Simulation modeling method of intelligent control system for sulfuric acid production

Abstract

The invention discloses a simulation modeling method of an intelligent control system for sulfuric acid production, and belongs to the field of acid making simulation modeling, and the method comprises the steps: determining the parameter type of the sulfuric acid production control system, and dividing the parameter type into an agent observation variable parameter and an agent action variable parameter; the intelligent agent observation variable parameters serve as main output variables of the system simulation model, and the intelligent agent action variable parameters serve as input variables of the system simulation model; an integral system model framework is constructed according to the characteristics of an acid making treatment system and the actual target of sulfuric acid treatment control in each stage. The method aims at learning and constructing a simulation model from historical data on the basis that extra interaction with a real environment is not needed, changes and behaviors of the environment can be simulated and predicted through the environment simulated by the simulation model, and model parameters and structures are automatically adjusted to adapt to changes of the production environment, so that the accuracy and adaptability of the model are improved.

Description

Technical Field

[0001] The invention relates to the field of simulation modeling of acid production, and in particular to a simulation modeling method of an intelligent control system for sulfuric acid production. Background Art

[0002] The acid production process has typical nonlinear, large time lag, multiple inputs and multiple outputs, and strong coupling characteristics, which makes it very difficult to build an accurate acid production simulation model. In addition, complex acid production models usually require a lot of computing resources and time overhead, which makes it difficult to meet the needs of real-time control based on acid production.

[0003] The most common approach to building a system simulation model is to use deep learning methods to perform supervised learning from expert data (also known as behavioral cloning), but this often does not work well because the transition probability function (or agent strategy) in the environment model is usually much more complex than the data distribution function in deep learning. It is generally recognized that the data generated by actual production systems is highly temporal, and direct use of supervised learning will accumulate compound errors. Summary of the invention

[0004] The purpose of the present invention is to provide a simulation modeling method for an intelligent control system of sulfuric acid production, which effectively solves the problem that the existing simulation model of the acid production system is poor in effect and the direct use of supervised learning will cause errors.

[0005] The technical solution of the present invention is as follows:

[0006] A simulation modeling method for an intelligent control system for sulfuric acid production comprises the following steps:

[0007] Step 1: Determine the parameter types of the sulfuric acid production control system and divide them into agent observation variable parameters and agent action variable parameters;

[0008] Step 2: Take the agent observation variable parameters as the main output variables of the system simulation model, and the agent action variable parameters as the input variables of the system simulation model;

[0009] Step 3: Based on the characteristics of the acid treatment system and the actual objectives of sulfuric acid treatment control at each stage, an overall system model framework is constructed;

[0010] ht+1=fθ(ht,ot,at)

[0011] ot+1=gφ(ht+1,ot)

[0012] Step 4: After completing the model design, determine the overall framework of the system and generate a simulation model of the adversarial imitation learning algorithm training system;

[0013] Step 5: After the training is completed, conduct corresponding tests and then use the simulation model for strategy learning.

[0014] Furthermore, ot is the output variable, including tower top pressure, steam flow, air flow, liquid sulfur flow, tower liquid level, acid balance, sulfuric acid cooling temperature, fan speed, heat exchanger temperature and furnace coal amount.

[0015] Furthermore, at is the system input variable, including steam flow rate, liquid sulfur flow rate, fan outlet air volume and absorption tower inlet temperature.

[0016] Furthermore, ht is the state variable inside the system, fθ is the state transition model of the system, gφ is the observation model, and Θ=(θ,φ) is the parameter of the model.

[0017] Furthermore, the test content includes:

[0018] ①, Derivation test, given the initial conditions, the trajectory deduced by the simulation model should be as close to the real trajectory as possible;

[0019] ② Response curve test to ensure that the relationship between the action and output learned by the model is consistent with the relationship between the action and output in physical reality;

[0020] ③. Data distribution test: when the simulation model is accurate enough to simulate the acid-making system, the data distribution of the output generated by the model based on historical data should be similar to the data distribution of the actual output.

[0021] Furthermore, after completing the model design, the network structure inside the model is designed according to the acid production process mechanism, and a mechanism data model is constructed to achieve the integration of mechanism and data.

[0022] The advantages of the present invention are:

[0023] 1. The present invention can simulate and predict environmental changes and behaviors through the simulated environment, and automatically adjust model parameters and structure to adapt to changes in the production environment, thereby improving the accuracy and adaptability of the model.

[0024] 2. The system is based on a data-driven development toolkit based on a generative adversarial imitation learning algorithm, which aims to learn and build simulation models from historical data without the need for additional interaction with the real environment.

[0025] 3. The present invention uses a system simulation model to mathematically describe the dynamic characteristics of the processing system, such as pressure, steam flow, air flow, liquid sulfur flow, and liquid level of each tower. The model supports interactive deduction and can predict future changes in acid production based on any input control instruction sequence. Based on the predictive ability of the acid production system simulation model, model-based intelligent control can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is in the flow chart of the simulation modeling method for sulfuric acid production of the present invention;

[0027] Figure 2 It is a schematic diagram of the relationship between the output variables and the input variables of the present invention. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0029] like Figure 1 to Figure 2 As shown, the present invention provides a simulation modeling method for an intelligent control system of sulfuric acid production, which includes the following steps:

[0030] Step 1: Determine the parameter types of the sulfuric acid production control system and divide them into agent observation variable parameters and agent action variable parameters;

[0031] Step 2: Take the agent observation variable parameters as the main output variables of the system simulation model, and the agent action variable parameters as the input variables of the system simulation model;

[0032] Step 3: Based on the characteristics of the acid treatment system and the actual objectives of sulfuric acid treatment control at each stage, an overall system model framework is constructed;

[0033] ht+1=fθ(ht,ot,at)

[0034] ot+1=gφ(ht+1,ot)

[0035] Among them, ot is the output variable, including tower top pressure, steam flow, air flow, liquid sulfur flow, tower liquid level, acid balance, sulfuric acid cooling temperature, fan speed, heat exchanger temperature and furnace coal amount.

[0036] at is the system input variable, including steam flow rate, liquid sulfur flow rate, fan outlet air volume and absorption tower inlet temperature.

[0037] ht is the state variable inside the system, fθ is the state transition model of the system, gφ is the observation model, Θ=(θ,φ) is the parameter of the model

[0038] Step 4: After completing the model design, design the internal network structure of the model according to the acid production process mechanism, build a mechanism data model, realize the fusion of mechanism and data, determine the overall framework of the system, and generate a simulation model of the adversarial imitation learning algorithm training system;

[0039] Step 5: After the training is completed, conduct corresponding tests and then use the simulation model for strategy learning. The test content includes:

[0040] ①, Derivation test, given the initial conditions, the trajectory deduced by the simulation model should be as close to the real trajectory as possible;

[0041] ② Response curve test to ensure that the relationship between the action and output learned by the model is consistent with the relationship between the action and output in physical reality;

[0042] ③. Data distribution test: when the simulation model is accurate enough to simulate the acid-making system, the data distribution of the output generated by the model based on historical data should be similar to the data distribution of the actual output.

[0043] The method of using the above embodiment is as follows:

[0044] initialization:

[0045] At time step t=0, the hidden state h0 is initialized to a fixed value or a random value, and the first output o0 is initialized, if any.

[0046] Iterative process (for t = 1, 2, …, T): Use the hidden state ht of the current time step and the external input value at (the output ot-1 of the previous time step, in the original equation, ot is used as the input of fθ, not the direct input of gφ), calculate the hidden state ht+1 of the next time step: ht+1 = fθ(ht, at, ot) (adjust the input according to actual situation) Use the hidden state ht+1 of the next time step and the output ot of the current time step (or according to the equation, directly ht+1, if gφ does not require ot as input), calculate the output ot+1 of the next time step: ot+1 = gφ(ht+1,ot) (adjust according to actual situation).

[0047] However, in practical applications, gφ may not depend directly on ot, but only on ht+1. If gφ does not require ot, the equation is simplified to ot+1=gφ(ht+1), output: At each time step, you can choose to output ot.

[0048] In practical applications, the architecture and equations are adjusted according to the needs of specific problems.

[0049] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A simulation modeling method for an intelligent control system of sulfuric acid production, characterized in that: The steps include: Step 1: Determine the parameter types of the sulfuric acid production control system and divide them into agent observation variable parameters and agent action variable parameters; Step 2: Take the agent observation variable parameters as the main output variables of the system simulation model, and the agent action variable parameters as the input variables of the system simulation model; Step 3: Based on the characteristics of the acid treatment system and the actual objectives of sulfuric acid treatment control at each stage, an overall system model framework is constructed; ht+1=fθ(ht,ot,at) ot+1=gφ(ht+1,ot) Step 4: After completing the model design, determine the overall framework of the system and generate a simulation model of the adversarial imitation learning algorithm training system; Step 5: After the training is completed, conduct corresponding tests and then use the simulation model for strategy learning.

2. The simulation modeling method of the intelligent control system for sulfuric acid production according to claim 1, characterized in that: Among them, ot is the output variable, including tower top pressure, steam flow, air flow, liquid sulfur flow, tower liquid level, acid balance, sulfuric acid cooling temperature, fan speed, heat exchanger temperature and furnace coal amount.

3. The simulation modeling method of the intelligent control system for sulfuric acid production according to claim 2, characterized in that: at is the system input variable, including steam flow rate, liquid sulfur flow rate, fan outlet air volume and absorption tower inlet temperature.

4. The simulation modeling method of the intelligent control system for sulfuric acid production according to claim 3, characterized in that: ht is the state variable inside the system, fθ is the state transition model of the system, gφ is the observation model, and Θ=(θ,φ) is the parameter of the model.

5. The simulation modeling method of the intelligent control system for sulfuric acid production according to claim 1, characterized in that: The test contents include: ①, Derivation test, given the initial conditions, the trajectory deduced by the simulation model should be as close to the real trajectory as possible; ② Response curve test to ensure that the relationship between the action and output learned by the model is consistent with the relationship between the action and output in physical reality; ③. Data distribution test: when the simulation model is accurate enough to simulate the acid-making system, the data distribution of the output generated by the model based on historical data should be similar to the data distribution of the actual output.

6. The simulation modeling method of the intelligent control system for sulfuric acid production according to claim 1, characterized in that: After completing the model design, the network structure inside the model is designed according to the acid production process mechanism, and a mechanism data model is constructed to achieve the integration of mechanism and data.

Images