Anhydrous ethanol synthesis production monitoring and control system

Abstract

The invention relates to the technical field of absolute ethyl alcohol synthesis, and discloses an absolute ethyl alcohol synthesis production monitoring and control system, which integrates a reaction kettle monitoring module, a synthesis equipment module and an intelligent control module, and realizes comprehensive monitoring and intelligent control of an absolute ethyl alcohol synthesis process. The reaction kettle monitoring module collects data such as temperature, pressure, liquid level and reactant concentration in real time, and a basis is provided for intelligent control. The synthesis equipment module comprises a reaction kettle, a stirrer, a heating / cooling device and a rectifying tower and is responsible for synthesis and purification of absolute ethyl alcohol. The intelligent control module constructs an absolute ethyl alcohol synthesis decision-making model, models a synthesis process into a Markov decision-making process, and automatically adjusts an equipment working mode according to real-time data and preset parameters so as to optimize production. According to the system, the energy utilization efficiency is effectively improved, the production cost is reduced, and safe production is ensured.

Description

Technical Field

[0001] The invention relates to the technical field of anhydrous ethanol synthesis, in particular to an anhydrous ethanol synthesis production monitoring and control system. Background Art

[0002] As an important organic chemical raw material, anhydrous ethanol is widely used in many fields such as medicine, chemical industry, and energy.

[0003] At present, the synthetic production of anhydrous ethanol usually faces the following problems:

[0004] In the production process of anhydrous ethanol, key equipment such as reactors, agitators, heating / cooling devices, and distillation towers consume a lot of energy. Due to the lack of effective real-time monitoring and control methods, operators often find it difficult to accurately grasp the energy consumption of equipment, resulting in low energy efficiency and high production costs.

[0005] There are potential safety risks in the production process. The synthesis reaction of anhydrous ethanol is a complex chemical process involving extreme conditions such as high temperature and high pressure. If the temperature, pressure and other parameters in the reactor are not effectively controlled, dangerous situations such as overpressure and overtemperature are likely to occur, posing serious safety hazards to production. Once a safety accident occurs, it will not only cause equipment damage and production interruption, but may also pose a threat to the life safety of operators. Summary of the invention

[0006] The object of the present invention is to provide a monitoring and control system for the production of anhydrous ethanol synthesis to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a monitoring and control system for anhydrous ethanol synthesis production, the system comprising:

[0008] Reactor monitoring module, integrating temperature sensor, pressure sensor, liquid level sensor and reactant concentration sensor, for real-time monitoring and collection of temperature, pressure, liquid level and reactant concentration data in the anhydrous ethanol synthesis reactor;

[0009] Synthesis equipment module, including reactor, stirrer, heating / cooling device and distillation column, used for the synthesis and purification of anhydrous ethanol;

[0010] The intelligent control module automatically adjusts the working mode of the synthesis equipment based on the data collected by the reactor monitoring module, the current status of the synthesis equipment and the preset production parameters, so as to efficiently control the synthesis production of anhydrous ethanol.

[0011] Preferably, the implementation method of the intelligent control module is:

[0012] Construct anhydrous ethanol synthesis decision model, and model the anhydrous ethanol synthesis process as a Markov decision process. The specific definitions include:

[0013] State space: includes the temperature, pressure, liquid level, reactant concentration, state of the synthesis equipment and vector representation of the current production stage in the reactor;

[0014] Action space: contains the possible actions that the synthesis equipment may take during the production process, including adjusting the stirring speed, heating or cooling the reactor, adjusting the reactant feed rate, and starting or stopping the distillation column;

[0015] Reward: A reward function designed based on the output, quality, equipment energy consumption and production safety indicators of anhydrous ethanol is used to evaluate the results of actions;

[0016] Policy: A rule for choosing an action in a given state, optimized through learning to maximize the cumulative reward.

[0017] Preferably, based on the goal of anhydrous ethanol synthesis, a reward function of the anhydrous ethanol synthesis decision model is designed, wherein the reward function includes reward mechanisms such as increased output, quality optimization, reduced equipment energy consumption, improved production safety, and avoidance of production accidents, while setting penalties for synthesis failures or equipment failures.

[0018] Preferably, the reward function is designed in the following manner:

[0019] Design a production reward function, and calculate the production reward value based on the deviation between the real-time production of anhydrous ethanol and the preset target production. The specific algorithm formula is:

[0020] R production =k production × (actual output - target output) 2

[0021] Among them, k production is the output reward coefficient. When the actual output is higher than the target output, R production is a positive reward; when the actual output is lower than the target output, R production Negative rewards;

[0022] Design a quality reward function and calculate the quality reward value based on the deviation between the quality index of anhydrous ethanol and the preset quality standard. The specific algorithm formula is:

[0023]

[0024] Among them, k quality is the quality reward coefficient, n is the number of quality indicators, when the actual quality is better than the quality standard, R quality is a positive reward; when the actual quality is worse than the quality standard, Rquality Negative rewards;

[0025] Design an energy consumption penalty function, and calculate the energy consumption penalty value according to the deviation between the real-time energy consumption of the synthetic equipment and the preset energy consumption budget. The specific algorithm formula is:

[0026] P energy =k energy × (actual energy consumption - energy consumption budget)

[0027] Among them, k energy is the energy penalty coefficient. When the actual energy consumption is higher than the energy budget, P energy is a negative penalty; when the actual energy consumption is lower than the energy budget, P energy is zero or positive reward;

[0028] Design a safety reward function and calculate the safety reward value based on the deviation between the safety index in the production process and the preset safety threshold. The specific algorithm formula is:

[0029]

[0030] Among them, k safegy is the security reward coefficient, m is the number of security indicators, and when the actual security indicator is better than the security threshold, R safety is a positive reward; when the actual safety index is worse than the safety threshold, R safety For negative punishment;

[0031] Design an accident penalty function. When a safety accident occurs during the production process, calculate the accident penalty value based on the severity and impact range of the accident. The specific algorithm formula is:

[0032] P accident =k accident ×(Accident Level)

[0033] Among them, k accident is the accident penalty coefficient, and the accident level is quantified according to the severity and impact scope of the accident;

[0034] Combining the above rewards and penalties, the total reward value is calculated. The specific algorithm formula is:

[0035] R total =R production +R quality -P energy +R safety -P accident

[0036] The total reward value R total As the input of the decision-making model in the intelligent control module, it guides the model to learn the optimal control strategy.

[0037] Preferably, a deep Q network DQN algorithm is used to iteratively train the anhydrous ethanol synthesis decision model.

[0038] Preferably, the step of training the anhydrous ethanol synthesis decision model comprises:

[0039] Step 1: Initialize two deep neural networks with the same structure. One is an online updated network, which is used to calculate the Q value of the current strategy selection and perform iterative Q value updates and gradient descent back propagation. The other is a target network that periodically copies the weights of the former and is used to calculate the target Q value.

[0040] Step 2: Set up an experience replay buffer to store the four-tuple data [state s t , action a t , reward r_t, new state s t+1 ] is stored in the playback buffer. When training the Q network, some data is randomly sampled from the playback buffer for training;

[0041] Step 3: Use the following formula to update the Q value:

[0042] Q(s,a)=r+γ*max a' Q target (s,a′)

[0043] Among them, Q(s,a) represents the Q value of taking action a in state s, r represents the immediate reward, γ represents the discount factor, and Q target (s',a') represents the Q value of taking action a' in the new state s' calculated by the target network;

[0044] Step 4: Copy the parameters in the online network to the target network;

[0045] Step 5: Use the gradient descent method to update the network weights and calculate the gradient through the back-propagation algorithm to minimize the difference between the predicted Q value and the actual Q value;

[0046] Step 6: Repeat steps 2 to 5 until the performance indicators of the model meet the requirements and the training of the anhydrous ethanol synthesis decision model is completed.

[0047] Preferably, in step 5, the formula for updating the parameters of the gradient descent algorithm is:

[0048]

[0049] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0050] Preferably, during the model training process, the mean square error (MSE) algorithm is used to evaluate the performance of the model:

[0051] For each state-action pair (s, a), calculate the error between the predicted Q value and the actual Q value and square it. Then average the squared errors of all state-action pairs to get the mean square error (MSE). The specific formula is as follows:

[0052]

[0053] Where N is the total number of state-action pairs sampled during the evaluation process, and Σ represents the summation operation. The smaller the MSE value, the closer the predicted Q value of the model is to the actual Q value, and the better the performance of the model. Conversely, the larger the MSE value, the worse the prediction ability of the model, and further optimization or retraining is needed.

[0054] Preferably, the performance of the anhydrous ethanol synthesis decision model in practical applications is evaluated by combining comprehensive simulation testing with field application verification.

[0055] Preferably, the method for conducting comprehensive simulation test and field application verification on the anhydrous ethanol synthesis decision model includes:

[0056] Using computational chemistry simulation software and thermodynamic models, a highly simulated anhydrous ethanol synthesis simulation environment is constructed. This environment is used to accurately reflect the physical and chemical changes in the actual synthesis process, as well as the effects of different operating conditions on synthesis efficiency, energy consumption, and by-product generation.

[0057] The anhydrous ethanol synthesis decision model to be evaluated is applied in a simulation environment, and the optimal synthesis scheme is generated according to the input raw material conditions, reaction parameters and equipment status, including the setting of reaction temperature, pressure, catalyst type and dosage, and reaction time;

[0058] Adopt the recognized traditional control algorithm, conduct simulation tests under the same conditions, and generate the corresponding synthesis scheme;

[0059] In the simulation environment, the model recommendation scheme and the traditional control algorithm scheme were executed respectively, and the anhydrous ethanol synthesis efficiency, equipment energy consumption, environmental emissions, and temperature and pressure control stability during the reaction process were recorded and compared.

[0060] Based on the simulation test, a representative anhydrous ethanol production unit was selected for field application verification; according to the scheme in the simulation test, the model recommendation scheme and the traditional control algorithm scheme were implemented respectively, and actual operation data were collected;

[0061] Combining the results of simulation tests with field application verification, a comprehensive evaluation of the anhydrous ethanol synthesis decision model was conducted.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] This system provides a comprehensive and effective solution to the problems of low energy efficiency and potential safety hazards in the current anhydrous ethanol synthesis production. By integrating the reactor monitoring module, synthesis equipment module and intelligent control module, this system realizes real-time monitoring and intelligent control of the anhydrous ethanol synthesis production process, bringing significant beneficial effects.

[0064] In terms of energy utilization, the system collects energy consumption data of key equipment in real time through the reactor monitoring module, such as temperature, pressure, liquid level and reactant concentration, and provides operators with accurate energy consumption feedback. Based on these data and combined with preset production parameters, the intelligent control module automatically adjusts the working mode of the synthesis equipment, such as stirring speed, power of heating or cooling reactor, reactant feed rate, etc., to achieve optimal energy utilization.

[0065] In terms of production safety, the system can timely detect and warn potential safety risks, such as overpressure, overtemperature and other dangerous situations, by real-time monitoring of key parameters such as temperature and pressure in the reactor. The intelligent control module can quickly respond to these warning signals and automatically take corresponding control measures, such as adjusting the power of the heating / cooling device, starting or stopping the distillation tower, etc., to ensure the safety of the production process.

[0066] The intelligent control module of this system builds an anhydrous ethanol synthesis decision model and models the anhydrous ethanol synthesis process as a Markov decision process, thus achieving precise control and optimization of the production process. By defining elements such as state space, action space, rewards and strategies, the system can automatically learn and optimize production strategies to maximize cumulative rewards, that is, to achieve high yield, high quality, low energy consumption and safe production of anhydrous ethanol. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a working principle diagram of the present invention;

[0068] Figure 2 Schematic diagram of the design of the reward function;

[0069] Figure 3 This is a training flow chart for the anhydrous ethanol synthesis decision model;

[0070] Figure 4 Flowchart for the performance evaluation of the anhydrous ethanol synthesis decision model. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] See also Figure 1-4 The present invention provides a technical solution: a monitoring and control system for anhydrous ethanol synthesis production, the system comprising:

[0073] Reactor monitoring module, which integrates high-precision temperature sensors, pressure sensors, liquid level sensors and reactant concentration sensors. These sensors are arranged at key positions of the anhydrous ethanol synthesis reactor to achieve real-time and accurate monitoring of its internal temperature, pressure, liquid level and reactant concentration.

[0074] The synthesis equipment module includes key equipment such as reactor, agitator, heating / cooling device and distillation tower. The reactor is the main place for chemical reaction, and it is equipped with a stirrer to promote the full mixing of reactants; the heating / cooling device is responsible for adjusting the temperature of the reactor according to the reaction requirements; the distillation tower is used for subsequent product purification to ensure that the quality of anhydrous ethanol meets the predetermined standards.

[0075] The intelligent control module is responsible for receiving real-time data from the reactor monitoring module, and automatically adjusting the working mode of the synthesis equipment through complex algorithm analysis based on the current status of the synthesis equipment and the preset production parameters, so as to achieve efficient control of anhydrous ethanol synthesis production.

[0076] The implementation method of the intelligent control module is:

[0077] A decision model for the synthesis of anhydrous ethanol was constructed. The model abstracted the synthesis process of anhydrous ethanol into a Markov decision process, which is specifically defined as follows:

[0078] State space: The state space is a multidimensional vector that fully describes the various key parameters in the anhydrous ethanol synthesis process, including the temperature, pressure, liquid level, reactant concentration, current state of the synthesis equipment, and the production stage in the reactor. This state information provides comprehensive environmental perception capabilities for the decision-making model.

[0079] Action space: The action space includes all the actions that the synthesis equipment may take during the production process, such as adjusting the stirring speed to change the mixing degree of the reactants, heating or cooling the reactor to control the reaction temperature, adjusting the feed rate of the reactants to optimize the reaction conditions, and starting or stopping the distillation column to purify the product.

[0080] Reward: The reward function is a key indicator for evaluating the quality of action results. It is calculated comprehensively based on multiple dimensions such as the output, quality, equipment energy consumption, and production safety of anhydrous ethanol. Through the reward function, the decision-making model can maximize production efficiency and energy utilization while ensuring production safety and product quality.

[0081] Strategy: Strategy is the rule for decision-making models to choose actions under given conditions. Through continuous learning and optimization, the strategy can gradually adapt to changes in the production environment, thereby maximizing the cumulative reward in the long-term production process. This learning strategy enables the intelligent control module to autonomously adjust and optimize the production process, achieving efficient, stable and safe production of anhydrous ethanol synthesis.

[0082] The present invention will be further described below in conjunction with Examples 1 to 3:

[0083] Embodiment 1:

[0084] The design of the reward function is the core part of the intelligent control module. The reward function directly affects the learning effect of the decision model and the optimization of the control strategy. Figure 2 ,This embodiment describes in detail the design method of the reward function and its algorithm formula with specific examples.

[0085] The reward function is designed in the following ways:

[0086] Output reward function: The output reward function is designed to motivate the system to increase the output of anhydrous ethanol while avoiding waste of resources caused by overproduction. Assuming the preset target output is 1000 kg / hour, the actual output is monitored in real time by the flow meter of the reactor. The design of the output reward function is as follows:

[0087] When the actual output is lower than the target output, a -10 reward value will be given for every 1% decrease in output;

[0088] When the actual output is equal to the target output, a reward value of 0 is given;

[0089] When the actual output is higher than the target output but not more than 105%, a bonus of 5 will be given for every 1% increase in output;

[0090] When the actual output exceeds the target output by 105%, a reward value of -2 is given for every 1% increase in output to punish overproduction.

[0091] The algorithm formula of the output reward function is:

[0092]

[0093] if actual output > target output; R production =0;

[0094]

[0095]

[0096] Quality reward function: The quality reward function is designed to ensure that the quality of anhydrous ethanol meets the preset standards. Assume that the quality standards of anhydrous ethanol include a purity of not less than 99.5% and a moisture content of not more than 0.2%. The quality reward function is designed as follows:

[0097] When the purity of anhydrous ethanol increases by 0.1% (above 99.5%), a bonus of 10 is given;

[0098] When the water content of anhydrous ethanol decreases by 0.01% (below 0.2%), a reward value of 5 is given;

[0099] When the purity of anhydrous ethanol is lower than 99.5% or the water content is higher than 0.2%, a bonus value of -20 is given for every 0.1% or 0.01% deviation from the standard.

[0100] The algorithm formula of the quality reward function is:

[0101]

[0102] Among them, k quality is the quality reward coefficient, n is the number of quality indicators, when the actual quality is better than the quality standard, R quality is a positive reward; when the actual quality is worse than the quality standard, R quality Negative reward.

[0103] Energy consumption penalty function: The energy consumption penalty function aims to reduce the energy consumption of the synthesis equipment and improve energy efficiency. Assume that the preset energy consumption budget is 500 kWh / h. The design of the energy consumption penalty function is as follows:

[0104] When the actual energy consumption is lower than the energy budget, no penalty will be imposed;

[0105] When the actual energy consumption is higher than the energy budget but not more than 110%, a penalty value of -1 is given for every 1% increase in energy consumption;

[0106] When the actual energy consumption exceeds 110% of the energy budget, a penalty value of -5 is given for every 1% increase in energy consumption.

[0107] The algorithm formula of the energy consumption penalty function is:

[0108] P energy =k energy × (actual energy consumption - energy consumption budget)

[0109] Among them, k energyis the energy penalty coefficient. When the actual energy consumption is higher than the energy budget, P energy is a negative penalty; when the actual energy consumption is lower than the energy budget, P energy Zero or positive reward.

[0110] Safety reward function: The safety reward function is designed to ensure safety during the production process. Assume that the safety pressure range of the equipment is 0-100 Pascals and the safety temperature range is 0-50 degrees Celsius. The design of the safety reward function is as follows:

[0111] No rewards or penalties are given when equipment pressure or temperature is within safe ranges;

[0112] When the equipment pressure or temperature exceeds the safety range but does not exceed 110%, a penalty value of -2 is given for every 1% deviation;

[0113] When the equipment pressure or temperature exceeds 110% of the safety range, a penalty value of -10 is given for every 1% deviation, and an emergency shutdown is triggered.

[0114] The algorithm formula of the security reward function is:

[0115]

[0116] Among them, k safegy is the security reward coefficient, m is the number of security indicators, and when the actual security indicator is better than the security threshold, R safety is a positive reward; when the actual safety index is worse than the safety threshold, R safety Negative punishment.

[0117] Accident penalty function: The accident penalty function is designed to prevent safety accidents from occurring during the production process. Assume that the accident levels are divided into minor, moderate, and severe, corresponding to level values ​​of 1, 2, and 3, respectively. The design of the accident penalty function is as follows:

[0118] When a minor accident occurs, a penalty value of -100 is given;

[0119] When a medium accident occurs, a penalty value of -500 is given;

[0120] When a serious accident occurs, a penalty value of -1000 is given and the machine is shut down immediately.

[0121] The algorithm formula of the accident penalty function is:

[0122] P accident =k accident ×(Accident Level)

[0123] Among them, k accident is the accident penalty coefficient, and the accident level is quantified according to the severity and impact scope of the accident.

[0124] Combining the above rewards and penalties, the total reward value is calculated. The specific algorithm formula is:

[0125] R total =R production +R quality -P energy +R safety -P accident

[0126] The total reward value R total As the input of the decision-making model in the intelligent control module, it guides the model to learn the optimal control strategy.

[0127] Embodiment 2:

[0128] See attached Figure 3 , this embodiment is used to describe in detail the implementation method of anhydrous ethanol synthesis decision model training based on the deep Q network (DQN) algorithm.

[0129] The steps for training the anhydrous ethanol synthesis decision model include:

[0130] Step 1: Initialize two deep neural networks with the same structure, called the online network (evaluatenetwork) and the target network (target network). Each network consists of several fully connected layers, and the last layer outputs the Q value corresponding to each action.

[0131] Step 2: Set an experience replay buffer with a capacity of N. Each time the four-tuple data [state s t , action a t , reward r_t, new state s t+1 ] is stored in the playback buffer. When training the Q network, some data is randomly sampled from the playback buffer for training; when the buffer is full, the new data will overwrite the oldest data.

[0132] Step 3: Start model training:

[0133] Step 3.1, Initialization: Initialize the weights of the online network and the target network, and set hyperparameters such as discount factor γ, learning rate α, and exploration rate ε.

[0134] Step 3.2, Experience Sampling: In each iteration, randomly sample M quadruple data (s i ,a i ,r i ,s i+1 ).

[0135] Step 3.3, Q value update: For each sampled data, use the target network to calculate the target Q value: Use the following formula to update the Q value:

[0136] Q(s,a)=r+γ*max a' Q target (s,a′)

[0137] Among them, Q(s,a) represents the Q value of taking action a in state s, r represents the immediate reward, γ represents the discount factor, and Q target (s',a') represents the Q value of taking action a' in the new state s' calculated by the target network.

[0138] Step 3.4, loss calculation: Use the mean square error loss function to calculate the difference between the Q value predicted by the online network and the target Q value. For each state-action pair (s, a), calculate the error between the predicted Q value and the actual Q value, and square it, then average the squared errors of all state-action pairs to get the mean square error MSE. The specific formula is as follows:

[0139]

[0140] Where N is the total number of state-action pairs sampled during the evaluation process, and Σ represents the summation operation. The smaller the MSE value, the closer the predicted Q value of the model is to the actual Q value, and the better the performance of the model. Conversely, the larger the MSE value, the worse the prediction ability of the model, and further optimization or retraining is needed.

[0141] Step 3.5, Gradient Descent and Back Propagation: Use the gradient descent method to update the weights of the online network, and calculate the gradient through the back propagation algorithm. The formula for updating the parameters of the gradient descent algorithm is:

[0142]

[0143] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0144] Step 4: At certain intervals, copy the parameters in the online network to the target network;

[0145] Step 5: In each iteration, use the ε-greedy strategy to select actions. An action is randomly selected with probability ε, and the action with the largest current Q value is selected with probability 1-ε.

[0146] Step 6: Repeat steps 2 to 5 until the performance indicators of the model meet the requirements and the training of the anhydrous ethanol synthesis decision model is completed.

[0147] In the anhydrous ethanol synthesis process, the state space includes the temperature, pressure, liquid level and reactant concentration in the reactor, and the action space includes adjusting the stirring speed, heating or cooling the reactor, adjusting the reactant feed rate, etc. Through simulation and actual production data, state transition and reward information can be obtained. Using the above method, the anhydrous ethanol synthesis decision model is iteratively trained. During the training process, the decision-making ability of the model is optimized by continuously adjusting the hyperparameters and network structure. Finally, the trained model can provide accurate control strategies for the anhydrous ethanol synthesis process based on real-time state information.

[0148] Embodiment 3:

[0149] See attached Figure 4 This embodiment is used to describe in detail the implementation of an evaluation method for an anhydrous ethanol synthesis decision model.

[0150] Computer chemical simulation software, such as Aspen HYSYS or ChemCAD, is used to construct a simulation environment for the synthesis of anhydrous ethanol. The environment includes key equipment modules such as raw material pretreatment, reactor, separation tower, condenser, heater, and corresponding material flow, energy flow, and information flow. The synthesis process of anhydrous ethanol is simulated by inputting parameters such as the composition, flow rate, temperature, pressure, and other parameters of the raw materials, as well as the type, size, and operating conditions of the reactor.

[0151] The anhydrous ethanol synthesis decision model to be evaluated is imported into the simulation environment. Taking a certain synthesis as an example, the model recommends the optimal synthesis scheme based on the input raw material ethanol concentration, water content, catalyst type and dosage, reaction temperature, pressure and other conditions through internal algorithm calculation. The scheme includes the operating conditions of the reactor (such as the temperature is set to 250°C and the pressure is set to 5MPa), the selection of the catalyst (such as using a certain type of solid acid catalyst), and the reaction time (such as set to 4 hours).

[0152] Under the same conditions, the traditional PID control algorithm was used for simulation test. The PID algorithm automatically adjusted the heating, cooling, pressurization, and decompression operations according to the real-time data of the reactor temperature and pressure to maintain the stability of the reaction conditions. The efficiency, energy consumption, environmental emissions, and safety indicators of anhydrous ethanol synthesis under the traditional control algorithm were recorded.

[0153] The model recommendation scheme and the traditional control algorithm scheme were executed to simulate the synthesis process of anhydrous ethanol. The anhydrous ethanol yield, purity, equipment energy consumption (such as electricity and steam consumption), and environmental friendliness indicators (such as wastewater and exhaust gas emissions) under the two schemes were recorded and compared. At the same time, the temperature and pressure changes during the reaction process were monitored to evaluate the impact of the two schemes on production safety.

[0154] An anhydrous ethanol production plant was selected as the site for field application verification. According to the scheme in the simulation test, the model recommendation scheme and the traditional control algorithm scheme were implemented respectively. During the actual operation process, data such as the yield, purity, energy consumption, environmental emissions and safety records of anhydrous ethanol were collected and recorded.

[0155] Based on the data from simulation tests and field application verification, a comprehensive performance evaluation of the anhydrous ethanol synthesis decision model was conducted.

[0156] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0157] 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 monitoring and control system for anhydrous ethanol synthesis production, characterized in that: The system comprises: Reactor monitoring module, integrating temperature sensor, pressure sensor, liquid level sensor and reactant concentration sensor, for real-time monitoring and collection of temperature, pressure, liquid level and reactant concentration data in the anhydrous ethanol synthesis reactor; Synthesis equipment module, including reactor, stirrer, heating / cooling device and distillation column, used for the synthesis and purification of anhydrous ethanol; The intelligent control module automatically adjusts the working mode of the synthesis equipment based on the data collected by the reactor monitoring module, the current status of the synthesis equipment and the preset production parameters, so as to efficiently control the synthesis production of anhydrous ethanol.

2. The anhydrous ethanol synthesis production monitoring and control system according to claim 1, characterized in that: The implementation method of the intelligent control module is: Construct anhydrous ethanol synthesis decision model, and model the anhydrous ethanol synthesis process as a Markov decision process. The specific definitions include: State space: includes the temperature, pressure, liquid level, reactant concentration, state of the synthesis equipment and vector representation of the current production stage in the reactor; Action space: contains the possible actions that the synthesis equipment may take during the production process, including adjusting the stirring speed, heating or cooling the reactor, adjusting the reactant feed rate, and starting or stopping the distillation column; Reward: A reward function designed based on the output, quality, equipment energy consumption and production safety indicators of anhydrous ethanol is used to evaluate the results of actions; Policy: A rule for choosing an action in a given state, optimized through learning to maximize the cumulative reward.

3. The anhydrous ethanol synthesis production monitoring and control system according to claim 2, characterized in that: According to the goal of anhydrous ethanol synthesis, a reward function of the anhydrous ethanol synthesis decision model is designed. The reward function includes reward mechanisms such as output increase, quality optimization, equipment energy consumption reduction, production safety improvement and avoidance of production accidents. At the same time, penalties are set for synthesis failure or equipment failure.

4. The anhydrous ethanol synthesis production monitoring and control system according to claim 3, characterized in that: The reward function is designed in the following ways: Design a production reward function, and calculate the production reward value based on the deviation between the real-time production of anhydrous ethanol and the preset target production. The specific algorithm formula is: Among them, k production is the output reward coefficient. When the actual output is higher than the target output, R production is a positive reward; when the actual output is lower than the target output, R production Negative rewards; Design a quality reward function and calculate the quality reward value based on the deviation between the quality index of anhydrous ethanol and the preset quality standard. The specific algorithm formula is: Among them, k quality is the quality reward coefficient, n is the number of quality indicators, when the actual quality is better than the quality standard, R quality is a positive reward; when the actual quality is worse than the quality standard, R quality Negative rewards; Design an energy consumption penalty function, and calculate the energy consumption penalty value according to the deviation between the real-time energy consumption of the synthetic equipment and the preset energy consumption budget. The specific algorithm formula is: p energy =k energy × (actual energy consumption - energy consumption budget) Among them, k energy is the energy penalty coefficient. When the actual energy consumption is higher than the energy budget, P energy is a negative penalty; when the actual energy consumption is lower than the energy budget, P energy is zero or positive reward; Design a safety reward function and calculate the safety reward value based on the deviation between the safety index in the production process and the preset safety threshold. The specific algorithm formula is: Among them, k safegy is the security reward coefficient, m is the number of security indicators, and when the actual security indicator is better than the security threshold, R safety is a positive reward; when the actual safety index is worse than the safety threshold, R safety For negative punishment; Design an accident penalty function. When a safety accident occurs during the production process, calculate the accident penalty value based on the severity and impact range of the accident. The specific algorithm formula is: P accident =k accident ×(Accident Level) Among them, k accident is the accident penalty coefficient, and the accident level is quantified according to the severity and impact scope of the accident; Combining the above rewards and penalties, the total reward value is calculated. The specific algorithm formula is: R total =R production +R quality -P energy +R safety -P accident The total reward value P total As the input of the decision-making model in the intelligent control module, it guides the model to learn the optimal control strategy.

5. The anhydrous ethanol synthesis production monitoring and control system according to claim 4, characterized in that: The deep Q network DQN algorithm was selected to iteratively train the anhydrous ethanol synthesis decision model.

6. The anhydrous ethanol synthesis production monitoring and control system according to claim 5, characterized in that: The steps for training the anhydrous ethanol synthesis decision model include: Step 1: Initialize two deep neural networks with the same structure. One is an online updated network, which is used to calculate the Q value of the current strategy selection and perform iterative Q value updates and gradient descent back propagation. The other is a target network that periodically copies the weights of the former and is used to calculate the target Q value. Step 2: Set up an experience replay buffer to store the four-tuple data [state s t , action a t , reward r_t, new state s t+1 ] is stored in the playback buffer. When training the Q network, some data is randomly sampled from the playback buffer for training; Step 3: Use the following formula to update the Q value: Q(s,a)=r+γ*max a' Q target (s,a′) Among them, Q(s,a) represents the Q value of taking action a in state s, r represents the immediate reward, γ represents the discount factor, and Q target (s',a') represents the Q value of taking action a' in the new state s' calculated by the target network; Step 4: Copy the parameters in the online network to the target network; Step 5: Use the gradient descent method to update the network weights and calculate the gradient through the back-propagation algorithm to minimize the difference between the predicted Q value and the actual Q value; Step 6: Repeat steps 2 to 5 until the performance indicators of the model meet the requirements and the training of the anhydrous ethanol synthesis decision model is completed.

7. The anhydrous ethanol synthesis production monitoring and control system according to claim 6, characterized in that: In step 5, the formula for updating the parameters of the gradient descent algorithm is: Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

8. The anhydrous ethanol synthesis production monitoring and control system according to claim 6, characterized in that: During the model training process, the mean square error MSE algorithm is used to evaluate the performance of the model: For each state-action pair (s, a), calculate the error between the predicted Q value and the actual Q value and square it. Then average the squared errors of all state-action pairs to get the mean square error (MSE). The specific formula is as follows: Where N is the total number of state-action pairs sampled during the evaluation process, and Σ represents the summation operation. The smaller the MSE value, the closer the predicted Q value of the model is to the actual Q value, and the better the performance of the model. Conversely, the larger the MSE value, the worse the prediction ability of the model, and further optimization or retraining is needed.

9. The anhydrous ethanol synthesis production monitoring and control system according to claim 1, characterized in that: The performance of the anhydrous ethanol synthesis decision model in practical applications is evaluated by combining comprehensive simulation tests with field application verification.

10. The anhydrous ethanol synthesis production monitoring and control system according to claim 9, characterized in that: The method of comprehensive simulation testing and field application verification of the anhydrous ethanol synthesis decision model includes: Using computational chemistry simulation software and thermodynamic models, a highly simulated anhydrous ethanol synthesis simulation environment is constructed. This environment is used to accurately reflect the physical and chemical changes in the actual synthesis process, as well as the effects of different operating conditions on synthesis efficiency, energy consumption, and by-product generation. The anhydrous ethanol synthesis decision model to be evaluated is applied in a simulation environment, and the optimal synthesis scheme is generated according to the input raw material conditions, reaction parameters and equipment status, including the setting of reaction temperature, pressure, catalyst type and dosage, and reaction time; Adopt the recognized traditional control algorithm, conduct simulation tests under the same conditions, and generate the corresponding synthesis scheme; In the simulation environment, the model recommendation scheme and the traditional control algorithm scheme were executed respectively, and the anhydrous ethanol synthesis efficiency, equipment energy consumption, environmental emissions, and temperature and pressure control stability during the reaction process were recorded and compared. Based on the simulation test, a representative anhydrous ethanol production unit was selected for field application verification; according to the scheme in the simulation test, the model recommendation scheme and the traditional control algorithm scheme were implemented respectively, and actual operation data were collected; Combining the results of simulation tests with field application verification, a comprehensive evaluation of the anhydrous ethanol synthesis decision model was conducted.

Images