WO2024011747A1 - 一种用于六氟丁二烯制备的智能化冷却液循环控制系统 - Google Patents

一种用于六氟丁二烯制备的智能化冷却液循环控制系统 Download PDF

Info

Publication number
WO2024011747A1
WO2024011747A1 PCT/CN2022/119265 CN2022119265W WO2024011747A1 WO 2024011747 A1 WO2024011747 A1 WO 2024011747A1 CN 2022119265 W CN2022119265 W CN 2022119265W WO 2024011747 A1 WO2024011747 A1 WO 2024011747A1
Authority
WO
WIPO (PCT)
Prior art keywords
temperature
vector
corrected
feature vector
condenser
Prior art date
Application number
PCT/CN2022/119265
Other languages
English (en)
French (fr)
Inventor
胡进军
张鸿铨
朱军伟
吴雪瑛
张奎
Original Assignee
福建省杭氟电子材料有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 福建省杭氟电子材料有限公司 filed Critical 福建省杭氟电子材料有限公司
Publication of WO2024011747A1 publication Critical patent/WO2024011747A1/zh

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F28—HEAT EXCHANGE IN GENERAL
    • F28F—DETAILS OF HEAT-EXCHANGE AND HEAT-TRANSFER APPARATUS, OF GENERAL APPLICATION
    • F28F27/00—Control arrangements or safety devices specially adapted for heat-exchange or heat-transfer apparatus
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • the present invention relates to the field of intelligent control, and more specifically, to an intelligent coolant circulation control system for the preparation of hexafluorobutadiene and a control method thereof.
  • Hexafluorobutadiene has many industrial applications. It is not only a monomer for the preparation of various fluorine-containing polymer elastic materials, polyhexafluorobutadiene, but also a green and efficient dry etching gas with extremely low greenhouse effect. , which has higher etching selectivity than traditional plasma etching gases. In the preparation process of hexafluorobutadiene, it is an important step to purify the obtained product to meet the purity requirements of industrial grade or electronic grade.
  • Embodiments of the present application provide an intelligent coolant circulation control system and a control method for the preparation of hexafluorobutadiene, which extract the water-washed material in the condenser through an artificial intelligence control method.
  • the dynamic characteristic information of the temperature value of hexafluorobutadiene at the outlet of the condenser and the flow rate value of the cooling liquid is combined with the dynamic characteristics of the three in time series to perform the cooling Intelligent control of liquid flow rate to improve the effective utilization of energy.
  • an intelligent coolant circulation control system for the preparation of hexafluorobutadiene, which includes: a temperature data acquisition module for obtaining data from a first temperature sensor deployed in the condenser The collected temperature values of the water-washed material at multiple predetermined time points in the condenser and the temperature values of hexafluorobutadiene collected by the second temperature sensor deployed at the outlet of the condenser.
  • Temperature values at multiple predetermined time points at the outlet a cooling liquid circulation flow rate data acquisition module, used to obtain the flow rate values of the coolant at the multiple predetermined time points collected by a flow meter deployed in the condenser;
  • the first temperature data encoding module is used to pass the temperature values of the washed material at multiple predetermined time points in the condenser through a first time sequence encoder including a one-dimensional convolution layer to obtain the temperature of the object to be cooled.
  • Time-series feature vector a second temperature data encoding module, configured to pass the temperature values of the hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser through a second time-series encoding including a one-dimensional convolution layer
  • the flow rate data encoding module is used to pass the flow rate values of the cooling liquid at the plurality of predetermined time points through a third temporal encoder including a one-dimensional convolution layer to obtain the flow rate feature vector
  • the feature distribution A correction module configured to perform feature distribution correction on the temperature time series feature vector of the object to be cooled, the outlet temperature feature vector and the flow velocity feature vector respectively to obtain the corrected temperature time series feature vector of the object to be cooled and the corrected outlet temperature feature.
  • a Bayesian-like fusion module used to use a Bayesian-like probability model to fuse the corrected temperature time series feature vector of the object to be cooled, the corrected outlet temperature feature vector and the correction
  • the posterior flow velocity feature vector is used to obtain the posterior probability vector
  • the loop control result generation module is used to pass the posterior probability vector through the classifier to obtain the classification result, and the classification result is used to represent the coolant flow rate response at the current time point. increase or should decrease.
  • a control method for an intelligent coolant circulation control system for hexafluorobutadiene preparation which includes: obtaining the washed water collected by a first temperature sensor deployed in the condenser. The temperature values of the material at multiple predetermined time points in the condenser and the multiple temperature values of hexafluorobutadiene at the outlet of the condenser collected by the second temperature sensor deployed at the outlet of the condenser.
  • the temperature values at predetermined time points are passed through a first temporal encoder including a one-dimensional convolution layer to obtain a temperature temporal feature vector of the object to be cooled;
  • the temperature value at the time point is passed through a second temporal encoder including a one-dimensional convolution layer to obtain the outlet temperature feature vector;
  • the flow rate values of the coolant at the plurality of predetermined time points are passed through a third temporal encoder including a one-dimensional convolution layer.
  • the posterior probability vector is passed through a classifier to obtain a classification result, which is used to indicate that the coolant flow rate at the current point in time should be increased or decreased.
  • the intelligent coolant circulation control system and its control method for the preparation of hexafluorobutadiene use a convolutional neural network model based on deep learning to control the temperature inside the calciner.
  • the relevant features are dynamically extracted in real time, and the structural change characteristics and internal heat distribution characteristics of the calcined product are deeply mined, and then the temperature of the calciner is intelligently adjusted by combining the time series characteristic information of the three. To optimize energy while ensuring the quality of the final lithium fluoride product.
  • Figure 1 is an application scenario diagram of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • Figure 2 is a block diagram of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • Figure 3 is a block diagram of the first temperature data encoding module in the intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • Figure 4 is a flow chart of a control method of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • Figure 5 is a schematic structural diagram of a control method of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • hexafluorobutadiene has many industrial applications. It is not only a monomer for the preparation of various fluorine-containing polymer elastic materials, polyhexafluorobutadiene, but also a green and environmentally friendly material with extremely low greenhouse effect. A highly efficient dry etching gas that has higher etching selectivity than traditional plasma etching gases. In the preparation process of hexafluorobutadiene, it is an important step to purify the obtained product to meet the purity requirements of industrial grade or electronic grade.
  • the hexafluorobutadiene raw material is sequentially subjected to water washing, condensation, primary adsorption, secondary adsorption, primary distillation and secondary distillation to obtain a purified product of hexafluorobutadiene.
  • the adsorbent used in the first-stage adsorption includes silica gel and lithium hydroxide and boron oxide supported on the silica gel;
  • the adsorbent used in the second-stage adsorption is type A zeolite and/or type Y zeolite.
  • the hexafluorobutadiene raw material enters a water washing tower and is washed with water.
  • the flow rate of the hexafluorobutadiene entering the water washing tower is 2 to 10kg/h.
  • the flow rate of water is (1 ⁇ 3) ⁇ 10m3/h, and the temperature of the water washing tower still liquid is 20 ⁇ 30°C.
  • the water-washed material enters a condenser for condensation.
  • the outlet temperature of the hexafluorobutadiene at the condenser is 5 to 10°C.
  • the inventor of the present application found that the final purification accuracy and accuracy of hexafluorobutadiene can be improved by controlling the coolant flow rate of the coolant circulation system so that the outlet temperature of hexafluorobutadiene in the condenser is maintained within a predetermined range.
  • Product consistency the inventor of the present application found that the final purification accuracy and accuracy of hexafluorobutadiene can be improved by controlling the coolant flow rate of the coolant circulation system so that the outlet temperature of hexafluorobutadiene in the condenser is maintained within a predetermined range.
  • the temperature value of the washed material in the condenser is collected through a first temperature sensor deployed in the condenser and a second temperature sensor deployed at the outlet of the condenser collects hexafluorofluoride
  • the temperature value of butadiene at the outlet of the condenser is used to dynamically monitor the temperature, and the flow rate value of the cooling liquid is collected through a flow meter deployed in the condenser to conduct the flow rate dynamics of the cooling liquid. control.
  • the outlet temperature of hexafluorobutadiene at the condenser can be maintained within a predetermined range based on the dynamic control of the coolant flow rate, thereby improving the final purification accuracy and product consistency of hexafluorobutadiene.
  • the temperature values of the washed material at multiple predetermined time points in the condenser collected by the first temperature sensor deployed in the condenser and the temperature values collected by the first temperature sensor deployed in the condenser are obtained.
  • the second temperature sensor at the outlet of the condenser collects the temperature values of hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser, and obtains the temperature values from the flow meter deployed in the condenser.
  • the temperature values of the water-washed material at multiple predetermined time points in the condenser and the location of the hexafluorobutadiene are The temperature values at multiple predetermined time points at the outlet of the condenser are encoded through a time series encoder including a one-dimensional convolution layer to obtain a temperature time series feature vector of the object to be cooled and an outlet temperature feature vector.
  • the temporal encoder consists of alternately arranged fully connected layers and one-dimensional convolutional layers, which extract the dynamic correlation of the temperature value in the temporal dimension through one-dimensional convolutional coding.
  • Features and high-dimensional latent features of the temperature value are extracted through fully connected encoding.
  • the flow rate values of the coolant at multiple predetermined time points are passed through a one-dimensional convolution layer. Encoding is performed in the third time sequence encoder to extract the local implicit dynamic correlation feature information of the flow rate values of the coolant at the plurality of predetermined time points, thereby obtaining a flow rate feature vector.
  • the purpose of the technical solution of this application is based on new evidence, that is, on the premise of the change in the temperature value of the washed material in the condenser. , update the prior probability to get the posterior probability. Then, according to the Bayesian formula, the posterior probability is the prior probability multiplied by the probability of the event divided by the probability of the evidence. Therefore, in the technical solution of this application, a Bayesian-like probability model is used to fuse the corrected The coolant temperature time series feature vector, the corrected outlet temperature feature vector and the corrected flow velocity feature vector are used to obtain a posterior probability vector.
  • the formula of the Bayesian-like probability model is expressed as:
  • each position in the corrected flow velocity feature vector is the value of each position in the corrected flow velocity feature vector, and are respectively the value of each position in the corrected outlet temperature feature vector and the corrected object temperature time series feature vector, and is the value of each position in the posterior probability vector.
  • this type of condition boundary constraint performs boundary constraints on features through structural understanding of feature values and the class conditions to which they belong based on information rules, thereby avoiding the feature value set being in the classification target domain due to the out-of-distribution feature value of the set.
  • the excessive fragmentation of the decision-making area enables the feature distribution represented by the feature vector to obtain a robust conditional class boundary, thereby realizing the constraints of each feature distribution to the probabilistic classification target and improving the performance obtained by the Bayesian probability model.
  • the classification effect of the posterior probability feature distribution thereby improving the accuracy of classification.
  • this application proposes an intelligent coolant circulation control system for the preparation of hexafluorobutadiene, which includes: a temperature data acquisition module, used to obtain the temperature data collected by the first temperature sensor deployed in the condenser.
  • a cooling liquid circulation flow rate data acquisition module used to obtain the flow rate values of the coolant at the multiple predetermined time points collected by a flow meter deployed in the condenser;
  • first Temperature data encoding module used to pass the temperature values of the washed material at multiple predetermined time points in the condenser through a first temporal encoder including a one-dimensional convolution layer to obtain the temperature temporal characteristics of the object to be cooled Vector;
  • a second temperature data encoding module configured to pass the
  • FIG. 1 illustrates an application scenario diagram of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • a first temperature sensor for example, T1 as shown in Figure 1
  • the condenser for example, D as shown in Figure 1
  • the temperature values of the washed material for example, M as shown in Figure 1
  • T2 for example, T2 as shown in Figure 1
  • the condenser collects the flow rate values of the cooling liquid (eg, L as illustrated in FIG. 1 ) at the plurality of predetermined time points. Then, the obtained temperature values of the water-washed material at multiple predetermined time points in the condenser, the temperature values of the hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser, and The flow rate values of the coolant at the plurality of predetermined time points are input into a server deployed with an intelligent coolant circulation control algorithm for hexafluorobutadiene preparation (for example, server S as shown in Figure 1), where , the server can use an intelligent coolant circulation control algorithm for the preparation of hexafluorobutadiene to control the temperature values of the washed material at multiple predetermined time points in the condenser, the hexafluorobutadiene
  • Figure 2 illustrates a block diagram of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • the intelligent coolant circulation control system 200 for the preparation of hexafluorobutadiene according to the embodiment of the present application includes: a temperature data acquisition module 210, used to obtain the temperature data obtained from the first temperature sensor deployed in the condenser.
  • the temperature values of the water-washed material collected by the temperature sensor at multiple predetermined time points in the condenser and the temperature values of hexafluorobutadiene collected by the second temperature sensor deployed at the outlet of the condenser during the condensation The temperature values at multiple predetermined time points at the outlet of the condenser; the coolant circulation flow rate data acquisition module 220 is used to obtain the temperature values of the coolant at the multiple predetermined time points collected by the flow meter deployed in the condenser.
  • Flow rate value; the first temperature data encoding module 230 is used to obtain the temperature values of the washed material at multiple predetermined time points in the condenser through a first temporal encoder including a one-dimensional convolution layer.
  • the temperature time series feature vector of the object to be cooled; the second temperature data encoding module 240 is used to pass the temperature values of the hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser through a one-dimensional convolution layer.
  • the second temporal encoder is used to obtain the outlet temperature feature vector;
  • the flow rate data encoding module 250 is used to pass the flow rate values of the coolant at the plurality of predetermined time points through a third temporal encoder including a one-dimensional convolution layer to obtain Flow velocity feature vector;
  • feature distribution correction module 260 used to perform feature distribution correction on the temperature time series feature vector of the object to be cooled, the outlet temperature feature vector and the flow velocity feature vector respectively to obtain the corrected temperature time series characteristics of the object to be cooled vector, the corrected outlet temperature feature vector and the corrected flow velocity feature vector;
  • the Bayesian-like fusion module 270 is used to use a Bayesian-like probability model to fuse the corrected temperature time series feature vector of the object to be cooled, the corrected The outlet temperature feature vector and the corrected flow velocity feature vector are used to obtain a posterior probability vector;
  • the loop control result generation module 280 is used to pass the posterior probability vector through a classifier to obtain a classification result, and the classification result is Yu indicates that the coolant flow
  • the temperature data acquisition module 210 and the coolant circulation flow rate data acquisition module 220 are used to obtain the water-washed material collected by the first temperature sensor deployed in the condenser. Temperature values at multiple predetermined time points within the condenser and multiple predetermined times of hexafluorobutadiene at the outlet of the condenser collected by a second temperature sensor deployed at the outlet of the condenser temperature values at the plurality of predetermined time points, and obtain the flow rate values of the cooling liquid at the plurality of predetermined time points collected by the flow rate meter deployed in the condenser.
  • the coolant flow rate of the coolant circulation system is controlled so that the outlet temperature of hexafluorobutadiene in the condenser is maintained within a predetermined range, the final hexafluorobutadiene can be improved. Diene purification precision and product consistency. Therefore, in the technical solution of the present application, the temperature values of the washed material at multiple predetermined time points in the condenser are collected through the first temperature sensor deployed in the condenser and deployed at the outlet of the condenser.
  • the second temperature sensor at the outlet of the condenser collects the temperature values of hexafluorobutadiene at multiple predetermined time points to perform dynamic monitoring of the temperature, and collects multiple temperature values through the flow meter deployed in the condenser.
  • the flow rate value of the cooling liquid at a predetermined time point is used to dynamically control the flow rate of the cooling liquid. In this way, the outlet temperature of hexafluorobutadiene at the condenser can be maintained within a predetermined range based on the dynamic control of the coolant flow rate, thereby improving the final purification accuracy and product consistency of hexafluorobutadiene. .
  • the first temperature data encoding module 230 and the second temperature data encoding module 240 are used to store the washed material in the condenser for multiple predetermined times.
  • the temperature value of the point is passed through the first temporal encoder including a one-dimensional convolution layer to obtain the temperature temporal feature vector of the object to be cooled, and the hexafluorobutadiene is placed at multiple predetermined time points at the outlet of the condenser.
  • the temperature value is passed through a second temporal encoder containing a one-dimensional convolutional layer to obtain the outlet temperature feature vector.
  • the temperature values of the washed material at multiple predetermined time points in the condenser and The temperature values of the hexafluorobutadiene at a plurality of predetermined time points at the outlet of the condenser are respectively encoded through a first temporal encoder and a second temporal encoder including a one-dimensional convolution layer, to The time series characteristic vector of the temperature of the object to be cooled and the characteristic vector of the outlet temperature are obtained respectively.
  • first timing encoder and the second timing encoder have the same network structure.
  • first temporal encoder and the second temporal encoder are composed of alternately arranged fully connected layers and one-dimensional convolutional layers, which extract the temperature value through one-dimensional convolutional coding. Dynamic correlation features in the time series dimension and high-dimensional latent features of the temperature values are extracted through fully connected encoding.
  • the first temperature data encoding module includes: arranging the temperature values of the washed material at multiple predetermined time points in the condenser according to the time dimension into the first A temperature input vector; use the fully connected layer of the temporal encoder to fully connect the first temperature input vector with the following formula to extract the high-dimensional hidden features of the eigenvalues of each position in the first temperature input vector.
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel.
  • the second temperature data encoding module includes: a second temperature input vector construction unit for converting the polyethylene hexafluorobutadiene at the outlet of the condenser into The temperature values at a predetermined time point are arranged into a second temperature input vector according to the time dimension; a second fully connected encoding unit is used to use the fully connected layer of the timing encoder to fully perform a full connection on the second temperature input vector using the following formula: Concatenate coding to extract high-dimensional hidden features of the eigenvalues of each position in the second temperature input vector, where the formula is: where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents matrix multiplication; a second one-dimensional convolution coding unit is used to perform one-dimensional convolution coding on the second temperature input vector using the following formula using the one-dimensional convolution layer of the temporal encoder to extract the first High-dimensional implicit correlation features between the eigenvalues of each
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel.
  • Figure 3 illustrates a block diagram of the first temperature data encoding module in the intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • the first temperature data encoding module 230 includes: a first temperature input vector construction unit 231, used to convert the temperatures of the washed materials at multiple predetermined time points in the condenser.
  • the values are arranged into the first temperature input vector according to the time dimension; the first fully connected encoding unit 232 is used to use the fully connected layer of the temporal encoder to perform fully connected encoding on the first temperature input vector using the following formula to extract The high-dimensional implicit features of the eigenvalues of each position in the first temperature input vector, where the formula is: where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents matrix multiplication; the first one-dimensional convolution encoding unit 233 is used to use the one-dimensional convolution layer of the temporal encoder to perform one-dimensional convolution encoding on the first temperature input vector according to the following formula to extract the High-dimensional implicit correlation features between the eigenvalues of each position in the first temperature input vector, where the formula is:
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel.
  • the flow rate data encoding module 250 is used to pass the flow rate values of the cooling liquid at the plurality of predetermined time points through a third temporal encoder including a one-dimensional convolution layer to obtain the flow rate. Feature vector.
  • a third temporal encoder including a one-dimensional convolution layer to obtain the flow rate.
  • the flow rate value of the liquid is encoded through a third temporal encoder including a one-dimensional convolution layer to extract the local implicit dynamic correlation feature information of the flow rate value of the cooling liquid at the plurality of predetermined time points, thereby obtaining the flow rate feature. vector.
  • the first timing encoder, the second timing encoder and the third timing encoder have the same network structure.
  • the temporal encoder and the third temporal encoder are composed of alternately arranged fully connected layers and one-dimensional convolutional layers.
  • the flow rate data encoding module includes: a flow rate input vector construction unit for converting the hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser.
  • the temperature values are arranged into flow velocity input vectors according to the time dimension;
  • the flow velocity fully connected encoding unit is used to use the fully connected layer of the temporal encoder to perform fully connected encoding on the flow velocity input vector with the following formula to extract the flow velocity input
  • a flow velocity one-dimensional convolution encoding unit used to use the one-dimensional convolution layer of the temporal encoder to perform one-dimensional convolution encoding on the flow velocity input vector with the following formula to extract the flow velocity input vector
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel.
  • the feature distribution correction module 260 is used to perform feature distribution correction on the temperature time series feature vector of the object to be cooled, the outlet temperature feature vector and the flow velocity feature vector respectively to obtain The corrected temperature time series feature vector of the object to be cooled, the corrected outlet temperature feature vector and the corrected flow velocity feature vector. It should be understood that before using the Bayesian probability model, the temperature time series feature vector of the object to be cooled, the flow rate feature vector and the outlet temperature feature vector need to be mapped to the probability space first.
  • class condition boundary constraints are first applied to the temperature time series feature vector of the object to be cooled, the flow rate feature vector and the outlet temperature feature vector, so as to The corrected temperature time series feature vector of the object to be cooled, the corrected outlet temperature feature vector and the corrected flow velocity feature vector are obtained.
  • the feature distribution correction module is further configured to: perform the following calculation on the temperature time series feature vector of the object to be cooled, the outlet temperature feature vector and the flow rate feature vector respectively.
  • the characteristic distribution is corrected to obtain the corrected temperature time series feature vector of the object to be cooled, the corrected outlet temperature feature vector and the corrected flow velocity feature vector; wherein, the formula is:
  • the class condition boundary constraint is implemented by structural understanding of the feature values and the class conditions to which they belong based on information rules, thus avoiding the feature value set being out-of-distribution feature values of the set.
  • the resulting excessive fragmentation of the decision-making area within the classification target domain enables the feature distribution represented by the feature vector to obtain a robust conditional class boundary, thereby realizing the constraints of each feature distribution to the probabilistic classification target, and improving the The classification effect of the posterior probability feature distribution obtained by the Bayesian probability model further improves the accuracy of classification.
  • the Bayesian-like fusion module 270 is used to use a Bayesian-like probability model to fuse the corrected temperature time series feature vector of the object to be cooled and the corrected outlet temperature feature. vector and the corrected flow velocity feature vector to obtain the posterior probability vector.
  • the flow rate characteristic vector is a priori probability
  • the purpose of the technical solution of this application is based on new evidence, that is, on the premise of the change in the temperature value of the washed material in the condenser. , update the prior probability to get the posterior probability.
  • the posterior probability is the prior probability multiplied by the probability of the event divided by the probability of the evidence. Therefore, in the technical solution of this application, a Bayesian-like probability model is used to fuse the corrected The coolant temperature time series feature vector, the corrected outlet temperature feature vector and the corrected flow velocity feature vector are used to obtain a posterior probability vector.
  • the Bayesian-like fusion module is further used to: use the Bayesian-like probability model to fuse the corrected temperature time series feature vector of the object to be cooled and the The corrected outlet temperature feature vector and the corrected flow velocity feature vector are used to obtain the posterior probability vector; wherein, the formula is:
  • each position in the corrected flow velocity feature vector is the value of each position in the corrected flow velocity feature vector, and are respectively the value of each position in the corrected outlet temperature feature vector and the corrected object temperature time series feature vector, and is the value of each position in the posterior probability vector.
  • the cycle control result generation module 280 is used to pass the posterior probability vector through a classifier to obtain a classification result, and the classification result is used to represent the coolant flow rate at the current point in time. should be increased or should be decreased. More specifically, in the embodiment of the present application, the loop control result generation module is further configured to: use the classifier to process the posterior probability vector with the following formula to obtain the classification result, where: The above formula is: softmax ⁇ (W n ,B n ):...:(W 1 ,B 1 )
  • the intelligent coolant circulation control system 200 for the preparation of hexafluorobutadiene is clarified, which uses an artificial intelligence control method to extract the washed material in the condenser
  • the dynamic characteristic information of the temperature value within the condenser, the temperature value of hexafluorobutadiene at the outlet of the condenser and the flow rate value of the cooling liquid is combined, and the dynamic characteristics of the three in time series are combined to carry out the description. Intelligent control of coolant flow rate to improve the effective utilization of energy.
  • the intelligent coolant circulation control system 200 for the preparation of hexafluorobutadiene can be implemented in various terminal equipment, such as the intelligent coolant for the preparation of hexafluorobutadiene. Servers for loop control algorithms, etc.
  • the intelligent coolant circulation control system 200 for hexafluorobutadiene preparation according to the embodiment of the present application can be integrated into the terminal device as a software module and/or hardware module.
  • the intelligent coolant circulation control system 200 for hexafluorobutadiene preparation can be a software module in the operating system of the terminal equipment, or can be an application program developed for the terminal equipment; of course , the intelligent coolant circulation control system 200 for hexafluorobutadiene preparation can also be one of the many hardware modules of the terminal equipment.
  • the intelligent coolant circulation control system 200 for hexafluorobutadiene preparation and the terminal equipment can also be separate devices, and the intelligent coolant circulation control system 200 for hexafluorobutadiene preparation can also be separate devices.
  • the chemical coolant circulation control system 200 can be connected to the terminal device through a wired and/or wireless network, and transmit interactive information according to an agreed data format.
  • Figure 4 illustrates a flow chart of a control method of an intelligent coolant circulation control system for hexafluorobutadiene preparation.
  • the control method of the intelligent coolant circulation control system for the preparation of hexafluorobutadiene according to the embodiment of the present application includes the step: S110, obtaining the data collected by the first temperature sensor deployed in the condenser.
  • the temperature values at multiple predetermined time points in the condenser are passed through the first time series encoder including a one-dimensional convolution layer to obtain the temperature time series feature vector of the object to be cooled; S140, add the hexafluorobutadiene at the
  • the temperature values at multiple predetermined time points at the outlet of the condenser are passed through a second temporal encoder including a one-dimensional convolution layer to obtain the outlet temperature feature vector; S150, convert the flow rates of the cooling liquid at the multiple predetermined time points.
  • the value is passed through the third time series encoder including a one-dimensional convolution layer to obtain the flow rate feature vector; S160, perform feature distribution correction on the temperature time series feature vector of the object to be cooled, the outlet temperature feature vector and the flow speed feature vector respectively.
  • S170 use a Bayesian probability model to fuse the corrected temperature time series feature vector of the object to be cooled, the The corrected outlet temperature feature vector and the corrected flow velocity feature vector are used to obtain a posterior probability vector; and, S180, pass the posterior probability vector through a classifier to obtain a classification result, and the classification result is used to represent the current time point.
  • the coolant flow rate should be increased or decreased.
  • Figure 5 illustrates a schematic structural diagram of a control method of an intelligent coolant circulation control system for hexafluorobutadiene preparation according to an embodiment of the present application.
  • the obtained washed material is placed in the condenser
  • the temperature values of multiple predetermined time points (for example, P1 as illustrated in Figure 5) are passed through a first temporal encoder (for example, E1 as illustrated in Figure 5) including a one-dimensional convolution layer to obtain the values to be cooled
  • the material temperature time series feature vector for example, VF1 as shown in Figure 5
  • the obtained temperature values of the hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser for example, P2 as illustrated in Figure 5
  • a second temporal encoder e.g., E2 as illustrated in Figure 5 including a one-dimensional convolution
  • the temperature values of the washed material at multiple predetermined time points in the condenser collected by the first temperature sensor deployed in the condenser and the temperature values collected by the first temperature sensor deployed in the condenser are obtained.
  • the second temperature sensor at the outlet of the condenser collects the temperature values of hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser, and obtains the temperature values collected by the flow meter deployed in the condenser.
  • the coolant flow rate of the coolant circulation system is controlled so that the outlet temperature of hexafluorobutadiene at the condenser is maintained within a predetermined range, the final purification accuracy of hexafluorobutadiene can be improved. and product consistency. Therefore, in the technical solution of the present application, the temperature values of the washed material at multiple predetermined time points in the condenser are collected through the first temperature sensor deployed in the condenser and deployed at the outlet of the condenser.
  • the second temperature sensor at the outlet of the condenser collects the temperature values of hexafluorobutadiene at multiple predetermined time points to perform dynamic monitoring of the temperature, and collects multiple temperature values through the flow meter deployed in the condenser.
  • the flow rate value of the cooling liquid at a predetermined time point is used to dynamically control the flow rate of the cooling liquid. In this way, the outlet temperature of hexafluorobutadiene at the condenser can be maintained within a predetermined range based on the dynamic control of the coolant flow rate, thereby improving the final purification accuracy and product consistency of hexafluorobutadiene. .
  • the temperature values of the washed material at multiple predetermined time points in the condenser are passed through a first temporal encoder including a one-dimensional convolution layer to obtain the temperature to be determined.
  • Coolant temperature time series feature vector and pass the temperature values of the hexafluorobutadiene at multiple predetermined time points at the outlet of the condenser through a second time series encoder including a one-dimensional convolution layer to obtain the outlet temperature Feature vector.
  • the temperature values of the washed material at multiple predetermined time points in the condenser and The temperature values of the hexafluorobutadiene at a plurality of predetermined time points at the outlet of the condenser are respectively encoded through a first temporal encoder and a second temporal encoder including a one-dimensional convolution layer, to The time series characteristic vector of the temperature of the object to be cooled and the characteristic vector of the outlet temperature are obtained respectively.
  • the first timing encoder and the second timing encoder have the same network structure.
  • the first temporal encoder and the second temporal encoder are composed of alternately arranged fully connected layers and one-dimensional convolutional layers, which extract the temperature value through one-dimensional convolutional coding. Dynamic correlation features in the time series dimension and high-dimensional latent features of the temperature values are extracted through fully connected encoding.
  • step S150 the flow rate values of the cooling liquid at the plurality of predetermined time points are passed through a third temporal encoder including a one-dimensional convolution layer to obtain a flow rate feature vector.
  • a third temporal encoder including a one-dimensional convolution layer to extract the local implicit dynamic correlation feature information of the flow rate value of the cooling liquid at the plurality of predetermined time points, thereby obtaining the flow rate feature. vector.
  • the first timing encoder, the second timing encoder and the third timing encoder have the same network structure.
  • the temporal encoder and the third temporal encoder are composed of alternately arranged fully connected layers and one-dimensional convolutional layers.
  • step S160 feature distribution correction is performed on the temperature time series feature vector of the object to be cooled, the outlet temperature feature vector and the flow velocity feature vector respectively to obtain the corrected temperature time series feature vector of the object to be cooled, the corrected The rear outlet temperature feature vector and the corrected flow velocity feature vector.
  • the temperature time series feature vector of the object to be cooled, the flow rate feature vector and the outlet temperature feature vector need to be mapped to the probability space first.
  • mapping with a linear mapping scheme such as linearization, it is impossible to constrain the feature distribution expressed by the eigenvalue set of the feature vector to the probabilistic classification target, which will affect the posterior probability vector calculated using the Bayesian probability model. Classification effect.
  • class condition boundary constraints are first applied to the temperature time series feature vector of the object to be cooled, the flow rate feature vector and the outlet temperature feature vector, so as to The corrected temperature time series feature vector of the object to be cooled, the corrected outlet temperature feature vector and the corrected flow velocity feature vector are obtained.
  • a Bayesian-like probability model is used to fuse the corrected temperature time series feature vector of the object to be cooled, the corrected outlet temperature feature vector, and the corrected flow velocity feature vector to obtain a posteriori probability vector.
  • the flow rate characteristic vector is a priori probability
  • the purpose of the technical solution of this application is based on new evidence, that is, on the premise of the change in the temperature value of the washed material in the condenser. , update the prior probability to get the posterior probability.
  • the posterior probability is the prior probability multiplied by the probability of the event divided by the probability of the evidence. Therefore, in the technical solution of this application, a Bayesian-like probability model is used to fuse the corrected The coolant temperature time series feature vector, the corrected outlet temperature feature vector and the corrected flow velocity feature vector are used to obtain a posterior probability vector.
  • the posterior probability vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate that the coolant flow rate at the current point in time should be increased or decreased.
  • the classifier is used to process the posterior probability vector with the following formula to obtain the classification result, where the formula is: softmax ⁇ (W n ,B n ) :...:(W 1 ,B 1 )
  • the control method of the intelligent coolant circulation control system for the preparation of hexafluorobutadiene is clarified, which uses artificial intelligence control methods to extract the washed materials in the
  • the dynamic characteristic information of the temperature value in the condenser, the temperature value of hexafluorobutadiene at the outlet of the condenser and the flow rate value of the cooling liquid is combined with the dynamic characteristics of the three in time series. Intelligent control of the coolant flow rate to improve the effective utilization of energy.
  • each component or each step can be decomposed and/or recombined. These decompositions and/or recombinations shall be considered equivalent versions of this application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Mechanical Engineering (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Thermal Sciences (AREA)
  • Feedback Control In General (AREA)

Abstract

本申请涉及智能控制的领域，其具体地公开了一种用于六氟丁二烯制备的智能化冷却液循环控制系统，其通过人工智能控制方法，来提取出水洗后的物料在所述冷凝器内的温度值、六氟丁二烯在所述冷凝器的出口处的温度值以及冷却液的流速值这三者的动态特征信息，并融合这三者在时序上的动态特征来进行所述冷却液流速的智能控制，以提高能量的有效利用率。

Description

一种用于六氟丁二烯制备的智能化冷却液循环控制系统 技术领域
本发明涉及智能控制的领域，且更为具体地，涉及一种用于六氟丁二烯制备的智能化冷却液循环控制系统及其控制方法。
背景技术
六氟丁二烯在工业上有多方面的应用，不仅是制备多种含氟高分子弹性材料聚六氟丁二烯的单体，还是一种温室效应极低，绿色环保的高效干蚀刻气体，其比传统等离子蚀刻气体的蚀刻选择性更高。在六氟丁二烯的制备过程中，对所获得的产物进行提纯以满足工业级或者电子级的纯度要求是重要环节。
纯化气体的常用方法包括低温精馏法、物理吸收法、化学转化法、选择吸附法、冷凝、冷冻法及膜分离法，也有一些厂商将各种纯化方案进行结合来提高纯化效果。通过冷却液来进行冷凝是大多数六氟丁二烯纯化方案中都会用到的处理手段，冷凝效果会影响最终的纯化效果，也就是，冷却液的循环控制系统的控制会影响最终的纯化精度，另一方面，冷却液的循环控制系统的运行本身会耗费能量，采用适当的方式进行控制有利于提高能量有效利用率，以节能环保。
因此，期待一种优化的用于六氟丁二烯制备的冷却液循环控制系统。
发明内容
为了解决上述技术问题，提出了本申请。本申请的实施例提供了一种用于六氟丁二烯制备的智能化冷却液循环控制系统及其控制方法，其通过人工智能控制方法，来提取出水洗后的物料在所述冷凝器内的温度值、六氟丁二烯在所述冷凝器的出口处的温度值以及冷却液的流速值这三者的动态特征信息，并融合这三者在时序上的动态特征来进行所述冷却液流速的智能控制，以提高能量的有效利用率。
根据本申请的一个方面，提供了一种用于六氟丁二烯制备的智能化冷却液循环控制系统，其包括：温度数据采集模块，用于获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；冷却液循环流速数据采集模块，用于获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；第一温度数据编码模块，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；第二温度数据编码模块，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；流速数据编码模块，用于将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；特征分布校正模块，用于对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；类贝叶斯融合模块，用于使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及循环控制结果生成模块，用于将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
根据本申请的另一方面，一种用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法，其包括：获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及
将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
与现有技术相比，本申请提供的用于六氟丁二烯制备的智能化冷却液循环控制系统及其控制方法，其通过基于深度学习的卷积神经网络模型来对于煅烧器内部的温度关联特征进行实时动态地 提取，以及对于煅烧产物的结构变化特征以及内部的热量分布特征进行深层次地挖掘，进而结合这三者在时序上的特征信息来智能地调整所述煅烧器的温度，以在优化能源的同时保证最终氟化锂的成品质量。
附图说明
通过结合附图对本申请实施例进行更详细的描述，本申请的上述以及其他目的、特征和优势将变得更加明显。附图用来提供对本申请实施例的进一步理解，并且构成说明书的一部分，与本申请实施例一起用于解释本申请，并不构成对本申请的限制。在附图中，相同的参考标号通常代表相同部件或步骤。
图1为根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的应用场景图。
图2为根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的框图。
图3为根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统中第一温度数据编码模块的框图。
图4为根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法的流程图。
图5为根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法的架构示意图。
具体实施方式
下面，将参考附图详细地描述根据本申请的示例实施例。显然，所描述的实施例仅仅是本申请的一部分实施例，而不是本申请的全部实施例，应理解，本申请不受这里描述的示例实施例的限制。
场景概述
如前所述，六氟丁二烯在工业上有多方面的应用，不仅是制备多种含氟高分子弹性材料聚六氟丁二烯的单体，还是一种温室效应极低，绿色环保的高效干蚀刻气体，其比传统等离子蚀刻气体的蚀刻选择性更高。在六氟丁二烯的制备过程中，对所获得的产物进行提纯以满足工业级或者电子级的纯度要求是重要环节。
纯化气体的常用方法包括低温精馏法、物理吸收法、化学转化法、选择吸附法、冷凝、冷冻法及膜分离法，也有一些厂商将各种纯化方案进行结合来提高纯化效果。通过冷却液来进行冷凝是大多数六氟丁二烯纯化方案中都会用到的处理手段，冷凝效果会影响最终的纯化效果，也就是，冷却液的循环控制系统的控制会影响最终的纯化精度，另一方面，冷却液的循环控制系统的运行本身会耗费能量，采用适当的方式进行控制有利于提高能量有效利用率，以节能环保。因此，期待一种优化的用于六氟丁二烯制备的冷却液循环控制系统。
在现有技术中，将六氟丁二烯原料依次进行水洗、冷凝、一级吸附、二级吸附、一级精馏和二级精馏，得到六氟丁二烯提纯产品。这里，所述一级吸附所使用的吸附剂包括硅胶和负载在硅胶上的氢氧化锂和氧化硼；所述二级吸附所使用的吸附剂为A型沸石和/或Y型沸石。
在六氟丁二烯制备的过程中，首先，将所述六氟丁二烯原料进入水洗塔，进行水洗，这里，所述六氟丁二烯进入水洗塔的流量为2～10kg/h，水的流量为(1～3)×10m3/h，所述水洗塔釜液温度为20～30℃。然后，将所述水洗后的物料进入冷凝器进行冷凝，这里，所述六氟丁二烯在冷凝器的出口温度为5～10℃。
基于此，本申请发明人发现通过对于冷却液循环系统的冷却液流速进行控制以使得六氟丁二烯在冷凝器的出口温度保持在预定范围内可提高最终六氟丁二烯的纯化精度和产品一致性。因此，在本申请中，通过部署于冷凝器内的第一温度传感器采集水洗后的物料在所述冷凝器内的温度值以及部署于所述冷凝器的出口处的第二温度传感器采集六氟丁二烯在所述冷凝器的出口处的温度值来进行温度的动态监控，并且通过部署于所述冷凝器内的流速计采集所述冷却液的流速值以进行所述冷却液的流速动态控制。通过这样的方式，可以基于对所述冷却液流速的动态控制来使得六氟丁二烯在冷凝器的出口温度保持在预定范围内，进而提高最终六氟丁二烯的纯化精度和产品一致性。
具体地，在本申请的技术方案中，首先，获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值，并且获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值。然后，应可以理解，考虑到所述水洗后的物料在所述冷凝器内的温度值以及六氟丁二烯在所述冷凝器的出口处的温度值在时间维度上都具有动态性的规律，因此为了更为充分地挖掘出这种动态性的变化特征信息，将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值分别通过包含一维卷积层的时序编码器中进行编码处理，以分别得到待冷却物温度时序特征向量和出口温度特征向量。特别地，在一个具体示例中，所述时序编码器都由交替排列的全连接 层和一维卷积层组成，其通过一维卷积编码提取出所述温度值在时序维度上的动态关联特征和通过全连接编码提取所述温度值的高维隐含特征。
并且，对于所述冷却液的流速值，考虑到其在时间维度上也存在着动态性的规律特征，因此，将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器中进行编码，以提取出所述多个预定时间点的冷却液的流速值的局部隐含动态关联特征信息，从而得到流速特征向量。
应可以理解，考虑到该所述流速特征向量为先验概率，本申请的技术方案的目的是在新的证据，即在所述水洗后的物料在所述冷凝器内的温度值变化前提下，更新先验概率以得到后验概率。那么，根据贝叶斯公式，后验概率为先验概率乘以事件的概率除以证据的概率，因此，在本申请的技术方案中，使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量。在一个具体示例中，所述类贝叶斯概率模型的公式表示为：
Figure PCTCN2022119265-appb-000001
其中
Figure PCTCN2022119265-appb-000002
是所述校正后流速特征向量中的每个位置的值，
Figure PCTCN2022119265-appb-000003
和
Figure PCTCN2022119265-appb-000004
分别是所述校正后出口温度特征向量和所述校正后待冷却物温度时序特征向量中的每个位置的值，而
Figure PCTCN2022119265-appb-000005
是所述后验概率向量中的每个位置的值。
但是，在使用贝叶斯概率模型之前，需要首先将所述待冷却物温度时序特征向量、所述流速特征向量和所述出口温度特征向量映射到概率空间，在通过最大值归一化等线性映射方案进行映射时，无法实现对特征向量的特征值集合所表达的特征分布到概率化的分类目标的约束，从而会影响利用贝叶斯概率模型计算得到的后验概率向量的分类效果。
因此，在使用贝叶斯概率模型之前，首先对所述待冷却物温度时序特征向量、所述流速特征向量和所述出口温度特征向量进行类条件边界约束，具体为：
Figure PCTCN2022119265-appb-000006
Figure PCTCN2022119265-appb-000007
Figure PCTCN2022119265-appb-000008
其中
Figure PCTCN2022119265-appb-000009
和
Figure PCTCN2022119265-appb-000010
分别是所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量的第i个位置的特征值，
Figure PCTCN2022119265-appb-000011
和
Figure PCTCN2022119265-appb-000012
分别是所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量的第i个位置的特征值。
这里，该类条件边界约束通过对特征值及其所属的类条件进行基于信息规则的结构理解来进行特征的边界约束，从而避免了特征值集合由于集合的分布外特征值而导致在分类目标域内的决策区域的过度碎片化，使得特征向量表示的特征分布获得了稳健的条件化的类边界，从而实现了各个特征分布到概率化的分类目标的约束，提高了贝叶斯概率模型所获得的后验概率特征分布的分类效果，进而提高了分类的准确性。
基于此，本申请提出了一种用于六氟丁二烯制备的智能化冷却液循环控制系统，其包括：温度数据采集模块，用于获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；冷却液循环流速数据采集模块，用于获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；第一温度数据编码模块，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；第二温度数据编码模块，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；流速数据编码模块，用于将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；特征分布校正模块，用于对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；类贝叶斯融合模块，用于使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及，循环控制结果生成模块，用于将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
图1图示了根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的应用场景图。如图1所示，在该应用场景中，首先，通过部署于冷凝器(例如，如图1中所示意的D)内的第一温度传感器(例如，如图1中所示意的T1)采集水洗后的物料(例如，如图1中所示意的M)在所 述冷凝器内的多个预定时间点的温度值以及通过部署于所述冷凝器的出口处的第二温度传感器(例如，如图1中所示意的T2)采集六氟丁二烯(例如，如图1中所示意的H)在所述冷凝器的出口处的多个预定时间点的温度值，并且由部署于所述冷凝器内的流速计(例如，如图1中所示意的E)采集所述多个预定时间点的冷却液(例如，如图1中所示意的L)的流速值。然后，将获得的所述水洗后的物料在冷凝器内的多个预定时间点的温度值、所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值以及所述多个预定时间点的冷却液的流速值输入至部署有用于六氟丁二烯制备的智能化冷却液循环控制算法的服务器中(例如，如图1中所示意的服务器S)，其中，所述服务器能够以用于六氟丁二烯制备的智能化冷却液循环控制算法对所述水洗后的物料在冷凝器内的多个预定时间点的温度值、所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值以及所述多个预定时间点的冷却液的流速值进行处理，以生成用于表示当前时间点的冷却液流速应增大或应减小的分类结果。
在介绍了本申请的基本原理之后，下面将参考附图来具体介绍本申请的各种非限制性实施例。
示例性系统
图2图示了根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的框图。如图2所示，根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统200，包括：温度数据采集模块210，用于获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；冷却液循环流速数据采集模块220，用于获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；第一温度数据编码模块230，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；第二温度数据编码模块240，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；流速数据编码模块250，用于将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；特征分布校正模块260，用于对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；类贝叶斯融合模块270，用于使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及，循环控制结果生成模块280，用于将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
具体地，在本申请实施例中，所述温度数据采集模块210和所述冷却液循环流速数据采集模块220，用于获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值，并获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值。如前所述，应可以理解，考虑到若通过对于冷却液循环系统的冷却液流速进行控制，以使得六氟丁二烯在冷凝器的出口温度保持在预定范围内，可提高最终六氟丁二烯的纯化精度和产品一致性。因此，在本申请的技术方案中，通过部署于冷凝器内的第一温度传感器采集水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及部署于所述冷凝器的出口处的第二温度传感器采集六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值来进行温度的动态监控，并且通过部署于所述冷凝器内的流速计采集多个预定时间点的所述冷却液的流速值以进行所述冷却液的流速动态控制。通过这样的方式，可以基于对所述冷却液流速的动态控制来使得六氟丁二烯在冷凝器的出口温度保持在预定范围内，进而提高最终六氟丁二烯的纯化精度和产品一致性。
具体地，在本申请实施例中，所述第一温度数据编码模块230和所述第二温度数据编码模块240，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量，并将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量。应可以理解，考虑到所述水洗后的物料在所述冷凝器内的温度值以及所述六氟丁二烯在所述冷凝器的出口处的温度值在时间维度上都具有动态性的规律，因此为了更为充分地挖掘出这种动态性的变化特征信息，在本申请的技术方案中，进一步将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值分别通过包含一维卷积层的第一时序编码器和第二时序编码器中进行编码处理，以分别得到待冷却物温度时序特征向量和出口温度特征向量。特别地，所述第一时序编码器和所述第二时序编码器具有相同的网络结构。在一个具体示例中，所述第一时序编码器和所述第二时序编码器都由交替排列的全连接层和一维卷积层组成，其通过一维卷积编码提取出所述温度值在时序维度上的动态关联特征和通过全连接编码提取所述温度值的高维隐含 特征。
更具体地，在本申请实施例中，所述第一温度数据编码模块，包括：将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值按照时间维度排列为第一温度输入向量；使用所述时序编码器的全连接层以如下公式对所述第一温度输入向量进行全连接编码以提取出所述第一温度输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000013
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022119265-appb-000014
表示矩阵乘；使用所述时序编码器的一维卷积层以如下公式对所述第一温度输入向量进行一维卷积编码以提取出所述第一温度输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000015
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
更具体地，在本申请实施例中，所述第二温度数据编码模块，包括：第二温度输入向量构造单元，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值按照时间维度排列为第二温度输入向量；第二全连接编码单元，用于使用所述时序编码器的全连接层以如下公式对所述第二温度输入向量进行全连接编码以提取出所述第二温度输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000016
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022119265-appb-000017
表示矩阵乘；第二一维卷积编码单元，用于使用所述时序编码器的一维卷积层以如下公式对所述第二温度输入向量进行一维卷积编码以提取出所述第二温度输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000018
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
图3图示了根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统中第一温度数据编码模块的框图。如图3所示，所述第一温度数据编码模块230，包括：第一温度输入向量构造单元231，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值按照时间维度排列为第一温度输入向量；第一全连接编码单元232，用于使用所述时序编码器的全连接层以如下公式对所述第一温度输入向量进行全连接编码以提取出所述第一温度输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000019
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022119265-appb-000020
表示矩阵乘；第一一维卷积编码单元233，用于使用所述时序编码器的一维卷积层以如下公式对所述第一温度输入向量进行一维卷积编码以提取出所述第一温度输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000021
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
具体地，在本申请实施例中，所述流速数据编码模块250，用于将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量。应可以理解，对于所述冷却液的流速值，考虑到其在时间维度上也存在着动态性的规律特征，因此，在本申请的技术方案中，也将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器中进行编码，以提取出所述多个预定时间点的冷却液的流速值的局部隐含动态关联特征信息，从而得到流速特征向量。特别地，值得一提的是，所述第一时序编码器、所述第二时序编码器和所述第三时序编码器具有相同的网络结构，所述第一时序编码器、所述第二时序编码器和所述第三时序编码器由交替排列的全连接层和一维卷积层组成。
更具体地，在本申请实施例中，所述流速数据编码模块，包括：流速输入向量构造单元，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值按照时间维度排列为流速输入向量；流速全连接编码单元，用于使用所述时序编码器的全连接层以如下公式对所述流速输入向量进行全连接编码以提取出所述流速输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000022
Figure PCTCN2022119265-appb-000023
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022119265-appb-000024
表示矩阵乘；流速一维卷积编码单元，用于使用所述时序编码器的一维卷积层以如下公式对所述流速输入向量进行 一维卷积编码以提取出所述流速输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022119265-appb-000025
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
具体地，在本申请实施例中，所述特征分布校正模块260，用于对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量。应可以理解，在使用贝叶斯概率模型之前，需要首先将所述待冷却物温度时序特征向量、所述流速特征向量和所述出口温度特征向量映射到概率空间，但是，在通过最大值归一化等线性映射方案进行映射时，无法实现对特征向量的特征值集合所表达的特征分布到概率化的分类目标的约束，从而会影响利用贝叶斯概率模型计算得到的后验概率向量的分类效果。因此，在本申请的技术方案中，在使用贝叶斯概率模型之前，首先对所述待冷却物温度时序特征向量、所述流速特征向量和所述出口温度特征向量进行类条件边界约束，以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量。
更具体地，在本申请实施例中，所述特征分布校正模块，进一步用于：以如下公式对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量；其中，所述公式为：
Figure PCTCN2022119265-appb-000026
Figure PCTCN2022119265-appb-000027
Figure PCTCN2022119265-appb-000028
其中
Figure PCTCN2022119265-appb-000029
和
Figure PCTCN2022119265-appb-000030
分别是所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量的第i个位置的特征值，
Figure PCTCN2022119265-appb-000031
和
Figure PCTCN2022119265-appb-000032
分别是所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量的第i个位置的特征值。应可以理解，这里，该所述类条件边界约束通过对特征值及其所属的类条件进行基于信息规则的结构理解来进行特征的边界约束，从而避免了特征值集合由于集合的分布外特征值而导致在分类目标域内的决策区域的过度碎片化，使得特征向量表示的特征分布获得了稳健的条件化的类边界，从而实现了各个所述特征分布到概率化的分类目标的约束，提高了所述贝叶斯概率模型所获得的后验概率特征分布的分类效果，进而提高了分类的准确性。
具体地，在本申请实施例中，所述类贝叶斯融合模块270，用于使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量。应可以理解，考虑到该所述流速特征向量为先验概率，本申请的技术方案的目的是在新的证据，即在所述水洗后的物料在所述冷凝器内的温度值变化前提下，更新先验概率以得到后验概率。那么，根据贝叶斯公式，后验概率为先验概率乘以事件的概率除以证据的概率，因此，在本申请的技术方案中，使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量。
更具体地，在本申请实施例中，所述类贝叶斯融合模块，进一步用于：使用类贝叶斯概率模型以如下公式来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到所述后验概率向量；其中，所述公式为：
Figure PCTCN2022119265-appb-000033
其中
Figure PCTCN2022119265-appb-000034
是所述校正后流速特征向量中的每个位置的值，
Figure PCTCN2022119265-appb-000035
和
Figure PCTCN2022119265-appb-000036
分别是所述校正后出口温度特征向量和所述校正后待冷却物温度时序特征向量中的每个位置的值，而
Figure PCTCN2022119265-appb-000037
是所述后验概率向量中的每个位置的值。
具体地，在本申请实施例中，所述循环控制结果生成模块280，用于将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。更具体地，在本申请实施例中，所述循环控制结果生成模块，进一步用于：使用所述分类器以如下公式对所述后验概率向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述后验概率向量。
综上，基于本申请实施例的所述用于六氟丁二烯制备的智能化冷却液循环控制系统200被阐明， 其通过人工智能控制方法，来提取出水洗后的物料在所述冷凝器内的温度值、六氟丁二烯在所述冷凝器的出口处的温度值以及冷却液的流速值这三者的动态特征信息，并融合这三者在时序上的动态特征来进行所述冷却液流速的智能控制，以提高能量的有效利用率。
如上所述，根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统200可以实现在各种终端设备中，例如用于六氟丁二烯制备的智能化冷却液循环控制算法的服务器等。在一个示例中，根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统200可以作为一个软件模块和/或硬件模块而集成到终端设备中。例如，该用于六氟丁二烯制备的智能化冷却液循环控制系统200可以是该终端设备的操作系统中的一个软件模块，或者可以是针对于该终端设备所开发的一个应用程序；当然，该用于六氟丁二烯制备的智能化冷却液循环控制系统200同样可以是该终端设备的众多硬件模块之一。
替换地，在另一示例中，该用于六氟丁二烯制备的智能化冷却液循环控制系统200与该终端设备也可以是分立的设备，并且该用于六氟丁二烯制备的智能化冷却液循环控制系统200可以通过有线和/或无线网络连接到该终端设备，并且按照约定的数据格式来传输交互信息。
示例性方法
图4图示了用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法的流程图。如图4所示，根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法，包括步骤：S110，获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；S120，获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；S130，将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；S140，将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；S150，将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；S160，对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；S170，使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及，S180，将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
图5图示了根据本申请实施例的用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法的架构示意图。如图5所示，在所述用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法的网络架构中，首先，将获得的所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值(例如，如图5中所示意的P1)通过包含一维卷积层的第一时序编码器(例如，如图5中所示意的E1)以得到待冷却物温度时序特征向量(例如，如图5中所示意的VF1)；接着，将获得的所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值(例如，如图5中所示意的P2)通过包含一维卷积层的第二时序编码器(例如，如图5中所示意的E2)以得到出口温度特征向量(例如，如图5中所示意的VF2)；然后，将获得的所述多个预定时间点的冷却液的流速值(例如，如图5中所示意的P3)通过包含一维卷积层的第三时序编码器(例如，如图5中所示意的E3)以得到流速特征向量(例如，如图5中所示意的VF3)；接着，对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量(例如，如图5中所示意的V1)、校正后出口温度特征向量(例如，如图5中所示意的V2)和校正后流速特征向量(例如，如图5中所示意的V3)；然后，使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量(例如，如图5中所示意的VF)；以及，最后，将所述后验概率向量通过分类器(例如，如图5中所示意的圈S)以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
更具体地，在步骤S110和S120中，获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值，并获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值。应可以理解，考虑到若通过对于冷却液循环系统的冷却液流速进行控制，以使得六氟丁二烯在冷凝器的出口温度保持在预定范围内，可提高最终六氟丁二烯的纯化精度和产品一致性。因此，在本申请的技术方案中，通过部署于冷凝器内的第一温度传感器采集水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及部署于所述冷凝器的出口处的第二温度传感器采集六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值来进行温 度的动态监控，并且通过部署于所述冷凝器内的流速计采集多个预定时间点的所述冷却液的流速值以进行所述冷却液的流速动态控制。通过这样的方式，可以基于对所述冷却液流速的动态控制来使得六氟丁二烯在冷凝器的出口温度保持在预定范围内，进而提高最终六氟丁二烯的纯化精度和产品一致性。
更具体地，在步骤S130和步骤S140中，将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量，并将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量。应可以理解，考虑到所述水洗后的物料在所述冷凝器内的温度值以及所述六氟丁二烯在所述冷凝器的出口处的温度值在时间维度上都具有动态性的规律，因此为了更为充分地挖掘出这种动态性的变化特征信息，在本申请的技术方案中，进一步将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值分别通过包含一维卷积层的第一时序编码器和第二时序编码器中进行编码处理，以分别得到待冷却物温度时序特征向量和出口温度特征向量。特别地，所述第一时序编码器和所述第二时序编码器具有相同的网络结构。在一个具体示例中，所述第一时序编码器和所述第二时序编码器都由交替排列的全连接层和一维卷积层组成，其通过一维卷积编码提取出所述温度值在时序维度上的动态关联特征和通过全连接编码提取所述温度值的高维隐含特征。
更具体地，在步骤S150中，将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量。应可以理解，对于所述冷却液的流速值，考虑到其在时间维度上也存在着动态性的规律特征，因此，在本申请的技术方案中，也将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器中进行编码，以提取出所述多个预定时间点的冷却液的流速值的局部隐含动态关联特征信息，从而得到流速特征向量。特别地，值得一提的是，所述第一时序编码器、所述第二时序编码器和所述第三时序编码器具有相同的网络结构，所述第一时序编码器、所述第二时序编码器和所述第三时序编码器由交替排列的全连接层和一维卷积层组成。
更具体地，在步骤S160中，对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量。应可以理解，在使用贝叶斯概率模型之前，需要首先将所述待冷却物温度时序特征向量、所述流速特征向量和所述出口温度特征向量映射到概率空间，但是，在通过最大值归一化等线性映射方案进行映射时，无法实现对特征向量的特征值集合所表达的特征分布到概率化的分类目标的约束，从而会影响利用贝叶斯概率模型计算得到的后验概率向量的分类效果。因此，在本申请的技术方案中，在使用贝叶斯概率模型之前，首先对所述待冷却物温度时序特征向量、所述流速特征向量和所述出口温度特征向量进行类条件边界约束，以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量。
更具体地，在步骤S170中，使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量。应可以理解，考虑到该所述流速特征向量为先验概率，本申请的技术方案的目的是在新的证据，即在所述水洗后的物料在所述冷凝器内的温度值变化前提下，更新先验概率以得到后验概率。那么，根据贝叶斯公式，后验概率为先验概率乘以事件的概率除以证据的概率，因此，在本申请的技术方案中，使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量。
更具体地，在步骤S180中，将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。具体地，在本申请实施例中，使用所述分类器以如下公式对所述后验概率向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述后验概率向量。
综上，基于本申请实施例的所述用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法被阐明，其通过人工智能控制方法，来提取出水洗后的物料在所述冷凝器内的温度值、六氟丁二烯在所述冷凝器的出口处的温度值以及冷却液的流速值这三者的动态特征信息，并融合这三者在时序上的动态特征来进行所述冷却液流速的智能控制，以提高能量的有效利用率。
以上结合具体实施例描述了本申请的基本原理，但是，需要指出的是，在本申请中提及的优点、优势、效果等仅是示例而非限制，不能认为这些优点、优势、效果等是本申请的各个实施例必须具备的。另外，上述公开的具体细节仅是为了示例的作用和便于理解的作用，而非限制，上述细节并不限制本申请为必须采用上述具体的细节来实现。
本申请中涉及的器件、装置、设备、系统的方框图仅作为例示性的例子并且不意图要求或暗示必须按照方框图示出的方式进行连接、布置、配置。如本领域技术人员将认识到的，可以按任意方式连 接、布置、配置这些器件、装置、设备、系统。诸如“包括”、“包含”、“具有”等等的词语是开放性词汇，指“包括但不限于”，且可与其互换使用。这里所使用的词汇“或”和“和”指词汇“和/或”，且可与其互换使用，除非上下文明确指示不是如此。这里所使用的词汇“诸如”指词组“诸如但不限于”，且可与其互换使用。
还需要指出的是，在本申请的装置、设备和方法中，各部件或各步骤是可以分解和/或重新组合的。这些分解和/或重新组合应视为本申请的等效方案。
提供所公开的方面的以上描述以使本领域的任何技术人员能够做出或者使用本申请。对这些方面的各种修改对于本领域技术人员而言是非常显而易见的，并且在此定义的一般原理可以应用于其他方面而不脱离本申请的范围。因此，本申请不意图被限制到在此示出的方面，而是按照与在此公开的原理和新颖的特征一致的最宽范围。
为了例示和描述的目的已经给出了以上描述。此外，此描述不意图将本申请的实施例限制到在此公开的形式。尽管以上已经讨论了多个示例方面和实施例，但是本领域技术人员将认识到其某些变型、修改、改变、添加和子组合。

Claims (10)

  1. 一种用于六氟丁二烯制备的智能化冷却液循环控制系统，其特征在于，包括：温度数据采集模块，用于获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；冷却液循环流速数据采集模块，用于获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；第一温度数据编码模块，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；第二温度数据编码模块，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；流速数据编码模块，用于将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；特征分布校正模块，用于对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；类贝叶斯融合模块，用于使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及循环控制结果生成模块，用于将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
  2. 根据权利要求1所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述第一温度数据编码模块，包括：第一温度输入向量构造单元，用于将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值按照时间维度排列为第一温度输入向量；第一全连接编码单元，用于使用所述时序编码器的全连接层以如下公式对所述第一温度输入向量进行全连接编码以提取出所述第一温度输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
    Figure PCTCN2022119265-appb-100001
    其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
    Figure PCTCN2022119265-appb-100002
    表示矩阵乘；第一一维卷积编码单元，用于使用所述时序编码器的一维卷积层以如下公式对所述第一温度输入向量进行一维卷积编码以提取出所述第一温度输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
    Figure PCTCN2022119265-appb-100003
    其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
  3. 根据权利要求2所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述第二温度数据编码模块，包括：第二温度输入向量构造单元，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值按照时间维度排列为第二温度输入向量；第二全连接编码单元，用于使用所述时序编码器的全连接层以如下公式对所述第二温度输入向量进行全连接编码以提取出所述第二温度输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
    Figure PCTCN2022119265-appb-100004
    其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
    Figure PCTCN2022119265-appb-100005
    表示矩阵乘；第二一维卷积编码单元，用于使用所述时序编码器的一维卷积层以如下公式对所述第二温度输入向量进行一维卷积编码以提取出所述第二温度输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
    Figure PCTCN2022119265-appb-100006
    其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
  4. 根据权利要求3所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述流速数据编码模块，包括：流速输入向量构造单元，用于将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值按照时间维度排列为流速输入向量；流速全连接编码单元，用于使用所述时序编码器的全连接层以如下公式对所述流速输入向量进行全连接编码以提取出所述流速输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
    Figure PCTCN2022119265-appb-100007
    其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
    Figure PCTCN2022119265-appb-100008
    表示矩阵乘；流速一维卷积编码单元，用于使用所述时序编码器的一维卷积层以如下公式对所述流速输入向量进行一维卷积编码以提取出所述流速输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
    Figure PCTCN2022119265-appb-100009
    其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸。
  5. 根据权利要求4所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述第一时序编码器、所述第二时序编码器和所述第三时序编码器具有相同的网络结构，所述第一时序编码器、所述第二时序编码器和所述第三时序编码器由交替排列的全连接层和一维卷积层组成。
  6. 根据权利要求5所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述特征分布校正模块，进一步用于以如下公式对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量；其中，所述公式为：
    Figure PCTCN2022119265-appb-100010
    Figure PCTCN2022119265-appb-100011
    Figure PCTCN2022119265-appb-100012
    其中
    Figure PCTCN2022119265-appb-100013
    和
    Figure PCTCN2022119265-appb-100014
    分别是所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量的第i个位置的特征值，
    Figure PCTCN2022119265-appb-100015
    和
    Figure PCTCN2022119265-appb-100016
    分别是所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量的第i个位置的特征值。
  7. 根据权利要求6所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述类贝叶斯融合模块，进一步用于：使用类贝叶斯概率模型以如下公式来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到所述后验概率向量；其中，所述公式为：
    Figure PCTCN2022119265-appb-100017
    其中
    Figure PCTCN2022119265-appb-100018
    是所述校正后流速特征向量中的每个位置的值，
    Figure PCTCN2022119265-appb-100019
    和
    Figure PCTCN2022119265-appb-100020
    分别是所述校正后出口温度特征向量和所述校正后待冷却物温度时序特征向量中的每个位置的值，而
    Figure PCTCN2022119265-appb-100021
    是所述后验概率向量中的每个位置的值。
  8. 根据权利要求7所述的用于六氟丁二烯制备的智能化冷却液循环控制系统，其中，所述循环控制结果生成模块，进一步用于：使用所述分类器以如下公式对所述后验概率向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述后验概率向量。
  9. 一种用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法，其特征在于，包括：
    获取由部署于冷凝器内的第一温度传感器采集的水洗后的物料在所述冷凝器内的多个预定时间点的温度值以及由部署于所述冷凝器的出口处的第二温度传感器采集的六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值；获取由部署于所述冷凝器内的流速计采集的所述多个预定时间点的冷却液的流速值；将所述水洗后的物料在所述冷凝器内的多个预定时间点的温度值通过包含一维卷积层的第一时序编码器以得到待冷却物温度时序特征向量；将所述六氟丁二烯在所述冷凝器的出口处的多个预定时间点的温度值通过包含一维卷积层的第二时序编码器以得到出口温度特征向量；将所述多个预定时间点的冷却液的流速值通过包含一维卷积层的第三时序编码器以得到流速特征向量；对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量；使用类贝叶斯概率模型来融合所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量以得到后验概率向量；以及将所述后验概率向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却液流速应增大或应减小。
  10. 根据权利要求9所述的用于六氟丁二烯制备的智能化冷却液循环控制系统的控制方法，其中，对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到校正后待冷却物温度时序特征向量、校正后出口温度特征向量和校正后流速特征向量，包括：
    以如下公式对所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向量分别进行特征分布校正以得到所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量；其中，所述公式为：
    Figure PCTCN2022119265-appb-100022
    Figure PCTCN2022119265-appb-100023
    Figure PCTCN2022119265-appb-100024
    其中
    Figure PCTCN2022119265-appb-100025
    和
    Figure PCTCN2022119265-appb-100026
    分别是所述待冷却物温度时序特征向量、所述出口温度特征向量和所述流速特征向 量的第i个位置的特征值，
    Figure PCTCN2022119265-appb-100027
    和
    Figure PCTCN2022119265-appb-100028
    分别是所述校正后待冷却物温度时序特征向量、所述校正后出口温度特征向量和所述校正后流速特征向量的第i个位置的特征值。
PCT/CN2022/119265 2022-07-15 2022-09-16 一种用于六氟丁二烯制备的智能化冷却液循环控制系统 WO2024011747A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210829857.9 2022-07-15
CN202210829857.9A CN115342679B (zh) 2022-07-15 2022-07-15 一种用于六氟丁二烯制备的智能化冷却液循环控制系统

Publications (1)

Publication Number Publication Date
WO2024011747A1 true WO2024011747A1 (zh) 2024-01-18

Family

ID=83948229

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/119265 WO2024011747A1 (zh) 2022-07-15 2022-09-16 一种用于六氟丁二烯制备的智能化冷却液循环控制系统

Country Status (2)

Country Link
CN (1) CN115342679B (zh)
WO (1) WO2024011747A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117965214A (zh) * 2024-04-01 2024-05-03 新疆凯龙清洁能源股份有限公司 一种天然气脱二氧化碳制合成气的方法和系统

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115342679B (zh) * 2022-07-15 2023-04-07 福建省杭氟电子材料有限公司 一种用于六氟丁二烯制备的智能化冷却液循环控制系统
CN115903705B (zh) * 2022-11-30 2023-07-14 福建省杭氟电子材料有限公司 用于电子级六氟丁二烯制备的生产管理控制系统
CN117046140B (zh) * 2023-02-28 2024-03-08 福建天甫电子材料有限公司 用于电子级氢氟酸精馏的装置
CN116819957B (zh) * 2023-03-29 2024-03-08 福建省龙德新能源有限公司 电子级六氟磷酸锂的尾气处理系统及其方法
CN116555976B (zh) * 2023-05-15 2023-10-13 肃宁县中原纺织有限责任公司 适用于火炮装备的绝热性材料制备方法及其系统
CN117404853B (zh) * 2023-12-14 2024-03-08 山西省水利建筑工程局集团有限公司 隧道掘进机外用循环水冷却系统及方法

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587357A (zh) * 2009-05-12 2009-11-25 上海微电子装备有限公司 温度控制方法及其温度控制系统
US20110048012A1 (en) * 2009-09-02 2011-03-03 Cummins Intellectual Properties, Inc. Energy recovery system and method using an organic rankine cycle with condenser pressure regulation
CN206973986U (zh) * 2017-02-14 2018-02-06 上海再启生物技术有限公司 低温冷却循环泵
CN108007021A (zh) * 2017-11-07 2018-05-08 珠海格力电器股份有限公司 一种冷凝器控制系统、方法及冷水机组
CN208238584U (zh) * 2018-03-30 2018-12-14 无锡宝丰石化装备有限公司 一种分馏塔冷凝器
CN216913137U (zh) * 2022-04-01 2022-07-08 东莞市西米克电机制造有限公司 一种电主轴的智能控制冷却结构
CN115342679A (zh) * 2022-07-15 2022-11-15 福建省杭氟电子材料有限公司 一种用于六氟丁二烯制备的智能化冷却液循环控制系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587357A (zh) * 2009-05-12 2009-11-25 上海微电子装备有限公司 温度控制方法及其温度控制系统
US20110048012A1 (en) * 2009-09-02 2011-03-03 Cummins Intellectual Properties, Inc. Energy recovery system and method using an organic rankine cycle with condenser pressure regulation
CN206973986U (zh) * 2017-02-14 2018-02-06 上海再启生物技术有限公司 低温冷却循环泵
CN108007021A (zh) * 2017-11-07 2018-05-08 珠海格力电器股份有限公司 一种冷凝器控制系统、方法及冷水机组
CN208238584U (zh) * 2018-03-30 2018-12-14 无锡宝丰石化装备有限公司 一种分馏塔冷凝器
CN216913137U (zh) * 2022-04-01 2022-07-08 东莞市西米克电机制造有限公司 一种电主轴的智能控制冷却结构
CN115342679A (zh) * 2022-07-15 2022-11-15 福建省杭氟电子材料有限公司 一种用于六氟丁二烯制备的智能化冷却液循环控制系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117965214A (zh) * 2024-04-01 2024-05-03 新疆凯龙清洁能源股份有限公司 一种天然气脱二氧化碳制合成气的方法和系统

Also Published As

Publication number Publication date
CN115342679A (zh) 2022-11-15
CN115342679B (zh) 2023-04-07

Similar Documents

Publication Publication Date Title
WO2024011747A1 (zh) 一种用于六氟丁二烯制备的智能化冷却液循环控制系统
WO2024098604A1 (zh) 用于电子级四氟化碳制备的精馏控制系统及其方法
Geng et al. Novel transformer based on gated convolutional neural network for dynamic soft sensor modeling of industrial processes
Huang et al. A projective and discriminative dictionary learning for high-dimensional process monitoring with industrial applications
WO2024045247A1 (zh) 用于氟化铵生产的生产管理控制系统及其控制方法
WO2024000800A1 (zh) 用于六氟磷酸锂制备的能源管理控制系统及其控制方法
WO2024021258A1 (zh) 电子级氢氧化钾的智慧产线的控制系统及其控制方法
WO2024045243A1 (zh) 用于六氟磷酸锂制备的自动配料系统及其配料方法
WO2023226226A1 (zh) 用于电子级三氟甲烷制备的精馏控制系统及其控制方法
CN115599049B (zh) 用于无水氟化氢生产的能源管理控制系统及其控制方法
WO2024021254A1 (zh) 电子级三氟化氯的智能分离纯化系统
WO2024000798A1 (zh) 用于电子级氢氟酸制备的生产管理控制系统及其控制方法
CN116067188B (zh) 一种六氟磷酸锂制备用的余热回收系统
WO2024113599A1 (zh) 用于电子级六氟丁二烯制备的生产管理控制系统
CN115470857B (zh) 变电站全景数字孪生系统及其方法
CN104848708B (zh) 一种基于温度场和流速场的空冷岛阵列控制方法
CN116455941B (zh) 基于物联网的室内环境多源数据传输方法及系统
CN115866621B (zh) 一种基于鲸鱼算法的无线传感器网络覆盖方法
CN115851332A (zh) 从合成氨驰放气膜提氢尾气中制取天然气的方法及系统
Kristianto et al. Integration of K-means clustering and naïve bayes classification algorithms for smart ac monitoring and control in WSAN
Wu et al. Prediction of CO2 Solubility in Polymers by Radial Basis Function Artificial Neural Network Based on Chaotic Self‐adaptive Particle Swarm Optimization and Fuzzy Clustering Method
Rao et al. Design optimization of rotary regenerator using artificial bee colony algorithm
Mehrani et al. Adaptive sampling rate determination for energy efficiency of wireless body area networks
WO2024138944A1 (zh) 用于电子级四氟化碳的废气回收再利用方法及其系统
Lei et al. Hessian Regularization Semi-supervised Extreme Learning Machine for Superheat Identification in Aluminum Reduction Cell