportional and integral coefficients obtained through experimental tuning; to ensure dimensional consistency, the dimensions of each coefficient should be: proportional coefficient It is dimensionless; integral coefficient The dimension of is one part of time. proportionality coefficient Dimensions and The dimensions are consistent, and the dimensions of the state vector are cubed. Integral coefficient The dimension of the state vector is the cube of the dimension of the state vector divided by time. .

[0031] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. Industrial internet-based exhaust gas treatment equipment control system, characterized by, The method comprises the following steps: a distributed state prediction module is used to obtain a posterior state estimation vector of a previous time, a current control input vector and a neighborhood influence vector, and a spatiotemporal coupling prediction model is used to predict a future state of the exhaust treatment equipment to obtain a prior predictive state vector; a self-organizing control module is used to calculate an attractive force vector and a repulsive force vector in a potential field based on the prior predictive state vector and a preset global target state vector, to fuse the two to generate a total virtual force vector, and to generate a self-organizing control instruction based on the total virtual force vector; a system entropy evaluation module is used to collect actual state values of the exhaust treatment equipment, to calculate a state deviation based on the actual state values and the global target state vector, and to determine a processing efficiency entropy; a self-organizing parameter correction module is used to compare the processing efficiency entropy with a preset entropy threshold value in response to the processing efficiency entropy, and to generate a self-organizing parameter correction signal for adjusting operating parameters of the self-organizing control module according to a comparison result.

2. The industrial internet-based exhaust treatment device control system of claim 1, wherein, The generation step of the total virtual force vector is specifically as follows: the attractive force vector is determined based on a difference between the global target state vector and the prior predictive state vector and in combination with an attractive force gain coefficient; the repulsive force vector is determined based on the prior predictive state vector and a state vector of a predicted disturbance source; the total virtual force vector is generated by fusing the attractive force vector and the repulsive force vector.

3. The industrial internet-based exhaust treatment device control system of claim 2, wherein, The determination step of the repulsive force vector is specifically as follows: a state space distance between the prior predictive state vector and the state vector of the disturbance source is calculated; the state space distance is compared with a preset disturbance influence threshold value; when the state space distance is less than or equal to the disturbance influence threshold value, the repulsive force vector is calculated according to the state space distance and a repulsive force gain coefficient; when the state space distance is greater than the disturbance influence threshold value, the repulsive force vector is determined as a zero vector.

4. The industrial internet-based exhaust treatment device control system of claim 1, wherein, The acquisition step of the prior predictive state vector is specifically as follows: a time evolution term is obtained by transforming the posterior state estimation vector through a state transition matrix; a control action term is obtained by transforming the control input vector through a control input matrix; a space propagation term is obtained by transforming the neighborhood influence vector through a spatiotemporal coupling weight matrix; the prior predictive state vector is obtained by fusing the time evolution term, the control action term and the space propagation term.

5. The industrial internet-based exhaust treatment device control system of claim 4, wherein, Parameters of the state transition matrix and the control input matrix are derived from computational fluid dynamics simulation of the exhaust treatment equipment; parameters of the spatiotemporal coupling weight matrix are derived from online learning optimization.

6. The industrial internet-based exhaust treatment device control system of claim 1, wherein, The determination step of the processing efficiency entropy is specifically as follows: actual state values collected by all sensors are compared with the global target state vector to calculate state deviations of the sensors; a value range of the state deviations is divided into preset deviation intervals, a number of sensors falling in each deviation interval is counted, and a probability distribution of the state deviations is determined; the processing efficiency entropy is calculated based on the probability distribution of the state deviations by using a Shannon entropy formula.

7. The industrial internet-based exhaust treatment device control system of claim 1, wherein, The generation step of the self-organizing parameter correction signal is specifically as follows: When the processing performance entropy is monitored to exceed the entropy threshold value, a correction mechanism is started to generate the self-organizing parameter correction signal; When the processing performance entropy does not exceed the entropy threshold value, the current operating parameters are maintained, and the self-organizing parameter correction signal is not generated.

8. The industrial internet-based exhaust treatment device control system of claim 3, wherein, The parameters of the attraction gain coefficient and the repulsion gain coefficient are derived from experimental setting; and the value of the disturbance influence threshold is calculated according to the process safety boundary, the maximum response capability of the system and the prediction time margin.

Images