An industrial gas cooling optimization control system, storage medium and device

Abstract

An industrial gas cooling optimization control system, a storage medium and equipment belong to the technical field of industrial automation control. In order to solve the problem that the control precision of the existing industrial gas cooling system (especially the final cooling tower and other equipment) needs to be improved, the system comprises a data acquisition and preprocessing module for acquiring system running state data, and a model predictive control module for determining the optimal control amount of the cooling water mass flow based on the system running state data and the ARMAX dynamic model. When predicting, the ARMAX dynamic model is used to obtain the gas outlet temperature at time k based on the historical data matrix and the future input matrix as input, and the control increment of the cooling water mass flow at time k is determined based on the input and output relationship corresponding to the prediction process of the ARMAX dynamic model, and then the control amount of the cooling water mass flow at time k is obtained by combining the control amount of the cooling water mass flow at time k.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of industrial automation control, and particularly relates to an intelligent control system for a heat exchange process, a storage medium and equipment. BACKGROUND

[0002] In modern coal chemical industry, steel smelting and other industrial processes, the cooling of high-temperature gas (such as coal gas, coke oven gas) is a key process. Taking a typical final cooling tower as an example, its main function is to use circulating cooling water to directly or indirectly exchange heat with high-temperature gas, so as to reduce the gas temperature from a relatively high level (such as 45-90℃) to a stable low temperature range (such as 25-35℃) required by the downstream process. The control accuracy and stability of the gas outlet temperature directly affect the quality of the subsequent products, production safety, equipment operation life and energy consumption level of the entire system.

[0003] At present, the methods used in industry to control such gas cooling processes mainly include manual control, traditional PID control and mechanism model-based control.

[0004] Manual control mainly relies on the experience of the operator to manually adjust the frequency of the cooling water pump or the opening of the control valve to change the cooling water flow, thereby indirectly controlling the gas outlet temperature. This method is highly dependent on the proficiency and responsibility of the operator, and has a lagging response and rough adjustment, making it difficult to cope with frequent changes in working conditions (such as fluctuations in coal gas flow, composition and inlet temperature), often resulting in large fluctuations in outlet temperature, causing energy waste or production indicators to be unqualified.

[0005] Traditional PID control is the most widely used automatic control strategy in the current industrial field. The actual temperature of the gas outlet is measured by a sensor, and compared with the preset target value to obtain a deviation. Then the controller calculates a control output signal according to the proportional (P), integral (I) and derivative (D) components of the deviation to drive the valve or water pump and other actuators. However, PID is a "post-feedback" control that only starts to adjust after the deviation has occurred. For heat exchange systems with large inertia and pure lag characteristics, the control effect is often poor, and overshoot and oscillation are prone to occur. In addition, in the production process, the flow and inlet temperature of the coal gas are the main and frequent disturbance sources. The PID controller is inherently unable to predict the impact of these disturbances, and can only passively compensate, resulting in significant peak and valley fluctuations in the outlet temperature when the disturbance occurs. Finally, the performance of the PID controller is highly dependent on the tuning of its three parameters (Kp, Ki, Kd). The gas cooling process is a typical nonlinear and time-varying system, and a fixed set of PID parameters is difficult to achieve optimal control effect in all working conditions, and online optimization and adaptive adjustment of parameters is very difficult.

[0006] To overcome the limitations of traditional PID control, some advanced control strategies attempt to establish a mathematical model of the controlled object. One method is to establish a mechanism model describing the heat exchange process based on basic physical laws such as thermodynamics and fluid mechanics. For example, by the law of conservation of heat, the heat release on the gas side is estimated according to the temperature rise and flow of cooling water, so as to predict the outlet temperature. However, the internal heat transfer process of real industrial equipment is extremely complex, affected by many uncertain factors such as fouling, uneven fluid distribution, and changes in physical properties. Mechanism models often require a large number of simplifications and idealized assumptions, resulting in low model accuracy and an inability to accurately reflect the actual dynamic characteristics. Control based on such a model will inevitably have large deviations. In addition, once the equipment ages or the working conditions migrate to a large extent, these mechanism models based on fixed parameters will fail and require complex re-identification and calibration by professional technicians.

[0007] In summary, existing industrial gas cooling control technologies, whether relying on human experience, classic PID feedback control, or based on simplified mechanism models, are difficult to achieve precise, smooth, and rapid control of the gas outlet temperature under strong interference, large lag, and nonlinear complex working conditions. Therefore, it is urgent to develop an advanced control system and method that can accurately describe the dynamic characteristics of the cooling system and has forward-looking prediction capabilities to solve the above-mentioned deficiencies in existing technologies. SUMMARY

[0008] The present application aims to solve the problem of improving the control accuracy of existing industrial gas cooling systems, especially the final cooling tower and other equipment.

[0009] An industrial gas cooling optimization control system, comprising:

[0010] A data acquisition and preprocessing module for acquiring system operating state data, including cooling water mass flow, gas outlet temperature, and other state data;

[0011] A model predictive control module for determining the optimal control amount of cooling water mass flow based on system operating state data and an ARMAX dynamic model; the model predictive control module includes a data construction unit and a prediction unit;

[0012] The data construction unit is used to construct a historical data matrix corresponding to the system operating state data and a future input matrix FuturesInputs;

[0013] Prediction unit: based on the historical data matrix and the future input matrix FuturesInputs as input, the ARMAX dynamic model is used to obtain the gas outlet temperature at time k, and based on the input and output relationship corresponding to the prediction process of the ARMAX dynamic model, the control increment of the cooling water mass flow at time k is determined, and then combined with the control amount of the cooling water mass flow at time k, that is, the optimal control amount. the control amount of the cooling water mass flow at time k, that is, the optimal control amount.

[0014] Further, the other state data includes cooling water inlet temperature, gas inlet temperature, and gas mass flow.

[0015] Further, the process of constructing the historical data matrix corresponding to the system running state data includes:

[0016] For the system running state data, the historical data of length S before the current time k is constructed into a matrix, wherein the cooling water mass flow data is a column in the historical data matrix as input prediction data of the ARMAX dynamic model; the gas outlet temperature is a column in the historical data matrix as output prediction data of the ARMAX dynamic model.

[0017] Further, the process of constructing the future input matrix FuturesInputs corresponding to the system running state data includes:

[0018] For the current time k, the other state data at time k is taken as data in the prediction time domain P, that is, the other state data at time k is taken as data in the future P times; the cooling water mass flow in the prediction time domain P including time k is determined according to the output of the prediction unit; in the process of predicting the cooling water mass flow by the prediction unit, the ARMAX dynamic model takes the actual cooling water mass flow at time k-1 as input prediction data, and obtains the gas outlet temperature in the prediction time domain P including time k.

[0019] Further, the process of determining the control increment of the cooling water mass flow at time k based on the input and output relationship corresponding to the prediction process of the ARMAX dynamic model includes:

[0020] A decision variable vector U of length P is defined, that is, the control increment of the future P steps:

[0021]

[0022] wherein, represents the control increment of the future P steps for time k;

[0023] According to the decision variable vector U and the control amount The complete control amount trajectory in the future P steps is calculated by accumulation and computation, i.e.

[0024]

[0025] wherein, represents the control amount at the time point k; is a variable from 0 to is a variable from 0 to

[0026] The calculated control amount trajectory is filled into the column of the FutureInputs matrix;

[0027] The historical data matrix, the prediction time domain P and the FutureInputs are taken as inputs, and the ARMAX model is called as a prediction function. According to the dynamic equation of the model, the prediction function performs iterative calculation for P steps to obtain the output prediction value sequence of the system in the future P steps:

[0028]

[0029] wherein, represents the prediction of the gas outlet temperature at the time point k+i at the time point k; is the output prediction value sequence of the system in the future P steps;

[0030] The total cost is obtained based on the difference between the predicted value of the gas outlet temperature and the set gas outlet temperature and the sum of the complete control increments in P steps, and the optimal control increment sequence is obtained by minimizing the total cost , i.e. the optimal control increment U;

[0031] Further, the actual control amount that should be executed at the current time point k is determined:

[0032]

[0033] wherein represents a new cooling water mass flow set value, is the increment at the time point k in

[0034] Further, the total cost ; wherein:

[0035] is the tracking cost:

[0036]

[0037] wherein, represents the set gas outlet temperature,​​​​ is the weight of the control cost;

[0038] is the control cost: used to penalize the drastic change of the control variable; it is calculated as:

[0039]

[0040] where, is the weight of the control cost; is the square of the control increment at time t.

[0041] Further, the ARMAX dynamic model is pre-trained, and the pre-training process of the ARMAX dynamic model includes:

[0042] The mathematical expression of the ARMAX model is as follows:

[0043]

[0044] where y(t) is the system output at time t, i.e., the gas outlet temperature; u(t) is the system input vector at time t; e(t) is a white noise sequence, representing unmeasurable disturbance; is the backshift operator, i.e., ; nk is the pure time delay between the system input and output; A(q), B(q), C(q) are polynomials with respect to the backshift operator , whose coefficients are parameters that need to be determined by data identification:

[0045]

[0046]

[0047]

[0048] where, , , are polynomial coefficients; , , are structure parameters of the set ARMAX model, i.e., the order of the polynomial;

[0049] ​Input training data, the training data including cooling water mass flow, gas outlet temperature and other state data, adjust all unknown coefficients in the polynomials A(q), B(q), C(q) during the training process, so that the error between the predicted output of the model and the real historical output when given the historical input is minimized; after the training is completed, an ARMAX model containing all identified coefficients is obtained.

[0050] Further, the structure parameters of the ARMAX model is 50, , The dimension of is other state data plus 1, is 1; pure time delay , The dimension of is other state data plus 1.

[0051] A computer storage medium, the storage medium stores at least one instruction, the at least one instruction is loaded and executed by the processor to realize the industrial gas cooling optimization control system.

[0052] An industrial gas cooling optimization control device, the device comprises a processor and a memory, the memory stores at least one instruction, the at least one instruction is loaded and executed by the processor to realize the industrial gas cooling optimization control system.

[0053] Beneficial effects:

[0054] The industrial gas cooling optimization control system provided by the application can realize accurate, stable and rapid control of the gas outlet temperature under complex working conditions with strong interference, large lag and nonlinearity by combining off-line data-driven modeling with online rolling optimization. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 It is a low-temperature water inlet temperature data signal diagram collected;

[0056] Figure 2 It is a low-temperature water flow data signal diagram collected;

[0057] Figure 3 It is a coal gas flow data signal diagram collected;

[0058] Figure 4 It is a coal gas inlet temperature data signal diagram collected;

[0059] Figure 5 It is a coal gas outlet temperature data signal diagram collected;

[0060] Figure 6 It is an ARMAX prediction model schematic diagram taking coal gas final cooling as an example;

[0061] Figure 7 ARMAX prediction effect display;

[0062] Figure 8 Optimization control module schematic diagram. DETAILED DESCRIPTION

[0063] The technical solution of the present application includes the following core modules:

[0064] 1. Data acquisition and preprocessing module, responsible for collecting and integrating multi-dimensional time series data from the control system or historical database of the industrial gas cooling system. These data should at least include: cooling water inlet temperature, gas inlet temperature, gas mass flow, cooling water mass flow as system input, and gas outlet temperature as system output.

[0065] 2. Data-driven dynamic modeling module, receiving the preprocessed historical data set to build an accurate dynamic mathematical model for predicting the gas outlet temperature.

[0066] 3. Model predictive control (MPC) module, which is an online core control module that makes optimization decisions in a rolling and limited time domain.

[0067] 4. Control execution module, receiving the optimal control instruction (i.e. new cooling water flow set value) calculated by the MPC module. The instruction is issued to the underlying physical execution mechanism, such as a regulating valve or a variable frequency water pump, to accurately adjust the actual flow of cooling water, thereby completing the closed-loop control of the entire cooling system.

[0068] The specific embodiments will be described in detail below. Specific embodiment one:

[0070] The industrial gas cooling optimization control system integrating dynamic modeling and predictive control described in this embodiment includes:

[0071] Data acquisition and preprocessing module: through a pre-set interface, connect with the distributed control system or historical database of the industrial gas cooling system to obtain the original time series data reflecting the system running state. In this embodiment, the data is provided in the form of CSV format file, respectively from different measurement points and sensors. The collected original data set includes at least the following key variables: cooling water inlet temperature data (records the cooling water temperature entering the final cooling tower), gas inlet flow data (records the standard volume flow of coal gas entering the final cooling tower), final cooling tower inlet and outlet pressure and temperature data (including gas inlet temperature, gas outlet temperature, gas inlet pressure and gas outlet pressure), cooling water flow data (records the volume flow of circulating cooling water).

[0072] Next, the system performs preliminary data filtering according to the effectiveness of the actual production conditions. In this embodiment, the system selects one or more stable running intervals for data acquisition and filtering. At the same time, in order to unify the time granularity of the data and reduce data redundancy, the system performs equal-interval down-sampling processing on the filtered data segments. For example, take one sample every 4 data points, thereby converting the original high-frequency data into low-frequency data with a unified sampling time.

[0073] In order to make the subsequent gas outlet temperature prediction model based on the autoregressive moving average model with external input (ARMAX, Auto-Regressive Moving Average with eXogenous inputs) have a clear physical meaning, the system performs unit conversion on part of the variables. In the original data, the gas flow and water flow are both volume flow (m³ / h), which need to be converted to mass flow (kg / s) to be more accurate for heat calculation. For the calculation of gas mass flow ( ): multiply the read gas volume flow data by the preset gas standard density ( , for example, 0.49 kg / Nm³), and then divide by 3600 to complete the conversion from Nm³ / h to kg / s. For the calculation of cooling water mass flow ( ), the same applies, multiply the read cooling water volume flow data by the density of water ( , for example, 1000 kg / m³), and then divide by 3600 to complete the conversion from m³ / h to kg / s.

[0074] Due to the possibility of sensor failure or communication interruption, there may be invalid or missing values (represented as NaN in the program) in the processed data sequence. In order to ensure the continuity of the data in time, the system uses a forward filling method to fill in the gaps of these data. The specific implementation is as follows: the system first locates the positions of all NaN values in the data matrix, and then replaces each NaN value with the previous valid observation value in the time series. This method is suitable for the characteristics of industrial data inertia, and can simply and effectively repair the breakpoints of the data chain.

[0075] At this point, the data acquisition and preprocessing module provides an ideal data basis for the subsequent construction of a high-precision and robust ARMAX dynamic prediction model. For demonstration, the present invention takes the gas final cooling process as an example to demonstrate this method, through the above data processing process, the collected data is shown as Figures 1-5 , Figures 1-5 The horizontal axis time step unit is step, i.e. the number of time steps, and Figure 1 , for example, the horizontal axis 0.5, 1, 1.5, combined with the 10 5, 0.5, 1, 1.5 of the horizontal axis actually represent the 0.5x10 5 , 1, 1.5 of the horizontal axis actually represent the 0.5x10 5 , 1, 1.5 of the horizontal axis actually represent the 0.5x10 5 , 1, 1.5 of the horizontal axis actually represent the 0.5x10

[0076] Data-driven dynamic modeling module: Next, the system utilizes the historical operation data collected from the actual production process, through a series of data processing and system identification steps, to train and generate an ARMAX model that can accurately describe the dynamic behavior of the system. This model can accurately predict the changes of the key output variable (gas outlet temperature) in the future time step according to a set of known input variables (such as the temperature and flow of cooling water / gas), as shown in Figure 6 .

[0077] Specifically, in this embodiment, the input vector of the model contains four dimensions: u1 (cooling water inlet temperature), u2 (gas inlet temperature), u3 (gas mass flow), and u4 (cooling water mass flow). The output one-dimensional variable is the gas outlet temperature. The above input / output data and sampling time are packaged into a standard system identification data object, that is, u1-u4 and the gas outlet temperature are all input to the model.

[0078] In this invention, the module specifically uses ARMAX (autoregressive moving average model with external input) to build the prediction model. It is particularly suitable for describing the industrial gas cooling process involved in this invention because its structure can model the three key dynamic characteristics of the system separately:

[0079] (1) Inertia - modeled by the autoregressive (AR) part. That is, the current gas outlet temperature y(t) is related not only to the current input, but also to the outlet temperatures at historical times y(t-1), y(t-2),... This accurately reflects the thermal inertia characteristics of large industrial equipment such as the final cooling tower.

[0080] (2) eXogenous Inputs - modeled by the external input (X) part: the running state of the system is directly driven by external variables, including the cooling water mass flow as a control variable, and the gas inlet temperature, gas mass flow, and cooling water inlet temperature as disturbance variables. The external input (X) part in the ARMAX model is specifically used to quantify the direct impact of these variables on the system output.

[0081] (3) Stochastic Noise - Modeled by the Moving Average (MA) part: Any real industrial process is accompanied by unmeasurable stochastic noise, such as measurement errors, environmental fluctuations, etc. These disturbances will affect the prediction accuracy. The Moving Average (MA) part in the ARMAX model can effectively capture and compensate for the dynamic characteristics of these random disturbances by modeling the historical prediction errors, thereby improving the robustness of the model.

[0082] The general mathematical expression of the ARMAX model is as follows:

[0083]

[0084] where: y(t) is the system output (gas outlet temperature) at time t. u(t) is the system input vector at time t. e(t) is a white noise sequence representing unmeasurable disturbances. is the backshift operator, i.e. nk is the pure time delay between system input and output. A(q), B(q), C(q) are polynomials in terms of the backshift operator whose coefficients are parameters that need to be determined by data identification:

[0085] (AR part, na is the autoregressive order)

[0086] (X part, nb is the external input order)

[0087] (MA part, nc is the moving average order)

[0088] In this invention, the structural parameters of the ARMAX model, i.e. the order of the polynomials, are set according to prior knowledge, research and multiple experiments on the cooling system process. These orders determine the complexity and dynamic description ability of the model. defines how many past time points the current outlet temperature is affected by its own temperature value. A higher value means that the system has stronger inertia. In this embodiment, is set to 50. defines the depth of influence of each input variable on the output. For example, indicates that the model considers the influence of each input value at the past 50 time points on the current output. defines the order of the noise model. In this embodiment, is set to 1. defines the pure delay time between the change of each input variable and its influence on the output. The change in each input will start to affect the output at the next sample time.

[0089] Next, the armax function in the Matlab System Identification Toolbox is called, and the training data, including the collected gas inlet temperature, gas outlet temperature, gas mass flow, cooling water inlet temperature, and cooling water mass flow, is input. The data is a large amount of data, and the function automatically reconfigures the data to meet the input format and set parameters (na, nb, nc, nk) of the ARMAX model and trains the ARMAX model. The algorithm automatically adjusts all unknown coefficients in the polynomials A(q), B(q), and C(q) to minimize the error (usually the sum of squares) between the predicted output and the actual historical output when the model is given the historical input. After training, the obtained ARMAX model object containing all identified coefficients, as well as its training date, sampling time, input and output variable names, and used order, is packaged into a structured data package.

[0090] Through the above steps, the complex and non-explicit dynamic characteristics of the industrial process contained in the historical data are successfully converted into an accurate, reliable, and computer-directly-callable mathematical model. This model is the premise and foundation for subsequent implementation of model predictive control (MPC) and provides a solid guarantee for its "predicting the future" and "optimizing decision-making" capabilities.

[0091] By separating the collected data into a data fitting set and a test set, and using the above method, after completing the ARMAX parameter identification in the fitting set, the prediction effect is shown on the test set, as shown in Figure 7 . Among them, the colored line segment is the ARMAX model prediction trend (displayed as a straight line due to post-processing), and the blue curve is the real gas outlet temperature observed by the sensor. In order to show the clarity, the prediction line segment and the true value temperature curve are manually offset. It can be seen that the change trend (line segment slope) of the ARMAX model predicted temperature is basically consistent with the true value.

[0092] Model predictive control (MPC) module: the core of the present application is an online running model predictive control (MPC) module, as shown in Figure 8 . This module replaces the traditional, reactive control logic and adopts a forward-looking, optimized control strategy based on model prediction. At each control time, it dynamically calculates the optimal control amount (i.e., cooling water mass flow) through a "prediction-correction-implementation" rolling cycle to achieve precise control of the gas outlet temperature.

[0093] At the system startup or entering the online control phase, the input received by the MPC module contains:

[0094] 1. Pre-trained ARMAX dynamic model: a mathematical model that can accurately predict the future dynamic behavior of the system, trained offline by the aforementioned "data-driven dynamic modeling module" and solidified.

[0095] 2. Historical data object: a data structure containing the historical input and output data of the system with a limited length S before the current time k. In this embodiment, S = 50, i.e., the historical data in this embodiment uses a 50 x 5 data structure, where 5 corresponds to u1-u4 and the gas outlet temperature, which is used to provide the necessary initial state information for the prediction function.

[0096] 3. Control and prediction parameters: hyperparameters set by the user or process engineer, mainly including prediction horizon (defines the number of time steps that the controller predicts the system behavior forward), control horizon (defines the length of the future control action sequence to be optimized), gas outlet temperature target set value, output tracking error weight and control amount change rate weight .

[0097] The MPC module executes the following steps in a fixed sampling cycle:

[0098] S301, in order to predict the future gas outlet temperature, the prediction function of the MPC is implemented by calling the forecast function of the System Identification Toolbox toolkit in Matlab, which nests the calculation method of the ARMAX model. In addition to historical data and set parameters, it also needs to construct a hypothetical future input matrix FuturesInputs, which has a dimension of P x 4 (P time steps, 4 input variables u1-u4), and later the gas outlet temperature will be added to realize multi-step prediction.

[0099] The construction process of the future input matrix FuturesInputs is as follows:

[0100] Let the current time be k, the MPC module obtains the latest system measurement values from the data acquisition module, including: cooling water inlet temperature u1(k), gas inlet temperature u2(k), gas mass flow rate u3(k), and the actual cooling water mass flow rate u4(k-1) at the previous time. These data are used to update the historical data object, i.e., as the time steps advance, the length of the historical data S will become shorter, and based on P x 4 corresponding to u1-u4 and the ARMAX output gas outlet temperature, the historical data object is updated, so that the historical data object maintains a limited length S.

[0101] For measurable external inputs (u1, u2, u3), the MPC module adopts a common simplification strategy: assume they will remain at their current latest measured values for the prediction horizon P into the future. That is:

[0102]

[0103]

[0104]

[0105] - The three-column element of matrix FuturesInputs represents.

[0106] For the control variable (cooling water mass flow u4), whose future trajectory is controllable, is the core of this optimization. The MPC module defines a decision variable vector U of length P, which is the control increment for the future P steps:

[0107]

[0108] where, represents the control increment for the future P steps at time .

[0109] The MPC module calculates the complete control amount trajectory within the future P steps according to the decision variable vector U and the control amount at time k-1, that is:

[0110]

[0111] where, represents the control amount at time ; is a variable from 0 to , is a variable from 0 to ;

[0112] The calculated control amount trajectory is filled into the fourth column of the FutureInputs matrix.

[0113] It should be noted that for time k, the above formula actually sets as the initial state of optimization.

[0114] S302, the MPC module calls the pre-trained ARMAX model as a prediction function with the historical data object (i.e. the past time's gas outlet temperature, gas inlet temperature, gas mass flow, cooling water inlet temperature and cooling water mass flow data), prediction horizon P and future input matrix FutureInputs of the current time as inputs, and the prediction function performs P-step iterative calculation according to the dynamic equation of the model to obtain the output prediction value sequence of the system in the future P steps:

[0115]

[0116] wherein, represents the prediction of the gas outlet temperature at k+i time at k time; is the output prediction value sequence of the system in the future P steps.

[0117] S303, this step is the core decision-making link of the MPC module, and the target of the MPC module is to find an optimal control increment sequence so that a pre-defined total cost function J reaches the minimum.

[0118] The total cost function J is designed as a multi-objective weighted sum in the following form:

[0119]

[0120] wherein,

[0121] (tracking cost): used for punishing the deviation between the predicted output and the target set value. The calculation formula is:

[0122]

[0123] wherein, represents the set gas outlet temperature (i.e. the target value of the controlled gas outlet temperature), is the weight of the tracking cost.

[0124] (control cost): used for punishing the sharp change of the control amount. The calculation formula is:

[0125]

[0126] wherein, is the weight of the control cost; is the square of , is the control increment at k time.

[0127] ​​The MPC module solves the above minimization problem by calling a numerical optimization solver (e.g. a quadratic programming, QP, solver) to obtain the optimal control increment sequence Uopt , i.e. the optimal control increment Uopt.

[0128] S304、According to the principle of the rolling horizon of model predictive control, although the optimal control sequence of the entire future P steps is calculated , the module only adopts and executes the first element of the sequence, i.e. to obtain the cooling water mass flow set value. The MPC module calculates the actual control amount that should be executed at the current time k:

[0129]

[0130] wherein represents the new cooling water mass flow set value, which is sent to the "control execution module" to be completed by the underlying execution mechanism.

[0131] When the next control time k+1 arrives, the system will repeat the entire process of steps S301 to S304, using new measurement information to perform a complete prediction and optimization again. This continuous "feedback correction-rolling optimization" mechanism enables the system to effectively cope with various disturbances, and exhibits excellent robustness and control performance. Specific implementation method two:

[0133] The embodiment is a computer storage medium, and the storage medium stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the industrial gas cooling optimization control system.

[0134] It should be understood that the instructions include a computer program product, software or computerized method corresponding to any method described in the present application; the instructions can be used to program a computer system or other electronic device. The computer storage medium can include a readable medium having instructions stored thereon, and can include but is not limited to a magnetic storage medium, an optical storage medium, a magneto-optical storage medium, a read-only memory (ROM), a random access memory (RAM), an erasable programmable memory (such as an EPROM and an EEPROM), and a flash memory layer, or other types of media suitable for storing electronic instructions. Specific implementation method three:

[0136] The embodiment is an industrial gas cooling optimization control device, the device comprises a processor and a memory, it should be understood that any device comprising the processor and the memory described in the application comprises other units, modules, etc. which display, interact, process, control, etc. through signals or instructions and other functions, and the device can also comprise other units, modules, etc. which display, interact, process, control, etc. through signals or instructions and other functions;

[0137] The memory stores at least one instruction, which is loaded and executed by the processor to implement the industrial gas cooling optimization control system.

[0138] Those skilled in the art should understand that the stored at least one instruction is a computer program product corresponding to the method or system. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes. The solutions in the embodiments of the present application can be implemented in various computer languages, such as object-oriented programming languages Java and interpreted scripting languages JavaScript, etc.

[0139] The present application is described with reference to flowcharts and / or block diagrams of the methods, systems and computer program products according to the embodiments of the present application, and can also be used for corresponding devices. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be realized by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for implementing the functions specified in one or more flows and / or blocks. Figure 1 The device for implementing the functions specified in one or more flows and / or blocks.

[0140] These computer program instructions can also be stored in a computer readable storage medium capable of guiding the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer readable storage medium produce a manufactured product comprising instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for implementing the functions specified in one or more flows and / or blocks. Figure 1 The device for implementing the functions specified in one or more flows and / or blocks.

[0141] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0142] While the preferred embodiments of the application have been described, additional variations and modifications can be employed by those skilled in the art without departing from the spirit and scope of the application. Therefore, it should be understood that the appended claims are intended to cover all such modifications and variations as falling within the scope of the application. Accordingly, the application is intended to embrace all such alterations, modifications, and variations that fall within the scope of the application. What is claimed is:

[0143] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. An industrial gas cooling optimization control system, characterized in that, include: Data acquisition and preprocessing module: used to acquire system operating status data, including cooling water mass flow rate, gas outlet temperature and other status data, including cooling water inlet temperature, gas inlet temperature and gas mass flow rate; Model predictive control module: Determines the optimal control quantity for cooling water mass flow rate based on system operating status data and ARMAX dynamic model; the model predictive control module includes a data construction unit and a prediction unit; Data construction unit: used to construct the historical data matrix corresponding to the system's operating status data, as well as the future input matrix FuturesInputs; The process of constructing a historical data matrix corresponding to system operating status data includes: For the system operating status data, a matrix of length S is constructed from the historical data up to the current time k. The cooling water mass flow rate data is one column in the historical data matrix and serves as the input prediction data for the ARMAX dynamic model. Each type of status data in the other status data is one column in the historical data matrix and serves as the input prediction data for the ARMAX dynamic model. The gas outlet temperature is one column in the historical data matrix and serves as the output prediction data for the ARMAX dynamic model. Prediction Unit: Based on historical data matrices and future input matrices (FuturesInputs), the gas outlet temperature at time k is obtained using an ARMAX dynamic model. Simultaneously, based on the input-output relationship corresponding to the prediction process of the ARMAX dynamic model, the control increment of the cooling water mass flow rate at time k is determined, and then combined with... The control quantity of cooling water mass flow rate at all times is obtained The control quantity of cooling water mass flow rate at any given time is the optimal control quantity.

2. The industrial gas cooling optimization control system according to claim 1, characterized in that, The process of constructing the FuturesInputs matrix corresponding to the system's operating status data includes: For the current time k, the other state data at time k are used as the data in the prediction time domain P, that is, the other state data at time k are used as the data in the future P times. The cooling water mass flow rate in the prediction time domain P, including time k, is determined according to the output of the prediction unit. In the process of the prediction unit predicting the cooling water mass flow rate, the ARMAX dynamic model uses the actual cooling water mass flow rate at time k-1 as the input prediction data and obtains the gas outlet temperature in the prediction time domain P, including time k.

3. The industrial gas cooling optimization control system according to claim 2, characterized in that, The process of determining the control increment of cooling water mass flow rate at time k based on the input-output relationship of the prediction process using the ARMAX dynamic model includes: Define a decision variable vector U of length P, representing the control increment for the next P steps: in, - Indicating targeting The control increment of P steps in the future at time 1; Based on the decision variable vector U and the control quantity at time k-1 The complete control trajectory within the next P steps is calculated by accumulating and summing the data. in, express The amount of control at any given moment; From 0 to variables, From 0 to Variables; The calculated control trajectory is filled into the columns of the FutureInputs matrix; Taking a historical data matrix, the prediction time domain P, and FutureInputs as inputs, the ARMAX model is called as the prediction function. The prediction function performs P iterative calculations forward based on the model's dynamic equations to obtain the sequence of predicted output values ​​for the system in the next P steps. in, This represents the prediction of the gas outlet temperature at time k+i from time k. This is the sequence of predicted output values ​​for the system in the next P steps; The total cost is obtained by calculating the difference between the predicted and set gas outlet temperatures, and the sum of complete control increments within P steps. The optimal control increment sequence is then obtained by minimizing the total cost. , That is, the optimal control increment U; Then determine the actual control quantity that should be executed at the current time k: in This indicates the new cooling water mass flow rate setpoint. for The value in the interval represents the increment at time k.

4. The industrial gas cooling optimization control system according to claim 3, characterized in that, The total cost ;in: To track costs: in, This indicates the set gas outlet temperature. Weighting for tracking costs; To control costs: used to penalize drastic changes in the control quantity; its calculation formula is: in, Weighting for cost control; for The square of, for Control increment at any given moment.

5. An industrial gas cooling optimization control system according to claim 4, characterized in that, The ARMAX dynamic model is pre-trained. The pre-training process of the ARMAX dynamic model includes: The mathematical expression for the ARMAX model is as follows: Where y(t) is the system output at time t, i.e., the gas outlet temperature; u(t) is the system input vector at time t; and e(t) is a white noise sequence, representing an unmeasurable disturbance. It is a shift operator, i.e. ;nk is the pure time delay between system input and output; A(q), B(q), C(q) are the time delays with respect to the shift operator. The polynomial has coefficients that are parameters that the model needs to identify through data: in, , , These are the polynomial coefficients; , , To set the structural parameters of the ARMAX model, i.e., the order of the polynomial; Input training data, which includes cooling water mass flow rate, gas outlet temperature, and other state data. During training, adjust all unknown coefficients in polynomials A(q), B(q), and C(q) to minimize the error between the model's predicted output and the actual historical output given historical input. After training, an ARMAX model containing all identified coefficients will be obtained.

6. The industrial gas cooling optimization control system according to claim 5, characterized in that, Structural parameters of ARMAX model It is 50. , The dimension is the other state data plus 1. =1; pure time delay , The dimension is the other state data plus 1.

7. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement an industrial gas cooling optimization control system as described in any one of claims 1 to 6.

8. An industrial gas cooling optimization and control device, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement an industrial gas cooling optimization control system according to any one of claims 1 to 6.

Images