Production process control system for refined fuel oil

Abstract

The invention discloses a production process control system for refined fuel oil, which belongs to the technical field of production control and comprises a reactant control module, a fuel oil separation control module and a storage environment monitoring module. The reactant control module collects raw material conveying data and controls the reaction process of conveying the raw materials to the reaction kettle; the fuel oil separation control module obtains fuel oil separation characteristics, predicts the fuel oil separation effect and controls separation operation parameters of fuel oil. And the storage environment monitoring module collects storage environment data of the refined fuel oil to obtain an operation prediction state, and regulates and controls environment parameters according to a prediction result. According to the invention, the conveying flow control unit in the reactant control module is combined with the raw material flow data collected in real time to calculate the opening adjustment amount of the adjusting valve, so that the raw material conveying flow is accurately regulated and controlled, and the raw material can enter the reaction kettle at a stable flow meeting the preset requirement; the problem that in the prior art, the production process of refined fuel oil is low in effect is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of production control, and in particular relates to a production process control system for refined fuel oil. Background Art

[0002] In recent years, economic development has led to a continuous increase in energy demand. As an important energy product, refined fuel oil plays an indispensable role in transportation, industrial production and many areas of people's livelihood. Under the general trend of continuous optimization and adjustment of the energy structure, the quality and output requirements of refined fuel oil are becoming increasingly stringent. On the one hand, increasingly stringent environmental protection regulations have prompted manufacturers to accurately control various component indicators in fuel oil, such as sulfur content and aromatic content, in order to reduce the pollution caused by combustion emissions to the environment; on the other hand, the expectations of various industries for the stability of fuel oil performance are also constantly increasing, which means that the production process needs to be more precise and efficient to ensure that the product meets the diverse usage needs.

[0003] At present, the refining and blending process of fuel oil usually adopts acid-base refining, hydrogenation refining, adsorption refining and other methods to treat fuel oil to improve its quality. However, in the existing technology, due to the lack of effective coordination of various links, problems such as unstable raw material delivery flow, poor reaction temperature control, low fuel oil separation efficiency, and weak storage environment control are prone to occur when different processes are connected, resulting in low production efficiency, easy shutdown waiting for materials, or unstable product quality due to process fluctuations, and the problem of low efficiency in the production process of refined fuel oil. Summary of the invention

[0004] In view of the above problems, the present invention is proposed. Therefore, the purpose of the present invention is to provide a production process control system for refined fuel oil, which can achieve precise regulation of the raw material delivery flow rate, so that the raw materials can enter the reactor at a stable flow rate that meets the preset requirements, thereby ensuring the stable and accurate input of the raw materials.

[0005] To achieve the above object, the present invention provides the following technical solutions: A production process control system for refined fuel oil, comprising a reactant control module, a fuel oil separation control module, and a storage environment monitoring module, wherein: Reactant control module: used to collect raw material delivery data and control the reaction process of raw material delivery to the reactor based on the raw material delivery data; Fuel oil separation control module: used to obtain fuel oil separation characteristics, predict fuel oil separation effects according to the fuel oil separation characteristics, and control fuel oil separation operation parameters according to the fuel oil separation effects; Storage environment monitoring module: used to collect storage environment data of refined fuel oil, derive operation prediction status based on storage environment data, and adjust environmental parameters according to the prediction results.

[0006] Preferably, the reactant control module includes a delivery flow control unit, a reaction temperature control unit, and a reactant concentration monitoring unit; The conveying flow control unit is used to monitor the real-time flow of raw material conveying and control the valve opening according to the real-time flow; The reaction temperature control unit is used to collect the real-time temperature of the reactor and the real-time flow rate of raw materials in real time, and dynamically adjust the temperature control power accordingly; The reactant concentration monitoring unit is used to monitor the reactant concentration changes in real time.

[0007] Preferably, controlling the valve opening according to the real-time flow rate comprises the following steps: When the raw materials begin to flow into the reactor, the flow sensor collects the real-time flow data of the raw materials in real time. ; Get the preset raw material flow setting value , calculate the adjustment amount of the regulating valve opening through the calculation formula of the adjustment amount, the formula is: ; in, Indicates the control adjustment amount of the regulating valve opening, is the proportionality coefficient, is the integration time constant, is the differential time constant, represents the flow deviation value at the current time t, , Indicates that in the interval Any time within The flow deviation value, , , Indicates time Raw material flow data; Based on the calculated value, adjust the opening of the regulating valve in real time, so that the raw material flow rate quickly approaches and stabilizes at a value close to within the allowable deviation range.

[0008] Preferably, dynamically adjusting the temperature to control power includes the following steps: Define reaction temperature deviation and temperature change rate as input variables, reaction temperature deviation is the difference between actual reaction temperature and expected reaction temperature; Define the temperature control power adjustment as the output variable; Obtaining a fuzzy set combination based on input variables and a fuzzy set of temperature control power adjustment amount from a database; A multilayer perceptron network structure with three hidden layers is selected as the neural network model, and the number of output layer nodes is set to 1; The historical reaction temperature deviation, historical temperature change rate, and historical temperature control power adjustment amount are normalized and divided into a training set and a test set; The neural network is trained using the training set. The weights and biases of the neural network are initialized to random values. In each training iteration, a small batch of data from the training set is input into the network, the predicted output is calculated, the mean square error loss function is calculated based on the predicted output and the actual label, the gradient is calculated through the back-propagation algorithm, and the weights and biases are updated according to the gradient and the learning rate. The trained neural network is tested using a test set to obtain a prediction model for the temperature control power adjustment amount; The real-time reaction temperature deviation and temperature change rate in the reactor are collected in real time, and input into the temperature control power adjustment prediction model to obtain the temperature control power adjustment at the current moment, and adjust the temperature control equipment power accordingly.

[0009] Preferably, the fuel oil separation control module includes a fuel oil separation effect prediction unit and a separation operation control unit; A fuel oil separation effect prediction unit is used to obtain historical separation data, extract fuel oil separation characteristics from it, and build a fuel oil separation effect prediction model based on the fuel oil separation characteristics, obtain real-time separation data, and output the fuel oil separation effect through the fuel oil separation effect prediction model; The separation operation control unit is used to control the separation operation parameters of the fuel oil according to the separation effect of the fuel oil. The separation operation parameters include adjusting the reflux ratio and the heating power.

[0010] Preferably, extracting the fuel oil separation characteristics comprises the following steps: Obtain historical separation data and construct a separation data set based on it ,in, represents the input data of the i-th sample, represents the corresponding output label, i is the sample number, and N is the number of samples in the data set; The CNN convolutional neural network is used to learn the processed data set, and the convolution kernel in the convolution layer is used to input data. Slide scan according to the set step size and perform convolution operation at each position , generate feature maps, where Represents the output result corresponding to the jth feature map in the lth layer of the CNN convolutional neural network, is the weight of the input data located in row m and column n in the j-th feature map convolution kernel, is the input data of the m+i row and n+j column of the l-1th layer, is the bias term of the jth feature map of the lth layer, l represents the index of the layer, j represents the index of the feature map, m and n represent the positions in the convolution kernel, the pooling layer processes the feature map by downsampling operation, the fully connected layer integrates and maps the features after convolution and pooling, and uses the ReLU activation function to perform nonlinear transformation on the neuron output; The CNN convolutional neural network is trained using the stochastic gradient descent algorithm to output the fuel oil separation feature vector F, where , k is the dimension of the feature vector.

[0011] Preferably, constructing a fuel oil separation effect prediction model according to the fuel oil separation characteristics comprises the following steps: The extracted fuel oil separation feature vector F is used as input to the fuel oil separation effect preparation model, and a fuel oil separation effect prediction model reflecting the separation effect and the operating parameters is trained; The fuel oil separation effect preparation model is: ; in, represents the predicted separation effect indicator vector, represents the fuel oil separation feature vector at the current time t, represents the operation parameter vector at the current time t, represents the external variable vector at the current time t, e is a natural constant, is the width parameter of the vth radial basis function, v is the center index of the radial basis function, V is the total number of centers of the radial basis function, Is , , A long vector concatenated from the input vectors. is the center of the vth radial basis function, is the weight of the vth radial basis function; Obtain the fuel oil separation feature vector F, the corresponding operation parameter vector U and the historical actual separation effect index vector Y in the historical separation data to construct a training data set and a test data set; Training the fuel oil separation effect preparation model based on the training data set; The trained fuel oil separation effect preparation model was tested using the test data set to obtain the fuel oil separation effect prediction model.

[0012] Preferably, controlling the separation operating parameters of the fuel oil according to the fuel oil separation effect comprises the following steps: Obtain real-time separation data and obtain the predicted separation effect index vector through the fuel oil separation effect prediction model ; Separation efficiency and product purity The weighted sum J of is taken as the optimization target, that is: ; in , is the corresponding weight coefficient, and the genetic algorithm is used to find the optimal reflux ratio that makes the optimization objective function reach the optimal value. and heating power ; The optimal reflux ratio calculated and heating power , automatically adjust the flow rate of the reflux pump and the power of the heating equipment.

[0013] Preferably, the storage environment monitoring module includes an environment monitoring unit and an environment prediction and control unit; An environmental monitoring unit, used to monitor the storage environment data of the refined fuel oil storage unit in real time, including liquid level, temperature, and pressure; The environment prediction and control unit is used to obtain a storage environment state prediction value based on the collected storage environment data, and adjust the environmental parameters according to the storage environment state prediction value.

[0014] Preferably, obtaining the operation prediction state according to the storage environment data and adjusting the environmental parameters according to the prediction result includes the following steps: Obtain historical operation data to form a data set ,in is the input feature vector of the hth sample, is the storage environment state corresponding to the hth sample, M is the number of samples, and h is the sample number; For the dataset Preprocess the data set and Divide into training set and test set; The prediction model is trained using the training set, and the trained prediction model is tested using the test set to obtain a storage environment prediction model; Acquire real-time storage environment data, input it into the storage environment prediction model, and output the storage environment status prediction value; When the difference between the predicted value of the storage environment state and the storage environment reference state value set in the database exceeds the allowable difference, the environment control is started to control the storage environment state value within the allowable difference, otherwise no processing is performed.

[0015] The beneficial effects of the present invention are: The present invention uses the raw material flow data collected in real time to accurately calculate the adjustment amount of the regulating valve opening according to the PID control algorithm through the delivery flow control unit in the reactant control module, thereby achieving precise regulation of the raw material delivery flow, so that the raw material can enter the reactor at a stable flow rate that meets the preset requirements, thereby ensuring the stability and accuracy of the raw material input amount, avoiding the instability of the reaction process and the impact on product quality due to flow fluctuations, and effectively solving the problem of low efficiency of the production process of refined fuel oil in the prior art.

[0016] The reaction temperature control unit of the reactant control module of the present invention adopts the reaction temperature deviation and the temperature change rate as input variables, combines the neural network with the multi-layer perceptron network structure, and constructs a temperature control power adjustment amount prediction model through training and testing based on fuzzy sets and mean square error loss functions. The temperature control power adjustment amount can be accurately output in real time according to the temperature change in actual production, so as to realize dynamic and precise adjustment of the reaction temperature, improve the accuracy and timeliness of the reaction temperature control, reduce the side reactions caused by abnormal temperature fluctuations, and improve the yield and quality of refined fuel oil.

[0017] The present invention uses a CNN convolutional neural network to extract fuel oil separation characteristics through a fuel oil separation effect prediction unit in a fuel oil separation control module, and constructs a separation effect prediction model based on these characteristics. The separation operation control unit uses a weighted sum of separation efficiency and product purity as an optimization target according to the prediction model, and determines the optimal operating parameters through a genetic algorithm to achieve refined and intelligent control of the fuel oil separation process, so that the separation operation can be dynamically adjusted according to actual conditions, thereby improving the efficiency of fuel oil separation and the purity of each fraction product, and optimizing resource utilization and product quality in the entire separation link.

[0018] The present invention first collects historical operation data through a storage environment monitoring module to construct and train a storage environment prediction model, then obtains storage environment data in real time and uses the model to derive a prediction value, and adjusts environmental parameters based on the comparison between the prediction result and the reference state value, thereby realizing real-time monitoring and precise regulation of the storage environment of refined fuel oil, ensuring that the storage environment is always in a suitable state, extending the quality stability period of refined fuel oil in the storage stage, reducing the risk of oil deterioration due to a poor storage environment, and improving the reliability of overall storage management. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the structure of the system of the present invention. DETAILED DESCRIPTION

[0020] This embodiment solves the problem of low efficiency of the production process of refined fuel oil in the prior art by providing a refined fuel oil production process control system. The reactant control module precisely controls the raw material flow and reaction temperature, and the fuel oil separation control module optimizes the separation operation parameters. Finally, the storage environment monitoring module finely controls the storage environment. The technical effects of ensuring stable raw material delivery, accurately controlling the reaction process, improving the fuel oil separation quality, and ensuring a suitable storage environment are achieved, thereby improving the overall quality and efficiency of refined fuel oil production.

[0021] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0022] like Figure 1 As shown, this embodiment provides a refined fuel oil production process control system, the system includes: a reactant control module, a fuel oil separation control module, and a storage environment monitoring module; Among them, the reactant control module is used to collect raw material delivery data and control the reaction process of delivering the raw materials to the reactor based on the raw material delivery data; Fuel oil separation control module: used to obtain fuel oil separation characteristics, predict fuel oil separation effects according to the fuel oil separation characteristics, and control fuel oil separation operation parameters according to the fuel oil separation effects; Storage environment monitoring module: used to collect storage environment data of refined fuel oil, derive operation prediction status based on storage environment data, and adjust environmental parameters according to the prediction results.

[0023] The reactant control module includes a delivery flow control unit, a reaction temperature control unit, and a reactant concentration monitoring unit; The conveying flow control unit is used to monitor the real-time flow of raw material conveying and control the valve opening according to the real-time flow; The reaction temperature control unit is used to collect the real-time temperature of the reactor and the real-time flow rate of raw materials in real time, and dynamically adjust the temperature control power accordingly; The reactant concentration monitoring unit is used to monitor the reactant concentration changes in real time.

[0024] Controlling the valve opening according to real-time flow includes the following steps: When the raw materials begin to flow into the reactor, the flow sensor collects the real-time flow data of the raw materials in real time. ; Get the preset raw material flow setting value , calculate the adjustment amount of the regulating valve opening through the calculation formula of the adjustment amount, the formula is: ; in, Indicates the control adjustment amount of the regulating valve opening, is the proportionality coefficient, is the integration time constant, is the differential time constant, represents the flow deviation value at the current time t, , Indicates that in the interval Any time within The flow deviation value, , , Indicates time Raw material flow data; Based on the calculated value, adjust the opening of the regulating valve in real time, so that the raw material flow rate quickly approaches and stabilizes at a value close to within the allowable deviation range.

[0025] In this embodiment, the PID control algorithm is used. By accurately calculating the adjustment amount of the regulating valve opening, it can respond to the real-time changes in the raw material flow in real time and accurately, so that the raw material flow quickly and stably approaches the preset value, overcoming the flow fluctuations caused by various internal and external factors, and ensuring the stability and accuracy of the raw material transportation link.

[0026] Dynamically adjusting temperature to control power includes the following steps: Define reaction temperature deviation and temperature change rate as input variables, reaction temperature deviation is the difference between actual reaction temperature and expected reaction temperature; Define the temperature control power adjustment as the output variable; Obtaining a fuzzy set combination based on input variables and a fuzzy set of temperature control power adjustment amount from a database; A multilayer perceptron network structure with three hidden layers is selected as the neural network model, and the number of output layer nodes is set to 1; The historical reaction temperature deviation, historical temperature change rate, and historical temperature control power adjustment are normalized and divided into a training set and a test set; The neural network is trained using the training set. The weights and biases of the neural network are initialized to random values. In each training iteration, a small batch of data from the training set is input into the network, the predicted output is calculated, the mean square error loss function is calculated based on the predicted output and the actual label, the gradient is calculated through the back-propagation algorithm, and the weights and biases are updated according to the gradient and the learning rate. The trained neural network is tested using a test set to obtain a prediction model for the temperature control power adjustment amount; The real-time reaction temperature deviation and temperature change rate in the reactor are collected in real time, and input into the temperature control power adjustment prediction model to obtain the temperature control power adjustment at the current moment, and adjust the temperature control equipment power accordingly.

[0027] In this embodiment, an advanced control strategy combining fuzzy sets and neural networks is adopted. By learning a large amount of historical data and inputting real-time data, the deviation of the reaction temperature and its changing trend can be captured, and the temperature control power adjustment amount can be accurately output to achieve real-time, dynamic and refined control of the reaction temperature, avoiding the lag and inaccuracy of traditional temperature control methods, effectively reducing side reactions caused by abnormal temperature, and improving the product quality and production efficiency of refined fuel oil.

[0028] The fuel oil separation control module includes a fuel oil separation effect prediction unit and a separation operation control unit; A fuel oil separation effect prediction unit is used to obtain historical separation data, extract fuel oil separation characteristics from it, and build a fuel oil separation effect prediction model based on the fuel oil separation characteristics, obtain real-time separation data, and output the fuel oil separation effect through the fuel oil separation effect prediction model; The separation operation control unit is used to control the separation operation parameters of the fuel oil according to the separation effect of the fuel oil. The separation operation parameters include adjusting the reflux ratio and the heating power.

[0029] In this embodiment, a prediction model is constructed based on historical and real-time data to know the fuel oil separation effect in advance, and the separation operation control unit adjusts key separation operation parameters such as reflux ratio and heating power according to the prediction results. Through the cooperation of the two, intelligent regulation of the fuel oil separation process is realized, the separation efficiency and product purity are improved, and it is ensured that fuel oil of different fractions can be separated according to high quality standards, thereby improving resource utilization efficiency and overall product quality.

[0030] Extracting the fuel oil separation characteristics includes the following steps: Obtain historical separation data and construct a separation data set based on it ,in, represents the input data of the i-th sample, represents the corresponding output label, i is the sample number, and N is the number of samples in the data set; The CNN convolutional neural network is used to learn the processed data set, and the convolution kernel in the convolution layer is used to input data. Slide scan according to the set step size and perform convolution operation at each position , generate feature maps, where Represents the output result corresponding to the jth feature map in the lth layer of the CNN convolutional neural network, is the weight of the input data located in row m and column n in the j-th feature map convolution kernel, is the input data of the m+i row and n+j column of the l-1th layer, is the bias term of the jth feature map of the lth layer, l represents the index of the layer, j represents the index of the feature map, m and n represent the position in the convolution kernel, the pooling layer processes the feature map by downsampling operation, the fully connected layer integrates and maps the features after convolution and pooling, and uses the ReLU activation function to perform nonlinear transformation on the neuron output; The CNN convolutional neural network is trained using the stochastic gradient descent algorithm to output the fuel oil separation feature vector F, where , k is the dimension of the feature vector, That is, the features processed by CNN convolutional neural network.

[0031] In this embodiment, by performing convolution, pooling, and full connection operations on a large amount of historical separation data, vector information that can reflect the essential characteristics of fuel oil separation is mined.

[0032] Constructing a fuel oil separation effect prediction model based on fuel oil separation characteristics includes the following steps: The extracted fuel oil separation feature vector F is used as input to the fuel oil separation effect preparation model, and a fuel oil separation effect prediction model reflecting the separation effect and the operating parameters is trained; The fuel oil separation effect preparation model is: ; in, represents the predicted separation effect indicator vector, represents the fuel oil separation feature vector at the current time t, represents the operation parameter vector at the current time t, represents the external variable vector at the current time t, e is a natural constant, is the width parameter of the vth radial basis function, v is the center index of the radial basis function, V is the total number of centers of the radial basis function, Is , , A long vector concatenated from the input vectors. is the center of the vth radial basis function, is the weight of the vth radial basis function; Obtain the fuel oil separation feature vector F, the corresponding operation parameter vector U and the historical actual separation effect index vector Y in the historical separation data to construct a training data set and a test data set; Training the fuel oil separation effect preparation model based on the training data set; The trained fuel oil separation effect preparation model was tested using the test data set to obtain the fuel oil separation effect prediction model.

[0033] In this embodiment, by establishing a fuel oil separation effect prediction model, the separation effect can be predicted in advance based on the fuel oil separation characteristics obtained in real time, so that the operator can adjust the separation operation parameters in time according to the prediction results to optimize the separation process.

[0034] Controlling the separation operating parameters of the fuel oil according to the fuel oil separation effect includes the following steps: Obtain real-time separation data and obtain the predicted separation effect index vector through the fuel oil separation effect prediction model ; Separation efficiency and product purity The weighted sum J of is taken as the optimization target, that is: ; in , is the corresponding weight coefficient. Through genetic algorithm, we can find the optimal reflux ratio that makes the optimization objective function reach the optimal value. and heating power ; The optimal reflux ratio calculated and heating power , automatically adjust the flow rate of the reflux pump and the power of the heating equipment.

[0035] In this embodiment, by setting the weighted sum J of separation efficiency and product purity as the optimization target, and combining the genetic algorithm to find the optimal reflux ratio and heating power, it is possible to weigh multiple aspects of the separation effect as a whole, find the best combination of operating parameters, and achieve precise optimization of the fuel oil separation process, thereby improving the separation efficiency, improving the purity of the product, and ensuring that the fuel oil separation link produces high-quality products to the greatest extent, while also improving the effective utilization of resources.

[0036] The storage environment monitoring module includes: an environment monitoring unit and an environment prediction and control unit; An environmental monitoring unit is used to monitor the storage environment data of the refined fuel oil storage unit in real time, and the storage environment data includes: liquid level, temperature, and pressure; The environment prediction and control unit is used to obtain a storage environment state prediction value based on the collected storage environment data, and adjust the environmental parameters according to the storage environment state prediction value.

[0037] In this embodiment, the environmental prediction and control unit performs state prediction and implements regulation based on these data, realizing real-time monitoring and active intervention of the storage environment, preventing the quality degradation of refined fuel oil due to changes in environmental factors in advance, ensuring the stability and quality reliability of oil products during the storage stage, and extending the storage period of oil products.

[0038] Determining the operation prediction status based on the storage environment data and adjusting the environment parameters based on the prediction results includes the following steps: Obtain historical operation data to form a data set ,in is the input feature vector of the hth sample, is the storage environment state corresponding to the hth sample, M is the number of samples, and h is the sample number; For the dataset Preprocess the data set and Divide into training set and test set; The prediction model is trained using the training set, and the trained prediction model is tested using the test set to obtain a storage environment prediction model; After obtaining the real-time storage environment data and normalizing it, it forms the input feature vector , ,in Is the real-time liquid level, is the real-time temperature, It is the real-time pressure, which is input into the storage environment prediction model and outputs the storage environment status prediction value ,in is the predicted liquid level value, is the predicted temperature value, is the predicted pressure value; When the difference between the predicted value of the storage environment state and the storage environment reference state value set in the database exceeds the allowable difference, the environment control is started to control the storage environment state value within the allowable difference, otherwise no processing is performed.

[0039] In this embodiment, by using historical data to build and train a storage environment prediction model, and then combining it with real-time data for prediction and regulation, it is possible to predict the changing trend of the storage environment in advance, take corresponding control measures in a timely manner, and accurately maintain the storage environment in a suitable state, thereby avoiding oil deterioration and loss caused by abnormal environmental parameters, ensuring that the quality of refined fuel oil is not damaged in the storage link, and effectively preserving the final results of the entire production process.

[0040] This embodiment also includes: a memory on which a computer processing program is stored; and a processor for executing the computer processing program in the memory to implement the above-mentioned system and algorithm calculation.

Claims

1. A production process control system for refined fuel oil, characterized in that: It includes reactant control module, fuel oil separation control module and storage environment monitoring module, among which: Reactant control module: used to collect raw material delivery data and control the reaction process of raw material delivery to the reactor based on the raw material delivery data; Fuel oil separation control module: used to obtain fuel oil separation characteristics, predict fuel oil separation effects according to the fuel oil separation characteristics, and control fuel oil separation operation parameters according to the fuel oil separation effects; Storage environment monitoring module: used to collect storage environment data of refined fuel oil, derive operation prediction status based on storage environment data, and adjust environmental parameters according to the prediction results.

2. A refined fuel oil production process control system as claimed in claim 1, characterized in that: The reactant control module includes a delivery flow control unit, a reaction temperature control unit, and a reactant concentration monitoring unit; The conveying flow control unit is used to monitor the real-time flow of raw material conveying and control the valve opening according to the real-time flow; The reaction temperature control unit is used to collect the real-time temperature of the reactor and the real-time flow rate of raw materials in real time, and dynamically adjust the temperature control power accordingly; The reactant concentration monitoring unit is used to monitor the reactant concentration changes in real time.

3. A refined fuel oil production process control system as claimed in claim 2, characterized in that: Controlling the valve opening according to real-time flow includes the following steps: When the raw materials begin to flow into the reactor, the flow sensor collects the real-time flow data of the raw materials in real time. ; Get the preset raw material flow setting value , calculate the adjustment amount of the regulating valve opening through the calculation formula of the adjustment amount, the formula is: ； in, Indicates the control adjustment amount of the regulating valve opening, is the proportionality coefficient, is the integration time constant, is the differential time constant, represents the flow deviation value at the current time t, , Indicates that in the interval Any time within The flow deviation value, , , Indicates time Raw material flow data; Based on the calculated value, adjust the opening of the regulating valve in real time, so that the raw material flow rate quickly approaches and stabilizes at a value close to within the allowable deviation range.

4. A refined fuel oil production process control system as claimed in claim 2, characterized in that: Dynamically adjusting temperature to control power includes the following steps: Define reaction temperature deviation and temperature change rate as input variables, reaction temperature deviation is the difference between actual reaction temperature and expected reaction temperature; Define the temperature control power adjustment amount as the output variable; Obtaining a fuzzy set combination based on input variables and a fuzzy set of temperature control power adjustment amount from a database; A multilayer perceptron network structure with three hidden layers is selected as the neural network model, and the number of output layer nodes is set to 1; The historical reaction temperature deviation, historical temperature change rate, and historical temperature control power adjustment amount are normalized and divided into a training set and a test set; The neural network is trained using the training set. The weights and biases of the neural network are initialized to random values. In each training iteration, a small batch of data from the training set is input into the network, the predicted output is calculated, the mean square error loss function is calculated based on the predicted output and the actual label, the gradient is calculated through the back-propagation algorithm, and the weights and biases are updated according to the gradient and the learning rate. The trained neural network is tested using a test set to obtain a prediction model for the temperature control power adjustment amount; The real-time reaction temperature deviation and temperature change rate in the reactor are collected in real time, and input into the temperature control power adjustment prediction model to obtain the temperature control power adjustment at the current moment, and adjust the temperature control equipment power accordingly.

5. A refined fuel oil production process control system as claimed in claim 1, characterized in that: The fuel oil separation control module includes a fuel oil separation effect prediction unit and a separation operation control unit; A fuel oil separation effect prediction unit is used to obtain historical separation data, extract fuel oil separation characteristics from it, and build a fuel oil separation effect prediction model based on the fuel oil separation characteristics, obtain real-time separation data, and output the fuel oil separation effect through the fuel oil separation effect prediction model; The separation operation control unit is used to control the separation operation parameters of the fuel oil according to the separation effect of the fuel oil. The separation operation parameters include adjusting the reflux ratio and the heating power.

6. A refined fuel oil production process control system as claimed in claim 5, characterized in that: Extracting the fuel oil separation characteristics includes the following steps: Obtain historical separation data and construct a separation data set based on it ,in, represents the input data of the i-th sample, represents the corresponding output label, i is the sample number, and N is the number of samples in the data set; The CNN convolutional neural network is used to learn the processed data set, and the convolution kernel in the convolution layer is used to input data. Slide scan according to the set step size and perform convolution operation at each position , generate feature maps, where Represents the output result corresponding to the jth feature map in the lth layer of the CNN convolutional neural network, is the weight of the input data located in row m and column n in the j-th feature map convolution kernel, is the input data of the m+i row and n+j column of the l-1th layer, is the bias term of the jth feature map of the lth layer, l represents the index of the layer, j represents the index of the feature map, m and n represent the positions in the convolution kernel, the pooling layer processes the feature map by downsampling operation, the fully connected layer integrates and maps the features after convolution and pooling, and uses the ReLU activation function to perform nonlinear transformation on the neuron output; The CNN convolutional neural network is trained using the stochastic gradient descent algorithm to output the fuel oil separation feature vector F, where , k is the dimension of the feature vector.

7. A refined fuel oil production process control system as claimed in claim 6, characterized in that: Constructing a fuel oil separation effect prediction model based on fuel oil separation characteristics includes the following steps: The extracted fuel oil separation feature vector F is used as input to the fuel oil separation effect preparation model, and a fuel oil separation effect prediction model reflecting the separation effect and the operating parameters is trained; The fuel oil separation effect preparation model is: ； in, represents the predicted separation effect indicator vector, represents the fuel oil separation feature vector at the current time t, represents the operation parameter vector at the current time t, represents the external variable vector at the current time t, e is a natural constant, is the width parameter of the vth radial basis function, v is the center index of the radial basis function, V is the total number of centers of the radial basis function, Is , , A long vector concatenated from the input vectors. is the center of the vth radial basis function, is the weight of the vth radial basis function; Obtain the fuel oil separation feature vector F, the corresponding operation parameter vector U and the historical actual separation effect index vector Y in the historical separation data to construct a training data set and a test data set; Training the fuel oil separation effect preparation model based on the training data set; The trained fuel oil separation effect preparation model was tested using the test data set to obtain the fuel oil separation effect prediction model.

8. A refined fuel oil production process control system as claimed in claim 1, characterized in that: Controlling the separation operating parameters of the fuel oil according to the fuel oil separation effect includes the following steps: Obtain real-time separation data and obtain the predicted separation effect index vector through the fuel oil separation effect prediction model ; Separation efficiency and product purity The weighted sum J of is taken as the optimization target, that is: ； in , is the corresponding weight coefficient, and the genetic algorithm is used to find the optimal reflux ratio that makes the optimization objective function reach the optimal value. and heating power ; The optimal reflux ratio calculated and heating power , automatically adjust the flow rate of the reflux pump and the power of the heating equipment.

9. A refined fuel oil production process control system as claimed in claim 1, characterized in that: The storage environment monitoring module includes an environment monitoring unit and an environment prediction and control unit; An environmental monitoring unit, used to monitor the storage environment data of the refined fuel oil storage unit in real time, including liquid level, temperature, and pressure; The environment prediction and control unit is used to obtain a storage environment state prediction value based on the collected storage environment data, and adjust the environmental parameters according to the storage environment state prediction value.

10. A refined fuel oil production process control system as claimed in claim 9, characterized in that: Determining the operation prediction status based on the storage environment data and adjusting the environment parameters based on the prediction results includes the following steps: Obtain historical operation data to form a data set ,in is the input feature vector of the hth sample, is the storage environment state corresponding to the hth sample, M is the number of samples, and h is the sample number; For the dataset Preprocess the data set and Divide into training set and test set; The prediction model is trained using the training set, and the trained prediction model is tested using the test set to obtain a storage environment prediction model; Acquire real-time storage environment data, input it into the storage environment prediction model, and output the storage environment status prediction value; When the difference between the predicted value of the storage environment state and the storage environment reference state value set in the database exceeds the allowable difference, the environment control is started to control the storage environment state value within the allowable difference, otherwise no processing is performed.

Images