WO2024000800A1 - 用于六氟磷酸锂制备的能源管理控制系统及其控制方法 - Google Patents

用于六氟磷酸锂制备的能源管理控制系统及其控制方法 Download PDF

Info

Publication number
WO2024000800A1
WO2024000800A1 PCT/CN2022/116744 CN2022116744W WO2024000800A1 WO 2024000800 A1 WO2024000800 A1 WO 2024000800A1 CN 2022116744 W CN2022116744 W CN 2022116744W WO 2024000800 A1 WO2024000800 A1 WO 2024000800A1
Authority
WO
WIPO (PCT)
Prior art keywords
temperature
vector
feature vector
pressure
feature
Prior art date
Application number
PCT/CN2022/116744
Other languages
English (en)
French (fr)
Inventor
赖育河
戴浩翔
谢光明
陈奕雯
詹秀玲
Original Assignee
福建省龙德新能源有限公司
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 福建省龙德新能源有限公司 filed Critical 福建省龙德新能源有限公司
Publication of WO2024000800A1 publication Critical patent/WO2024000800A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance

Definitions

  • the present invention relates to the field of intelligent production control, and more specifically, to an energy management control system for the preparation of lithium hexafluorophosphate and a control method thereof.
  • Lithium-ion batteries have the advantages of high platform voltage, good cycle performance, wide operating temperature range, high energy density, and no memory effect. They are widely used in mobile communications, portable electronic products, power tools, weapons assembly and other fields. They are currently used in power vehicles. It also has broad application prospects.
  • the electrolyte and the electrolyte in it are the key raw materials of lithium batteries and directly affect the performance of the prepared battery. Therefore, the electrolyte must meet the following requirements: high conductivity, stable chemical and electrochemical properties, wide usable temperature range and low price. Lithium hexafluorophosphate meets the above requirements and can serve as an excellent electrolyte.
  • lithium hexafluorophosphate Due to the continuous development of materials science, solutions for preparing lithium hexafluorophosphate are constantly emerging. Among different preparation schemes, the energy management issue of preparation schemes has been a topic that has gradually attracted attention in recent years. On the one hand, the preparation schemes are required to meet energy-saving requirements. On the other hand, due to the limitation of the preparation process itself, it is necessary to prepare products that meet quality requirements. of lithium hexafluorophosphate. That is, in the preparation scheme of lithium hexafluorophosphate, it is expected to provide an energy management control system that can save energy consumption as much as possible on the premise of preparing products that meet quality requirements.
  • the embodiments of the present application provide an energy management control system and a control method for the preparation of lithium hexafluorophosphate, which use artificial intelligence control technology to monitor three aspects: solid precipitation monitoring video, temperature value and pressure value in the distiller. Perform dynamic feature analysis, and then in the control of temperature and pressure at the current time point, the temperature and pressure can be adaptively adjusted based on the actual situation of the object to be distilled, so as to improve energy utilization and improve the efficiency and purity of the precipitated solids.
  • an energy management control system for the preparation of lithium hexafluorophosphate which includes: a solid precipitation monitoring video acquisition module for acquiring the scheduled time collected by an underwater camera deployed in the distiller The solid precipitation monitoring video of the segment; the status data collection module in the distiller, used to obtain the temperature values and pressure values at multiple predetermined time points in the predetermined time period collected by the temperature sensor and pressure sensor deployed in the distiller;
  • a temperature and pressure data encoding module configured to calculate the sum of the temperature input vectors after arranging the temperature values at the plurality of predetermined time points and the pressure values at the plurality of predetermined time points into temperature input vectors and pressure input vectors respectively.
  • the vector product between the transposed vectors of the pressure input vector is used to obtain the temperature-pressure time series correlation matrix;
  • a correlation coding module used to pass the temperature-pressure time series correlation matrix through the first convolutional neural network as a filter to obtain the correlation feature vector;
  • a video encoding module configured to pass the solid precipitation monitoring video of the predetermined time period through a second convolutional neural network using a three-dimensional convolution kernel to obtain a precipitation feature vector;
  • a feature distribution fusion module used to fuse the associated feature vector and the extracted feature vector to obtain a classification feature vector
  • An energy management result generation module is used to pass the classification feature vector through a classifier to obtain a classification result.
  • the classification result is used to indicate that the temperature at the current time point should increase or decrease, and that the pressure at the current time point should increase or should decrease.
  • the correlation coding module is further used for each layer of the first convolutional neural network to perform convolution processing on the input data in the forward transmission of the layer.
  • a convolution feature map To obtain a convolution feature map; perform mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; and perform nonlinear activation on the pooled feature map to obtain an activation feature map; wherein , the output of the last layer of the first convolutional neural network is the correlation feature vector, and the input of the first layer of the first convolutional neural network is the temperature-pressure time series correlation matrix.
  • the video encoding module includes: a sampling unit for extracting multiple key frames from the solid precipitation monitoring video of the predetermined time period at a predetermined sampling frequency; multi-layer a convolution unit, used to pass the plurality of key frames through multiple convolution layers of the second convolutional neural network to obtain a precipitation feature map; and a global mean pooling unit, used to process the precipitation feature map A global mean pooling process based on the feature matrix is performed to obtain the precipitation feature vector.
  • the multi-layer convolution unit is further used to use the multi-layer convolution layer of the second convolutional neural network to respectively process the input data in the forward transmission of the layer.
  • Convolution processing, pooling processing and nonlinear activation processing based on the three-dimensional convolution kernel are performed to output the precipitation feature map from the last convolution layer of the multi-layer convolution layer.
  • the feature distribution fusion module is further used to: fuse the correlation feature vector and the precipitation feature vector with the following formula to obtain the classification feature vector;
  • V c represents the classification feature vector
  • V 1 represents the association feature vector
  • V 2 represents the precipitation feature vector
  • 1 represents a norm of the vector
  • F represents The Frobenius norm of the matrix
  • ⁇ represents the dot product by position, means adding by position, Represents subtraction by position.
  • the energy management result generation module is further used to use the classifier to process the classification feature vector with the following formula to obtain the classification result, wherein: The above formula is: softmax ⁇ (W n ,B n ):...:( W 1 ,B 1 )
  • a control method for an energy management control system for the preparation of lithium hexafluorophosphate which includes:
  • the classification feature vector is passed through a classifier to obtain a classification result.
  • the classification result is used to indicate that the temperature at the current time point should increase or decrease, and the pressure at the current time point should increase or decrease.
  • the temperature-pressure time series correlation matrix is passed through the first convolutional neural network as a filter to obtain the correlation feature vector, including: the first convolution
  • Each layer of the neural network is performed separately in the forward pass of the layer: convolution processing is performed on the input data to obtain the convolution feature map; mean pooling based on the local feature matrix is performed on the convolution feature map to obtain the pooling feature and perform nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the associated feature vector, and the first convolution
  • the input of the first layer of the neural network is the temperature-pressure time series correlation matrix.
  • the solid precipitation monitoring video of the predetermined period of time is passed through the second convolutional neural network using a three-dimensional convolution kernel to obtain the precipitation feature vector, including:
  • the sampling frequency extracts multiple key frames from the solid precipitation monitoring video of the predetermined time period; passes the multiple key frames through the multi-layer convolution layer of the second convolutional neural network to obtain a precipitation feature map;
  • the precipitation feature map is subjected to global mean pooling processing based on the feature matrix to obtain the precipitation feature vector.
  • passing the multiple key frames through the multi-layer convolution layer of the second convolutional neural network to obtain a precipitation characteristic map includes: using the third The multi-layer convolution layer of the two-convolution neural network performs convolution processing, pooling processing and non-linear activation processing based on the three-dimensional convolution kernel on the input data in the forward pass of the layer to obtain the multi-layer convolution layer.
  • the last convolutional layer of the multilayer outputs the precipitation feature map.
  • fusing the correlation feature vector and the precipitation feature vector to obtain a classification feature vector includes: fusing the correlation feature vector and the precipitation feature with the following formula vector to obtain the classification feature vector;
  • V c represents the classification feature vector
  • V 1 represents the association feature vector
  • V 2 represents the precipitation feature vector
  • 1 represents a norm of the vector
  • F represents The Frobenius norm of the matrix
  • ⁇ represents the dot product by position, means adding by position, Represents subtraction by position.
  • passing the classification feature vector through a classifier to obtain a classification result includes: using the classifier to process the classification feature vector with the following formula to obtain The classification result, wherein the formula is: softmax ⁇ (W n ,B n ):...:(W 1 ,B 1 )
  • a computer-readable medium is provided with computer program instructions stored thereon, which when executed by a processor cause the processor to perform the above-mentioned preparation of lithium hexafluorophosphate. Control method of energy management control system.
  • the energy management control system and its control method for the preparation of lithium hexafluorophosphate use artificial intelligence control technology to monitor the solid precipitation monitoring video, the temperature value and the pressure value in the distiller. Dynamically analyze the characteristics of each aspect, and then in the control of the temperature and pressure at the current time point, the temperature and pressure can be adaptively adjusted based on the actual situation of the object to be distilled, so as to improve energy utilization and improve the efficiency and purity of the precipitated solids. .
  • Figure 1 is an application scenario diagram of an energy management control system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG. 2 is a block diagram of an energy management control system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • Figure 3 is a block diagram of a video encoding module in an energy management control system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • Figure 4 is a flow chart of a control method of an energy management control system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • Figure 5 is a schematic structural diagram of a control method of an energy management control system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • lithium-ion batteries have the advantages of high platform voltage, good cycle performance, wide operating temperature range, high energy density, and no memory effect. They are widely used in mobile communications, portable electronic products, power tools, weapon assembly and other fields. At present, it also has broad application prospects in power vehicles.
  • the electrolyte and the electrolyte in it are the key raw materials of lithium batteries and directly affect the performance of the prepared battery. Therefore, the electrolyte must meet the following requirements: high conductivity, stable chemical and electrochemical properties, wide usable temperature range and low price. Lithium hexafluorophosphate meets the above requirements and can serve as an excellent electrolyte.
  • lithium hexafluorophosphate Due to the continuous development of materials science, solutions for preparing lithium hexafluorophosphate are constantly emerging. Among different preparation schemes, the energy management issue of preparation schemes has been a topic that has gradually attracted attention in recent years. On the one hand, the preparation schemes are required to meet energy-saving requirements. On the other hand, due to the limitation of the preparation process itself, it is necessary to prepare products that meet quality requirements. of lithium hexafluorophosphate. That is, in the preparation scheme of lithium hexafluorophosphate, it is expected to provide an energy management control system that can save energy consumption as much as possible on the premise of preparing products that meet quality requirements.
  • step S6 Repeat step S4 to obtain high-purity lithium hexafluorophosphate.
  • the above-mentioned organic solvent is at least one of carbonates, alkanes, ethers, and nitrogen-containing organic substances.
  • the inventor of the present application considered that in this preparation scheme, the temperature and pressure control in step S4 are key steps in the energy control of this preparation scheme. It should be understood that under appropriate temperature and pressure control, the efficiency of solid precipitation will be improved and the purity of the precipitated solid will be better.
  • predetermined control programs are used to set the temperature and pressure values at different stages. That is, in the existing energy management control strategies, fixed programs are used to control the temperature and pressure values in each stage. Pressure and temperature, rather than adaptively adopting temperature and pressure based on the actual conditions of the object being distilled, in order to improve energy utilization and improve the efficiency and purity of the precipitated solids.
  • the actual dynamic situation of the distilled object is judged by using the monitoring video of the precipitated solid, and the dynamic change characteristic information of the temperature and pressure of the actual environment is collected through the temperature sensor and the pressure sensor, so that In the control strategy of energy management, the temperature and pressure can be adaptively adjusted based on the actual conditions of the object being distilled to improve energy utilization and improve the efficiency and purity of the precipitated solids.
  • the solid precipitation monitoring video of a predetermined period of time is collected through an underwater camera deployed in the distiller, and the temperature sensor and pressure sensor deployed in the distiller are used to collect the video of the predetermined period of time.
  • the temperature values at the multiple predetermined time points and the pressure values at the multiple predetermined time points are respectively arranged into a temperature input vector and After the pressure input vector is obtained, the vector product between the temperature input vector and the transposed vector of the pressure input vector is calculated to obtain a temperature-pressure time series correlation matrix. Further, the temperature-pressure time series correlation matrix is processed through the first convolutional neural network as a filter to extract the temperature values of the multiple predetermined time points and the temperature values of the multiple predetermined time points. The implicit correlation feature distribution information of the pressure value is obtained, thereby obtaining the correlation feature vector.
  • the solid precipitation monitoring video of the predetermined time period has the implicit dynamic change characteristics of solid precipitation in the time series dimension
  • the solid precipitation of the predetermined time period is The surveillance video is processed through a second convolutional neural network using a three-dimensional convolution kernel to obtain the extracted feature vector.
  • the two-dimensional convolution kernel as a filter of the first convolutional neural network that extracts temperature-pressure correlation, or the second volume that extracts image semantic-temporal correlation features in the video
  • the three-dimensional convolution kernel of the convolutional neural network all obtains feature vectors as expressions of deep associated semantic features. Due to the differences in feature dimensions and semantic expressions, before feature fusion, it is preferable to combine the associated feature vector V 1 with the Extract the feature vector V 2 for deep single-pair alignment.
  • V c represents the classification feature vector
  • V 1 represents the association feature vector
  • V 2 represents the precipitation feature vector
  • 1 represents a norm of the vector
  • F represents The Frobenius norm of the matrix
  • ⁇ represents the dot product by position, means adding by position, Represents subtraction by position.
  • the fused classification feature vector performs homography alignment based on the scene depth flow based on the feature depth information characteristics represented by the fusion vector, and adds depth perception based on the vector-based full-scene homography correlation matrix.
  • the bias term allows the feature vectors to be densely and deeply fused on the basis of possible feature distribution differences to obtain a classification feature vector V c with better classification effect, thereby improving the accuracy of classification.
  • this application proposes an energy management control system for the preparation of lithium hexafluorophosphate, which includes: a solid precipitation monitoring video collection module, used to obtain a predetermined period of time collected by an underwater camera deployed in the distiller. Solid precipitation monitoring video; a status data acquisition module in the distiller, used to obtain temperature values and pressure values at multiple predetermined time points in the predetermined time period collected by temperature sensors and pressure sensors deployed in the distiller; temperature and A pressure data encoding module, configured to arrange the temperature values at the plurality of predetermined time points and the pressure values at the plurality of predetermined time points into temperature input vectors and pressure input vectors respectively, and then calculate the temperature input vector and the pressure input vector.
  • FIG 1 illustrates an application scenario diagram of an energy management control system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • an underwater camera for example, C as shown in Figure 1 deployed in the distiller (for example, D as shown in Figure 1)
  • Collect the solid precipitation monitoring video for a predetermined period of time, and collect it through a temperature sensor (for example, T as shown in Figure 1) and a pressure sensor (for example, as shown in Figure 1) that are deployed in the distiller.
  • the obtained solid precipitation monitoring video of the predetermined time period and the temperature values and pressure values of multiple predetermined time points of the predetermined time period are input into a server deployed with an energy management control algorithm for the preparation of lithium hexafluorophosphate (for example, , the cloud server S) as shown in Figure 1, wherein the server can monitor the solid precipitation monitoring video of the predetermined time period and a plurality of predetermined time periods with an energy management control algorithm for the preparation of lithium hexafluorophosphate.
  • the temperature value and pressure value at the time point are processed to generate a classification result indicating that the temperature at the current time point should increase or decrease, and the pressure at the current time point should increase or decrease.
  • FIG. 2 illustrates a block diagram of an energy management control system for lithium hexafluorophosphate preparation according to an embodiment of the present application.
  • the energy management control system 200 for the preparation of lithium hexafluorophosphate according to the embodiment of the present application includes: a solid precipitation monitoring video collection module 210, which is used to obtain video collected by an underwater camera deployed in the distiller. Solid precipitation monitoring video of a predetermined period of time; the status data collection module 220 in the distiller is used to obtain the temperature values and multiple predetermined time points of the predetermined period of time collected by the temperature sensor and pressure sensor deployed in the distiller.
  • temperature and pressure data encoding module 230 used to arrange the temperature values at the plurality of predetermined time points and the pressure values at the plurality of predetermined time points into temperature input vectors and pressure input vectors, and then calculate the The vector product between the temperature input vector and the transposed vector of the pressure input vector is used to obtain the temperature-pressure time series correlation matrix;
  • the correlation encoding module 240 is used to pass the temperature-pressure time series correlation matrix through the first filter Convolutional neural network to obtain the associated feature vector;
  • video encoding module 250 used to pass the solid precipitation monitoring video of the predetermined time period through a second convolutional neural network using a three-dimensional convolution kernel to obtain the precipitation feature vector;
  • feature distribution fusion Module 260 is used to fuse the associated feature vector and the extracted feature vector to obtain a classification feature vector; and
  • an energy management result generation module 270 is used to pass the classification feature vector through a classifier to obtain a classification result, the The classification result is used to indicate that the temperature at the current time point should increase or decrease, and
  • the solid precipitation monitoring video collection module 210 and the distiller internal status data collection module 220 are used to obtain the predetermined time collected by the underwater camera deployed in the distiller.
  • the solid precipitation monitoring video of the segment is monitored, and the temperature values and pressure values at multiple predetermined time points of the predetermined time period collected by the temperature sensor and pressure sensor deployed in the distiller are obtained.
  • the control of temperature and pressure is a key step in the energy control of the preparation scheme. It should be understood that under appropriate temperature and pressure control, the efficiency of solid precipitation will be improved and the purity of the precipitated solid will be better.
  • predetermined control programs are used to set the temperature and pressure values at different stages.
  • the actual dynamic situation of the object to be distilled is determined by using the monitoring video of the precipitated solid, and the dynamic change characteristic information of the temperature and pressure of the actual environment is collected through the temperature sensor and the pressure sensor, so that in In the control strategy of energy management, the temperature and pressure can be adaptively adjusted based on the actual conditions of the object being distilled to improve energy utilization and improve the efficiency and purity of the precipitated solids. That is, specifically, first, the solid precipitation monitoring video of a predetermined period of time is collected through an underwater camera deployed in the distiller, and the temperature sensor and pressure sensor deployed in the distiller are used to collect the video of the predetermined period of time. Temperature values and pressure values at multiple predetermined time points.
  • the temperature values at the multiple predetermined time points and the pressure values at the multiple predetermined time points are respectively arranged into a temperature input vector and After the pressure input vector is obtained, the vector product between the temperature input vector and the transposed vector of the pressure input vector is calculated to obtain a temperature-pressure time series correlation matrix. Further, the temperature-pressure time series correlation matrix is processed through the first convolutional neural network as a filter to extract the temperature values of the multiple predetermined time points and the temperature values of the multiple predetermined time points. The implicit correlation feature distribution information of the pressure value is obtained, thereby obtaining the correlation feature vector.
  • the correlation coding module is further used to perform convolution processing on the input data in each layer of the first convolutional neural network in the forward pass of the layer.
  • the output of the last layer of the first convolutional neural network is the correlation feature vector, and the input of the first layer of the first convolutional neural network is the temperature-pressure time series correlation matrix.
  • the video encoding module 250 is used to pass the solid precipitation monitoring video of the predetermined period of time through a second convolutional neural network using a three-dimensional convolution kernel to obtain a precipitation feature vector. It should be understood that considering that the solid precipitation monitoring video of the predetermined time period has the implicit dynamic change characteristics of solid precipitation in the time series dimension, in order to extract such dynamic implicit characteristics, in the technical solution of the present application , and further process the solid precipitation monitoring video of the predetermined time period through a second convolutional neural network using a three-dimensional convolution kernel to obtain a precipitation feature vector.
  • the video encoding module includes: first, extracting multiple key frames from the solid precipitation monitoring video of the predetermined time period at a predetermined sampling frequency. Then, the multiple key frames are passed through multiple convolutional layers of the second convolutional neural network to obtain a precipitation feature map.
  • the multi-layer convolution layer of the second convolutional neural network is used to perform convolution processing and pooling based on the three-dimensional convolution kernel on the input data in the forward pass of the layer. processing and nonlinear activation processing to output the precipitation feature map from the last convolution layer of the multi-layer convolution layer.
  • the nonlinear activation function used in each layer of the multi-layer convolution layer is the Mish function
  • f(x) x.tanh(ln(1+e x )
  • FIG. 3 illustrates a block diagram of a video encoding module in an energy management control system for lithium hexafluorophosphate preparation according to an embodiment of the present application.
  • the video encoding module 250 includes: a sampling unit 251, used to extract multiple key frames from the solid precipitation monitoring video of the predetermined time period at a predetermined sampling frequency; a multi-layer convolution unit 252, Used to pass the multiple key frames through the multi-layer convolution layer of the second convolutional neural network to obtain a precipitation feature map; and, a global mean pooling unit 253, used to perform feature-based processing on the precipitation feature map. Global mean pooling of the matrix is performed to obtain the precipitation feature vector.
  • the feature distribution fusion module 260 and the energy management result generation module 270 are used to fuse the associated feature vector and the extracted feature vector to obtain a classification feature vector, and combine the The classification feature vector is passed through the classifier to obtain a classification result.
  • the classification result is used to indicate that the temperature at the current time point should increase or decrease, and the pressure at the current time point should increase or decrease. It should be understood that considering whether it is the two-dimensional convolution kernel as a filter of the first convolutional neural network that extracts temperature-pressure correlation, or the second volume that extracts image semantic-temporal correlation features in the video The three-dimensional convolution kernel of the convolutional neural network all obtains feature vectors as expressions of deep associated semantic features.
  • the correlation feature vector and the separation feature vector are fused to obtain a classification feature vector.
  • the classification feature vector can be processed through a classifier to obtain a representation that the temperature at the current time point should increase or decrease, and the pressure at the current time point should increase or decrease. classification results.
  • the classifier is used to process the classification feature vector with the following formula to obtain the classification result, wherein the formula is: softmax ⁇ (W n ,B n ):... :(W 1 ,B 1 )
  • the feature distribution fusion module is further configured to: fuse the associated feature vector and the extracted feature vector with the following formula to obtain the classification feature vector;
  • V c represents the classification feature vector
  • V 1 represents the association feature vector
  • V 2 represents the precipitation feature vector
  • 1 represents a norm of the vector
  • F represents The Frobenius norm of the matrix
  • ⁇ represents the dot product by position, means adding by position, Represents subtraction by position.
  • the fused classification feature vector performs homography alignment based on the scene depth flow based on the feature depth information characteristics represented by the fusion vector, and adds vector-based full scene homography.
  • the depth-sensing bias term of the correlation matrix should be associated, so that the feature vectors can be densely and deeply fused on the basis of possible feature distribution differences to obtain a classification feature vector V c with better classification effect, and thus Can improve classification accuracy.
  • the energy management control system 200 for the preparation of lithium hexafluorophosphate is clarified, which adopts artificial intelligence control technology to monitor three aspects: solid precipitation monitoring video, temperature value and pressure value in the distiller. Dynamic feature analysis is performed on the aspect, and then the temperature and pressure can be adaptively adjusted based on the actual situation of the object being distilled during the control of the temperature and pressure at the current time point, so as to improve energy utilization and improve the efficiency and purity of the precipitated solids.
  • the energy management control system 200 for the preparation of lithium hexafluorophosphate according to the embodiment of the present application can be implemented in various terminal devices, such as the server of the energy management control algorithm for the preparation of lithium hexafluorophosphate, etc.
  • the energy management control system 200 for the preparation of lithium hexafluorophosphate according to the embodiment of the present application can be integrated into the terminal device as a software module and/or hardware module.
  • the energy management control system 200 for the preparation of lithium hexafluorophosphate can be a software module in the operating system of the terminal equipment, or can be an application developed for the terminal equipment; of course, the energy management control system 200 for the preparation of lithium hexafluorophosphate
  • the energy management control system 200 can also be one of many hardware modules of the terminal device.
  • the energy management control system 200 for the preparation of lithium hexafluorophosphate and the terminal device can also be separate devices, and the energy management control system 200 for the preparation of lithium hexafluorophosphate can be connected via wired and/or wireless devices.
  • the network is connected to the terminal device and transmits interactive information according to the agreed data format.
  • Figure 4 illustrates a flow chart of a control method of an energy management control system for lithium hexafluorophosphate preparation.
  • the control method of the energy management control system for the preparation of lithium hexafluorophosphate according to the embodiment of the present application includes the step of: S110, obtaining the solids collected by the underwater camera deployed in the distiller for a predetermined period of time.
  • Separation monitoring video S120, obtain the temperature values and pressure values at multiple predetermined time points of the predetermined time period collected by the temperature sensor and pressure sensor deployed in the distiller; S130, convert the multiple predetermined time points to After the temperature values and the pressure values at the plurality of predetermined time points are arranged into a temperature input vector and a pressure input vector respectively, the vector product between the temperature input vector and the transposed vector of the pressure input vector is calculated to obtain temperature - Pressure time series correlation matrix; S140, pass the temperature-pressure time series correlation matrix through the first convolutional neural network as a filter to obtain the correlation feature vector; S150, pass the solid precipitation monitoring video of the predetermined time period through the use of a three-dimensional volume Accumulate the second convolutional neural network of the kernel to obtain the extracted feature vector; S160, fuse the associated feature vector and the extracted feature vector to obtain a classification feature vector; and, S170, pass the classification feature vector through a classifier to obtain Classification results, the classification results are used to indicate that the temperature at the current time point should increase
  • FIG. 5 illustrates a schematic architectural diagram of a control method of an energy management control system for lithium hexafluorophosphate preparation according to an embodiment of the present application.
  • the temperature values obtained at the plurality of predetermined time points (for example, as shown in Figure 5 P1) and the pressure values of the plurality of predetermined time points (for example, P2 as shown in Figure 5) are respectively arranged as a temperature input vector (for example, V1 as shown in Figure 5) and a pressure input vector (for example, as shown in Figure 5 , V2 as shown in Figure 5), calculate the vector product between the temperature input vector and the transposed vector of the pressure input vector to obtain the temperature-pressure time series correlation matrix (for example, as shown in Figure 5 M); then, pass the temperature-pressure time series correlation matrix through the first convolutional neural network as a filter (for example, CNN1 as shown in Figure 5) to obtain the correlation feature vector (
  • steps S110 and S120 the solid precipitation monitoring video of a predetermined period of time collected by the underwater camera deployed in the distiller is obtained, and the temperature sensor and pressure sensor deployed in the distiller are obtained.
  • the actual dynamic situation of the object to be distilled is determined by using the monitoring video of the precipitated solid, and the dynamic change characteristic information of the temperature and pressure of the actual environment is collected through the temperature sensor and the pressure sensor, so that in In the control strategy of energy management, the temperature and pressure can be adaptively adjusted based on the actual conditions of the object being distilled to improve energy utilization and improve the efficiency and purity of the precipitated solids. That is, specifically, first, the solid precipitation monitoring video of a predetermined period of time is collected through an underwater camera deployed in the distiller, and the temperature sensor and pressure sensor deployed in the distiller are used to collect the video of the predetermined period of time. Temperature values and pressure values at multiple predetermined time points.
  • steps S130 and S140 after arranging the temperature values at the plurality of predetermined time points and the pressure values at the plurality of predetermined time points into a temperature input vector and a pressure input vector respectively, the temperature is calculated
  • the vector product between the input vector and the transposed vector of the pressure input vector is used to obtain the temperature-pressure time series correlation matrix, and the temperature-pressure time series correlation matrix is passed through the first convolutional neural network as a filter to obtain the correlation Feature vector.
  • the temperature values at the multiple predetermined time points and the pressure values at the multiple predetermined time points are respectively arranged into a temperature input vector and After the pressure input vector is obtained, the vector product between the temperature input vector and the transposed vector of the pressure input vector is calculated to obtain a temperature-pressure time series correlation matrix. Further, the temperature-pressure time series correlation matrix is processed through the first convolutional neural network as a filter to extract the temperature values of the multiple predetermined time points and the temperature values of the multiple predetermined time points. The implicit correlation feature distribution information of the pressure value is obtained, thereby obtaining the correlation feature vector.
  • step S150 the solid precipitation monitoring video of the predetermined time period is passed through a second convolutional neural network using a three-dimensional convolution kernel to obtain a precipitation feature vector.
  • the solid precipitation monitoring video of the predetermined time period has the implicit dynamic change characteristics of solid precipitation in the time series dimension, in order to extract such dynamic implicit characteristics, in the technical solution of the present application , and further process the solid precipitation monitoring video of the predetermined time period through a second convolutional neural network using a three-dimensional convolution kernel to obtain a precipitation feature vector.
  • the associated feature vector and the extracted feature vector are fused to obtain a classification feature vector, and the classification feature vector is passed through a classifier to obtain a classification result.
  • the classification result is expressed by Yu indicates that the temperature at the current time point should increase or decrease, and the pressure at the current time point should increase or decrease. It should be understood that considering whether it is the two-dimensional convolution kernel as a filter of the first convolutional neural network that extracts temperature-pressure correlation, or the second volume that extracts image semantic-temporal correlation features in the video The three-dimensional convolution kernel of the convolutional neural network all obtains feature vectors as expressions of deep associated semantic features.
  • the correlation feature vector and the separation feature vector are fused to obtain a classification feature vector.
  • the classification feature vector can be processed through a classifier to obtain a representation that the temperature at the current time point should increase or decrease, and the pressure at the current time point should increase or decrease. classification results.
  • the classifier is used to process the classification feature vector with the following formula to obtain the classification result, wherein the formula is: softmax ⁇ (W n ,B n ):... :(W 1 ,B 1 )
  • control method of the energy management control system for the preparation of lithium hexafluorophosphate based on the embodiments of the present application has been clarified, which adopts artificial intelligence control technology to monitor the video, temperature value and pressure value in the distiller from the solid precipitation Dynamic feature analysis is performed in three aspects, and then the temperature and pressure can be adaptively adjusted based on the actual situation of the object being distilled in the control of the temperature and pressure at the current time point, so as to improve energy utilization and improve the efficiency and efficiency of precipitated solids. purity.
  • embodiments of the present application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to execute the “exemplary method” described above in this specification.
  • the steps in the functions of the control method of the energy management control system for the preparation of lithium hexafluorophosphate according to various embodiments of the present application are described in the section.
  • the computer program product can be used to write program codes for performing the operations of the embodiments of the present application in any combination of one or more programming languages, including object-oriented programming languages, such as Java, C++, etc. , also includes conventional procedural programming languages, such as the "C" language or similar programming languages.
  • the program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server execute on.
  • embodiments of the present application may also be a computer-readable storage medium having computer program instructions stored thereon.
  • the computer program instructions When the computer program instructions are run by a processor, the computer program instructions cause the processor to execute the above-mentioned "example method" part of this specification.
  • the computer-readable storage medium may be any combination of one or more readable media.
  • the readable medium may be a readable signal medium or a readable storage medium.
  • Readable storage media may include, for example, but are not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: electrical connection with one or more conductors, portable disk, hard disk, random access memory (RAM), read only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • 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)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Secondary Cells (AREA)

Abstract

本申请涉及智能生产控制的领域，其具体地公开了一种用于六氟磷酸锂制备的能源管理控制系统及其控制方法，其通过采用人工智能控制技术，来从固体析出监控视频、蒸馏器内的温度值和压强值三个方面进行动态地特征分析，进而在对于当前时间点的温度和压强的控制中能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。

Description

用于六氟磷酸锂制备的能源管理控制系统及其控制方法 技术领域
本发明涉智能生产控制的领域，且更为具体地，涉及一种用于六氟磷酸锂制备的能源管理控制系统及其控制方法。
背景技术
锂离子电池具有平台电压高、循环性能好、工作温度范围宽、能量密度大、无记忆效应等优点，广泛应用于移动通讯、便携式电子产品、电动工具、武器装配等领域，目前在动力汽车中也具有广阔的应用前景。电解液和其中的电解质是锂电池的关键原材料，并且直接影响所制备电池的性能，因此电解质必须满足：电导率高、化学及电化学性能稳定、可使用温度范围宽和价格低。六氟磷酸锂满足以上要求，并可作为优良的电解质。
由于材料科学的不断发展，制备六氟磷酸锂的方案不断涌现。在不同的制备方案中，制备方案的能源管理问题是近年来逐渐被关注的话题，其一方面要求制备方案能够满足节能要求，另一方面，受限于制备工艺自身，需制备出满足质量要求的六氟磷酸锂。也就是，在六氟磷酸锂的制备方案中，期待提供一种能源管理控制系统，其能够在满足制备出满足质量要求的产物的前提下，尽可能地节约能耗。
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磷酸锂制备的能源管理控制系统及其控制方法，其通过采用人工智能控制技术，来从固体析出监控视频、蒸馏器内的温度值和压强值三个方面进行动态地特征分析，进而在对于当前时间点的温度和压强的控制中能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
根据本申请的一个方面，提供了一种用于六氟磷酸锂制备的能源管理控制系统，其包括：固体析出监控视频采集模块，用于获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；蒸馏器内状态数据采集模块，用于获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；
温度和压强数据编码模块，用于将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算 所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；
关联编码模块，用于将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；
视频编码模块，用于将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；
特征分布融合模块，用于融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及
能源管理结果生成模块，用于将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
在上述用于六氟磷酸锂制备的能源管理控制系统中，所述关联编码模块，进一步用于所述第一卷积神经网络的各层在层的正向传递中分别进行：对输入数据进行卷积处理以得到卷积特征图；对所述卷积特征图进行基于局部特征矩阵的均值池化以得到池化特征图；以及，对所述池化特征图进行非线性激活以得到激活特征图；其中，所述第一卷积神经网络的最后一层的输出为所述关联特征向量，所述第一卷积神经网络的第一层的输入为所述温度-压强时序关联矩阵。
在上述用于六氟磷酸锂制备的能源管理控制系统中，所述视频编码模块，包括：采样单元，用于以预定采样频率从所述预定时间段的固体析出监控视频中提取多个关键帧；多层卷积单元，用于将所述多个关键帧通过所述第二卷积神经网络的多层卷积层以得到析出特征图；以及，全局均值池化单元，用于对所述析出特征图进行基于特征矩阵的全局均值池化处理以得到所述析出特征向量。
在上述用于六氟磷酸锂制备的能源管理控制系统中，所述多层卷积单元，进一步用于使用所述第二卷积神经网络的多层卷积层在层的正向传递中对输入数据分别进行基于所述三维卷积核的卷积处理、池化处理和非线性激活处理以由所述多层卷积层的最后一层卷积层输出所述析出特征图。
在上述用于六氟磷酸锂制备的能源管理控制系统中，所述多层卷积层的各层使用的非线性激活函数为Mish函数，所述Mish函数用公式表示为：f(x)＝x.tanh(ln(1+e x))。
在上述用于六氟磷酸锂制备的能源管理控制系统中，所述特征分布融合模块，进一步用于：以如下公式融合所述关联特征向量和所述析出特征向量以得到所述分类特征向量；
其中，所述公式为：
Figure PCTCN2022116744-appb-000001
其中，V c表示所述分类特征向量，V 1表示所述关联特征向量，V 2表示所述析出特征向量，||·|| 1表示向量的一范数，且||·|| F表示矩阵的Frobenius范数，⊙表示按位置点乘、
Figure PCTCN2022116744-appb-000002
表示按位置相加，
Figure PCTCN2022116744-appb-000003
表示按位置相减。
在上述用于六氟磷酸锂制备的能源管理控制系统中，所述能源管理结果生成模块，进一步用于使用所述分类器以如下公式对所述分类特征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为所述分类特征向量。
根据本申请的另一方面，一种用于六氟磷酸锂制备的能源管理控制系统的控制方法，其包括：
获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；
获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；
将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；
将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；
将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；
融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及
将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
在上述用于六氟磷酸锂制备的能源管理控制系统的控制方法中，将所述 温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量，包括：所述第一卷积神经网络的各层在层的正向传递中分别进行：对输入数据进行卷积处理以得到卷积特征图；对所述卷积特征图进行基于局部特征矩阵的均值池化以得到池化特征图；以及，对所述池化特征图进行非线性激活以得到激活特征图；其中，所述第一卷积神经网络的最后一层的输出为所述关联特征向量，所述第一卷积神经网络的第一层的输入为所述温度-压强时序关联矩阵。
在上述用于六氟磷酸锂制备的能源管理控制系统的控制方法中，将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量，包括：以预定采样频率从所述预定时间段的固体析出监控视频中提取多个关键帧；将所述多个关键帧通过所述第二卷积神经网络的多层卷积层以得到析出特征图；对所述析出特征图进行基于特征矩阵的全局均值池化处理以得到所述析出特征向量。
在上述用于六氟磷酸锂制备的能源管理控制系统的控制方法中，将所述多个关键帧通过所述第二卷积神经网络的多层卷积层以得到析出特征图，包括：使用所述第二卷积神经网络的多层卷积层在层的正向传递中对输入数据分别进行基于所述三维卷积核的卷积处理、池化处理和非线性激活处理以由所述多层卷积层的最后一层卷积层输出所述析出特征图。
在上述用于六氟磷酸锂制备的能源管理控制系统的控制方法中，所述多层卷积层的各层使用的非线性激活函数为Mish函数，所述Mish函数用公式表示为：f(x)＝x.tanh(ln(1+e x))。
在上述用于六氟磷酸锂制备的能源管理控制系统的控制方法中，融合所述关联特征向量和所述析出特征向量以得到分类特征向量，包括：以如下公式融合所述关联特征向量和所述析出特征向量以得到所述分类特征向量；
其中，所述公式为：
Figure PCTCN2022116744-appb-000004
其中，V c表示所述分类特征向量，V 1表示所述关联特征向量，V 2表示所述析出特征向量，||·|| 1表示向量的一范数，且||·|| F表示矩阵的Frobenius范数，⊙表示按位置点乘、
Figure PCTCN2022116744-appb-000005
表示按位置相加，
Figure PCTCN2022116744-appb-000006
表示按位置相减。
在上述用于六氟磷酸锂制备的能源管理控制系统的控制方法中，将所述分类特征向量通过分类器以得到分类结果，包括：使用所述分类器以如下公 式对所述分类特征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为所述分类特征向量。
根据本申请的又一方面，提供了一种计算机可读介质，其上存储有计算机程序指令，所述计算机程序指令在被处理器运行时使得所述处理器执行如上所述的用于六氟磷酸锂制备的能源管理控制系统的控制方法。
与现有技术相比，本申请提供的用于六氟磷酸锂制备的能源管理控制系统及其控制方法，其通过采用人工智能控制技术，来从固体析出监控视频、蒸馏器内的温度值和压强值三个方面进行动态地特征分析，进而在对于当前时间点的温度和压强的控制中能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
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磷酸锂制备的能源管理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磷酸锂制备的能源管理控制系统的框图。
图3为根据本申请实施例的用于六氟磷酸锂制备的能源管理控制系统中视频编码模块的框图。
图4为根据本申请实施例的用于六氟磷酸锂制备的能源管理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磷酸锂制备的能源管理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，锂离子电池具有平台电压高、循环性能好、工作温度范围宽、能量密度大、无记忆效应等优点，广泛应用于移动通讯、便携式电子产品、电动工具、武器装配等领域，目前在动力汽车中也具有广阔的应用前景。电解液和其中的电解质是锂电池的关键原材料，并且直接影响所制备电池的性能，因此电解质必须满足：电导率高、化学及电化学性能稳定、可使用温度范围宽和价格低。六氟磷酸锂满足以上要求，并可作为优良的电解质。
由于材料科学的不断发展，制备六氟磷酸锂的方案不断涌现。在不同的制备方案中，制备方案的能源管理问题是近年来逐渐被关注的话题，其一方面要求制备方案能够满足节能要求，另一方面，受限于制备工艺自身，需制备出满足质量要求的六氟磷酸锂。也就是，在六氟磷酸锂的制备方案中，期待提供一种能源管理控制系统，其能够在满足制备出满足质量要求的产物的前提下，尽可能地节约能耗。
在现有的六氟磷酸锂的制备方案中，其包括以下步骤：
S1：将五氧化二磷固体溶解分散在有机溶剂当中，配置成五氧化二磷为1wt％～100wt％的混合溶液，再加入F:P物质的量比为4.0～6.0:1的氟化合物，反应温度控制在-30℃～240℃，制得五氟化磷气体，干燥；
S2：将氟化锂固体悬浮在有机溶剂中，配置成F:P物质的量比为1:0.5～2的氟化锂悬浮液，并将温度控制在5℃～120℃；
S3：将制得的五氟化磷气体与氟化锂悬浮液持续反应富集并保持搅拌0.5～10小时，得到澄清溶液；
S4：将澄清溶液蒸发浓缩，通过分离，得到大量固体，并在10℃～120℃，0.01MPa～0.99MPa条件下干燥该固体0.5～10小时；
S5：将干燥后固体饱和溶解在有机溶剂中，得到澄清溶液；
S6：重复步骤S4，得到高纯六氟磷酸锂。
其中，上述的有机溶剂为碳酸酯类、烷烃类、醚类、含氮有机物中的至少一种。
相应地，本申请发明人考虑到在该制备方案中，步骤S4中的温度和压强控制是该制备方案的能源控制的关键步骤。应可以理解，在适当的温度和压强控制下，固体析出的效率会提高且析出固体的纯度会更优。但是，在现有的温度和压强控制策略中，采用预定控制程序来设定不同阶段的温度和压 强值，也就是，在现有的能源管理控制策略中，以固定的程序来控制各个阶段的压强和温度，而没有基于被蒸馏对象的实际情况来自适应地采取温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
基于此，在本申请的技术方案中，通过使用析出固体的监控视频来判断被蒸馏对象的实际动态情况，并通过温度传感器和压强传感器来采集实际环境的温度和压强的动态变化特征信息，这样在能源管理的控制策略中，能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
也就是，具体地，首先，通过部署于所述蒸馏器内的水下摄像头采集预定时间段的固体析出监控视频，并且通过部署于蒸馏器内的温度传感器和压强传感器采集所述预定时间段的多个预定时间点的温度值和压强值。应可以理解，考虑到温度和压强会相互影响，也就是，例如根据理想气体状态方程PV＝NTR，温度和压强成正比的关系。因此，在本申请的技术方案中，为了挖掘出这种实际的关联特征信息，将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵。进一步地，再将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络中进行处理，以提取出所述多个预定时间点的温度值和所述多个预定时间点的压强值的隐含关联特征分布信息，从而得到关联特征向量。
并且，考虑到所述预定时间段的固体析出监控视频是在时序维度上具有固体析出的隐含动态变化特征，因此为了提取出这种动态的隐含特征，将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络进行处理，以得到析出特征向量。
应可以理解，考虑到无论是提取温度-压强关联的所述第一卷积神经网络的作为过滤器的二维卷积核，还是提取视频中的图像语义-时序关联特征的所述第二卷积神经网络的三维卷积核，都获得作为深层关联语义特征表达的特征向量，而由于特征维度和语义表达上的差异，在特征融合之前，优选地将所述关联特征向量V 1和所述析出特征向量V 2进行深度单应对齐。
基于此，获得融合后的分类特征向量为：
Figure PCTCN2022116744-appb-000007
其中，V c表示所述分类特征向量，V 1表示所述关联特征向量，V 2表示所述析出特征向量，||·|| 1表示向量的一范数，且||·|| F表示矩阵的Frobenius范数，⊙表示按位置点乘、
Figure PCTCN2022116744-appb-000008
表示按位置相加，
Figure PCTCN2022116744-appb-000009
表示按位置相减。
这样，该融合后的所述分类特征向量通过根据融合向量表征的特征深度信息特性，将特征向量进行基于场景深度流的单应性对齐，并添加基于向量的全场景单应关联矩阵的深度感知的偏置项，从而将特征向量在可能存在特征分布差异的基础上进行稠密的深度融合，以获得具有更好的分类效果的分类特征向量V c，进而也就能够提高分类的准确性。
基于此，本申请提出了一种用于六氟磷酸锂制备的能源管理控制系统，其包括：固体析出监控视频采集模块，用于获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；蒸馏器内状态数据采集模块，用于获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；温度和压强数据编码模块，用于将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；关联编码模块，用于将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；视频编码模块，用于将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；特征分布融合模块，用于融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及，能源管理结果生成模块，用于将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
图1图示了根据本申请实施例的用于六氟磷酸锂制备的能源管理控制系统的应用场景图。如图1所示，在该应用场景中，首先，通过部署于所述蒸馏器(例如，如图1中所示意的D)内的水下摄像头(例如，如图1中所示意的C)采集预定时间段的固体析出监控视频，并且通过部署于蒸馏器内的温度传感器(例如，如图1中所示意的T)和压强传感器(例如，如图1中所示意的P)采集所述预定时间段的多个预定时间点的温度值和压强值。然后，将获得的所述预定时间段的固体析出监控视频以及所述预定时间段的多个预定时间点的温度值和压强值输入至部署有用于六氟磷酸锂制备的能源 管理控制算法的服务器中(例如，如图1中所示意的云服务器S)，其中，所述服务器能够以用于六氟磷酸锂制备的能源管理控制算法对所述预定时间段的固体析出监控视频以及所述预定时间段的多个预定时间点的温度值和压强值进行处理，以生成用于表示当前时间点的温度应增大或应减小，以及，当前时间点的压强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磷酸锂制备的能源管理控制系统的框图。如图2所示，根据本申请实施例的用于六氟磷酸锂制备的能源管理控制系统200，包括：固体析出监控视频采集模块210，用于获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；蒸馏器内状态数据采集模块220，用于获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；温度和压强数据编码模块230，用于将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；关联编码模块240，用于将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；视频编码模块250，用于将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；特征分布融合模块260，用于融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及，能源管理结果生成模块270，用于将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
具体地，在本申请实施例中，所述固体析出监控视频采集模块210和所述蒸馏器内状态数据采集模块220，用于获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频，并获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值。如前所述，考虑到在现有的制备方案中，温度和压强的控制是 该所述制备方案的能源控制的关键步骤。应可以理解，在适当的温度和压强控制下，固体析出的效率会提高且析出固体的纯度会更优。但是，在现有的温度和压强控制策略中，采用预定控制程序来设定不同阶段的温度和压强值，也就是，在现有的能源管理控制策略中，以固定的程序来控制各个阶段的压强和温度，而没有基于被蒸馏对象的实际情况来自适应地采取温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
因此，在本申请的技术方案中，通过使用析出固体的监控视频来判断被蒸馏对象的实际动态情况，并通过温度传感器和压强传感器来采集实际环境的温度和压强的动态变化特征信息，这样在能源管理的控制策略中，能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。也就是，具体地，首先，通过部署于所述蒸馏器内的水下摄像头采集预定时间段的固体析出监控视频，并且通过部署于蒸馏器内的温度传感器和压强传感器采集所述预定时间段的多个预定时间点的温度值和压强值。
具体地，在本申请实施例中，所述温度和压强数据编码模块230和所述关联编码模块240，用于将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵，并将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量。应可以理解，考虑到温度和压强会相互影响，也就是，例如根据理想气体状态方程PV＝NTR，温度和压强成正比的关系。因此，在本申请的技术方案中，为了挖掘出这种实际的关联特征信息，将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵。进一步地，再将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络中进行处理，以提取出所述多个预定时间点的温度值和所述多个预定时间点的压强值的隐含关联特征分布信息，从而得到关联特征向量。
更具体地，在本申请实施例中，所述关联编码模块，进一步用于：所述第一卷积神经网络的各层在层的正向传递中分别进行：对输入数据进行卷积处理以得到卷积特征图；对所述卷积特征图进行基于局部特征矩阵的均值池 化以得到池化特征图；以及，对所述池化特征图进行非线性激活以得到激活特征图；其中，所述第一卷积神经网络的最后一层的输出为所述关联特征向量，所述第一卷积神经网络的第一层的输入为所述温度-压强时序关联矩阵。
具体地，在本申请实施例中，所述视频编码模块250，用于将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量。应可以理解，考虑到所述预定时间段的固体析出监控视频是在时序维度上具有固体析出的隐含动态变化特征，因此为了提取出这种动态的隐含特征，在本申请的技术方案中，进一步将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络进行处理，以得到析出特征向量。
更具体地，在本申请实施例中，所述视频编码模块，包括：首先，以预定采样频率从所述预定时间段的固体析出监控视频中提取多个关键帧。然后，将所述多个关键帧通过所述第二卷积神经网络的多层卷积层以得到析出特征图。相应地，在一个具体示例中，使用所述第二卷积神经网络的多层卷积层在层的正向传递中对输入数据分别进行基于所述三维卷积核的卷积处理、池化处理和非线性激活处理以由所述多层卷积层的最后一层卷积层输出所述析出特征图。特别地，所述多层卷积层的各层使用的非线性激活函数为Mish函数，所述Mish函数用公式表示为：f(x)＝x.tanh(ln(1+e x))。应可以理解，与ReLU函数相比，Mish对负值的轻微允许会通过更好的梯度流，而不像ReLU中那样存在硬性的零边界。不同于ReLU，Mish在零处的导数存在，平滑的函数曲线会允许更好的信息进入神经网络，从而得到更好的准确性和泛化性。最后，对所述析出特征图进行基于特征矩阵的全局均值池化处理以得到所述析出特征向量。
图3图示了根据本申请实施例的用于六氟磷酸锂制备的能源管理控制系统中视频编码模块的框图。如图3所示，所述视频编码模块250，包括：采样单元251，用于以预定采样频率从所述预定时间段的固体析出监控视频中提取多个关键帧；多层卷积单元252，用于将所述多个关键帧通过所述第二卷积神经网络的多层卷积层以得到析出特征图；以及，全局均值池化单元253，用于对所述析出特征图进行基于特征矩阵的全局均值池化处理以得到所述析出特征向量。
具体地，在本申请实施例中，所述特征分布融合模块260和所述能源管 理结果生成模块270，用于融合所述关联特征向量和所述析出特征向量以得到分类特征向量，并将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。应可以理解，考虑到无论是提取温度-压强关联的所述第一卷积神经网络的作为过滤器的二维卷积核，还是提取视频中的图像语义-时序关联特征的所述第二卷积神经网络的三维卷积核，都获得作为深层关联语义特征表达的特征向量，而由于特征维度和语义表达上的差异，在特征融合之前，优选地将所述关联特征向量V 1和所述析出特征向量V 2进行深度单应对齐。也就是，在本申请的技术方案中，融合所述关联特征向量和所述析出特征向量以得到分类特征向量。进一步地，就可以将所述分类特征向量通过分类器中进行处理，以获得用于表示当前时间点的温度应增大或应减小，以及，当前时间点的压强应增大或应减小的分类结果。相应地，在一个具体示例中，使用所述分类器以如下公式对所述分类特征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为所述分类特征向量。
更具体地，在本申请实施例中，所述特征分布融合模块，进一步用于：以如下公式融合所述关联特征向量和所述析出特征向量以得到所述分类特征向量；
其中，所述公式为：
Figure PCTCN2022116744-appb-000010
其中，V c表示所述分类特征向量，V 1表示所述关联特征向量，V 2表示所述析出特征向量，||·|| 1表示向量的一范数，且||·|| F表示矩阵的Frobenius范数，⊙表示按位置点乘、
Figure PCTCN2022116744-appb-000011
表示按位置相加，
Figure PCTCN2022116744-appb-000012
表示按位置相减。应可以理解，这样，该融合后的所述分类特征向量通过根据融合向量表征的特征深度信息特性，将所述特征向量进行基于场景深度流的单应性对齐，并添加基于向量的全场景单应关联矩阵的深度感知的偏置项，从而将所述特征向量在可能存在特征分布差异的基础上进行稠密的深度融合，以获得具有更好的分类效果的分类特征向量V c，进而也就能够提高分类的准确性。
综上，基于本申请实施例的所述用于六氟磷酸锂制备的能源管理控制系统200被阐明，其通过采用人工智能控制技术，来从固体析出监控视频、蒸馏器内的温度值和压强值三个方面进行动态地特征分析，进而在对于当前时 间点的温度和压强的控制中能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
如上所述，根据本申请实施例的用于六氟磷酸锂制备的能源管理控制系统200可以实现在各种终端设备中，例如用于六氟磷酸锂制备的能源管理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磷酸锂制备的能源管理控制系统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，该用于六氟磷酸锂制备的能源管理控制系统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，该用于六氟磷酸锂制备的能源管理控制系统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，该用于六氟磷酸锂制备的能源管理控制系统200与该终端设备也可以是分立的设备，并且该用于六氟磷酸锂制备的能源管理控制系统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图示了用于六氟磷酸锂制备的能源管理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磷酸锂制备的能源管理控制系统的控制方法，包括步骤：S110，获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；S120，获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；S130，将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；S140，将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；S150，将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；S160，融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及，S170，将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
图5图示了根据本申请实施例的用于六氟磷酸锂制备的能源管理控制系统的控制方法的架构示意图。如图5所示，在所述用于六氟磷酸锂制备的能 源管理控制系统的控制方法的网络架构中，首先，将获得的所述多个预定时间点的温度值(例如，如图5中所示意的P1)和所述多个预定时间点的压强值(例如，如图5中所示意的P2)分别排列为温度输入向量(例如，如图5中所示意的V1)和压强输入向量(例如，如图5中所示意的V2)后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵(例如，如图5中所示意的M)；接着，将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络(例如，如图5中所示意的CNN1)以得到关联特征向量(例如，如图5中所示意的VF1)；然后，将获得的所述预定时间段的固体析出监控视频(例如，如图5中所示意的Q)通过使用三维卷积核的第二卷积神经网络(例如，如图5中所示意的CNN2)以得到析出特征向量(例如，如图5中所示意的VF2)；接着，融合所述关联特征向量和所述析出特征向量以得到分类特征向量(例如，如图5中所示意的VF)；以及，最后，将所述分类特征向量通过分类器(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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
更具体地，在步骤S110和步骤S120中，获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频，并获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值。应可以理解，考虑到在现有的制备方案中，温度和压强的控制是该所述制备方案的能源控制的关键步骤。应可以理解，在适当的温度和压强控制下，固体析出的效率会提高且析出固体的纯度会更优。但是，在现有的温度和压强控制策略中，采用预定控制程序来设定不同阶段的温度和压强值，也就是，在现有的能源管理控制策略中，以固定的程序来控制各个阶段的压强和温度，而没有基于被蒸馏对象的实际情况来自适应地采取温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
因此，在本申请的技术方案中，通过使用析出固体的监控视频来判断被蒸馏对象的实际动态情况，并通过温度传感器和压强传感器来采集实际环境的温度和压强的动态变化特征信息，这样在能源管理的控制策略中，能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。也就是，具体地，首先，通过部署于所述蒸馏 器内的水下摄像头采集预定时间段的固体析出监控视频，并且通过部署于蒸馏器内的温度传感器和压强传感器采集所述预定时间段的多个预定时间点的温度值和压强值。
更具体地，在步骤S130和步骤S140中，将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵，并将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量。应可以理解，考虑到温度和压强会相互影响，也就是，例如根据理想气体状态方程PV＝NTR，温度和压强成正比的关系。因此，在本申请的技术方案中，为了挖掘出这种实际的关联特征信息，将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵。进一步地，再将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络中进行处理，以提取出所述多个预定时间点的温度值和所述多个预定时间点的压强值的隐含关联特征分布信息，从而得到关联特征向量。
更具体地，在步骤S150中，将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量。应可以理解，考虑到所述预定时间段的固体析出监控视频是在时序维度上具有固体析出的隐含动态变化特征，因此为了提取出这种动态的隐含特征，在本申请的技术方案中，进一步将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络进行处理，以得到析出特征向量。
更具体地，在步骤S160和步骤S170中，融合所述关联特征向量和所述析出特征向量以得到分类特征向量，并将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。应可以理解，考虑到无论是提取温度-压强关联的所述第一卷积神经网络的作为过滤器的二维卷积核，还是提取视频中的图像语义-时序关联特征的所述第二卷积神经网络的三维卷积核，都获得作为深层关联语义特征表达的特征向量，而由于特征维度和语义表达上的差异，在特征融合之前，优选地将所述关联特征向量V 1和所述析出特征向量V 2进行深度单应对齐。也就是，在本申请的技术方案中，融合所述关联 特征向量和所述析出特征向量以得到分类特征向量。进一步地，就可以将所述分类特征向量通过分类器中进行处理，以获得用于表示当前时间点的温度应增大或应减小，以及，当前时间点的压强应增大或应减小的分类结果。相应地，在一个具体示例中，使用所述分类器以如下公式对所述分类特征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为所述分类特征向量。
综上，基于本申请实施例的所述用于六氟磷酸锂制备的能源管理控制系统的控制方法被阐明，其通过采用人工智能控制技术，来从固体析出监控视频、蒸馏器内的温度值和压强值三个方面进行动态地特征分析，进而在对于当前时间点的温度和压强的控制中能够基于被蒸馏对象的实际情况来自适应地调整温度和压强，以提高能源利用率且提高析出固体的效率和纯度。
示例性计算机程序产品和计算机可读存储介质
除了上述方法和设备以外，本申请的实施例还可以是计算机程序产品，其包括计算机程序指令，所述计算机程序指令在被处理器运行时使得所述处理器执行本说明书上述“示例性方法”部分中描述的根据本申请各种实施例的用于六氟磷酸锂制备的能源管理控制系统的控制方法中的功能中的步骤。
所述计算机程序产品可以以一种或多种程序设计语言的任意组合来编写用于执行本申请实施例操作的程序代码，所述程序设计语言包括面向对象的程序设计语言，诸如Java、C++等，还包括常规的过程式程序设计语言，诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户计算设备上部分在远程计算设备上执行、或者完全在远程计算设备或服务器上执行。
此外，本申请的实施例还可以是计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令在被处理器运行时使得所述处理器执行本说明书上述“示例性方法”部分中描述的用于六氟磷酸锂制备的能源管理控制系统的控制方法中的步骤。
所述计算机可读存储介质可以采用一个或多个可读介质的任意组合。可读介质可以是可读信号介质或者可读存储介质。可读存储介质例如可以包括 但不限于电、磁、光、电磁、红外线、或半导体的系统、系统或器件，或者任意以上的组合。可读存储介质的更具体的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。
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. 一种用于六氟磷酸锂制备的能源管理控制系统，其特征在于，包括：
    固体析出监控视频采集模块，用于获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；蒸馏器内状态数据采集模块，用于获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；温度和压强数据编码模块，用于将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；关联编码模块，用于将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；视频编码模块，用于将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；特征分布融合模块，用于融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及能源管理结果生成模块，用于将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
  2. 根据权利要求1所述的用于六氟磷酸锂制备的能源管理控制系统，其中，所述关联编码模块，进一步用于所述第一卷积神经网络的各层在层的正向传递中分别进行：对输入数据进行卷积处理以得到卷积特征图；对所述卷积特征图进行基于局部特征矩阵的均值池化以得到池化特征图；以及对所述池化特征图进行非线性激活以得到激活特征图；其中，所述第一卷积神经网络的最后一层的输出为所述关联特征向量，所述第一卷积神经网络的第一层的输入为所述温度-压强时序关联矩阵。
  3. 根据权利要求2所述的用于六氟磷酸锂制备的能源管理控制系统，其中，所述视频编码模块，包括：采样单元，用于以预定采样频率从所述预定时间段的固体析出监控视频中提取多个关键帧；多层卷积单元，用于将所述多个关键帧通过所述第二卷积神经网络的多层卷积层以得到析出特征图；以及全局均值池化单元，用于对所述析出特征图进行基于特征矩阵的全局均值池化处理以得到所述析出特征向量。
  4. 根据权利要求3所述的用于六氟磷酸锂制备的能源管理控制系统，其中，所述多层卷积单元，进一步用于使用所述第二卷积神经网络的多层卷积层在层的正向传递中对输入数据分别进行基于所述三维卷积核的卷积处 理、池化处理和非线性激活处理以由所述多层卷积层的最后一层卷积层输出所述析出特征图。
  5. 根据权利要求4所述的用于六氟磷酸锂制备的能源管理控制系统，其中，所述多层卷积层的各层使用的非线性激活函数为Mish函数，所述Mish函数用公式表示为：f(x)＝x.tanh(ln(1+e x))。
  6. 根据权利要求5所述的用于六氟磷酸锂制备的能源管理控制系统，其中，所述特征分布融合模块，进一步用于：以如下公式融合所述关联特征向量和所述析出特征向量以得到所述分类特征向量；其中，所述公式为：
    Figure PCTCN2022116744-appb-100001
    其中，V c表示所述分类特征向量，V 1表示所述关联特征向量，V 2表示所述析出特征向量，||·|| 1表示向量的一范数，且||·|| F表示矩阵的Frobenius范数，⊙表示按位置点乘、
    Figure PCTCN2022116744-appb-100002
    表示按位置相加，
    Figure PCTCN2022116744-appb-100003
    表示按位置相减。
  7. 根据权利要求6所述的用于六氟磷酸锂制备的能源管理控制系统，其中，所述能源管理结果生成模块，进一步用于使用所述分类器以如下公式对所述分类特征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为所述分类特征向量。
  8. 一种用于六氟磷酸锂制备的能源管理控制系统的控制方法，其特征在于，包括：获取由部署于所述蒸馏器内的水下摄像头采集的预定时间段的固体析出监控视频；获取由部署于蒸馏器内的温度传感器和压强传感器采集的所述预定时间段的多个预定时间点的温度值和压强值；将所述多个预定时间点的温度值和所述多个预定时间点的压强值分别排列为温度输入向量和压强输入向量后，计算所述温度输入向量和所述压强输入向量的转置向量之间的向量乘积以得到温度-压强时序关联矩阵；将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量；将所述预定时间段的固体析出监控视频通过使用三维卷积核的第二卷积神经网络以得到析出特征向量；融合所述关联特征向量和所述析出特征向量以得到分类特征向量；以及将所述分类特征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温度应增大或应减小，以及，当前时间点的压强应增大或应减小。
  9. 根据权利要求8所述的用于六氟磷酸锂制备的能源管理控制系统的 控制方法，其中，将所述温度-压强时序关联矩阵通过作为过滤器的第一卷积神经网络以得到关联特征向量，包括：所述第一卷积神经网络的各层在层的正向传递中分别进行：对输入数据进行卷积处理以得到卷积特征图；对所述卷积特征图进行基于局部特征矩阵的均值池化以得到池化特征图；以及对所述池化特征图进行非线性激活以得到激活特征图；其中，所述第一卷积神经网络的最后一层的输出为所述关联特征向量，所述第一卷积神经网络的第一层的输入为所述温度-压强时序关联矩阵。
  10. 根据权利要求9所述的用于六氟磷酸锂制备的能源管理控制系统的控制方法，其中，融合所述关联特征向量和所述析出特征向量以得到分类特征向量，包括：以如下公式融合所述关联特征向量和所述析出特征向量以得到所述分类特征向量；其中，所述公式为：
    Figure PCTCN2022116744-appb-100004
    其中，V c表示所述分类特征向量，V 1表示所述关联特征向量，V 2表示所述析出特征向量，||·|| 1表示向量的一范数，且||·|| F表示矩阵的Frobenius范数，⊙表示按位置点乘、
    Figure PCTCN2022116744-appb-100005
    表示按位置相加，
    Figure PCTCN2022116744-appb-100006
    表示按位置相减。
PCT/CN2022/116744 2022-06-30 2022-09-02 用于六氟磷酸锂制备的能源管理控制系统及其控制方法 WO2024000800A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210772608.0 2022-06-30
CN202210772608.0A CN115079572B (zh) 2022-06-30 2022-06-30 用于六氟磷酸锂制备的能源管理控制系统及其控制方法

Publications (1)

Publication Number Publication Date
WO2024000800A1 true WO2024000800A1 (zh) 2024-01-04

Family

ID=83258636

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/116744 WO2024000800A1 (zh) 2022-06-30 2022-09-02 用于六氟磷酸锂制备的能源管理控制系统及其控制方法

Country Status (2)

Country Link
CN (1) CN115079572B (zh)
WO (1) WO2024000800A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117669838A (zh) * 2024-01-31 2024-03-08 江西荧光磁业有限公司 钕铁硼磁铁生产的优化控制系统及方法

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115328228B (zh) * 2022-10-13 2023-04-07 新乡市合力鑫电源有限公司 一种高频开关电源
CN115586755B (zh) * 2022-10-28 2023-05-30 福建省龙德新能源有限公司 用于六氟磷酸锂制备的生产管理控制系统及其方法
CN115754107B (zh) * 2022-11-08 2023-06-23 福建省龙德新能源有限公司 用于六氟磷酸锂制备的自动化采样分析系统及其方法
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
CN115936682B (zh) * 2022-12-21 2023-11-07 江西有源工业废物回收处理有限公司 印制线路板的废料回收系统及其方法
CN116000297A (zh) * 2023-01-03 2023-04-25 赣州市光华有色金属有限公司 高强度钨镧丝的制备装置及其方法
CN116110507B (zh) * 2023-02-15 2024-02-06 浙江宏丰炉料有限公司 镁碳砖智能化生产方法及系统
CN116059681B (zh) * 2023-03-07 2023-06-09 福建省龙德新能源有限公司 用于控制结晶成核的超声控制方法及其系统
CN116785967B (zh) * 2023-04-28 2024-03-08 福建省龙德新能源有限公司 用于电子级六氟磷酸锂制备的自动配料系统
CN116907214B (zh) * 2023-05-09 2024-03-08 广东夏和瓷业有限公司 环保日用陶瓷的制备工艺及其系统
CN116854111B (zh) * 2023-05-23 2024-05-14 福建省龙德新能源有限公司 电子级六氟磷酸锂的纯化方法及其系统
CN116859848B (zh) * 2023-07-12 2024-01-26 普洛赛斯工程技术(宁波)有限公司 甲基异丁基酮生产装置的控制系统及其方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07182009A (ja) * 1993-12-22 1995-07-21 Kobe Steel Ltd 制御装置
CN106882819A (zh) * 2017-03-29 2017-06-23 东营石大胜华新能源有限公司 一种六氟磷酸锂的制备纯化方法
CN113139436A (zh) * 2021-03-31 2021-07-20 哈尔滨端点科技发展有限公司 基于图像识别的火锅温度自动控制方法、系统及介质
CN113731321A (zh) * 2021-10-08 2021-12-03 广西埃索凯循环科技有限公司 一种基于多传感技术的硫酸锌的生产系统
CN113856235A (zh) * 2021-09-29 2021-12-31 浙江大华技术股份有限公司 降温结晶控制方法、装置、电子设备和系统
CN114475350A (zh) * 2022-02-14 2022-05-13 杭州鸽然科技有限公司 智能充电系统及其工作方法
CN114548246A (zh) * 2022-02-11 2022-05-27 杭州吾好科技有限公司 智能设备的控制方法、系统和电子设备
CN114870416A (zh) * 2022-04-28 2022-08-09 福建德尔科技股份有限公司 用于电子级一氟甲烷制备的精馏控制系统及精馏控制方法

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102380220A (zh) * 2011-10-10 2012-03-21 上海师范大学 一种乙烯精馏塔塔釜乙烯浓度的测量方法
CA2874610A1 (en) * 2012-05-25 2013-11-28 Lanxess Deutschland Gmbh High-purity lithium hexafluorophosphate
JP5960907B2 (ja) * 2013-03-25 2016-08-02 三井化学株式会社 五フッ化リンの製造方法及び六フッ化リン酸リチウムの製造方法
CN104965967B (zh) * 2015-03-26 2018-02-13 华东理工大学 一种常减压蒸馏装置的收率实时预测方法
CN110052050B (zh) * 2019-04-29 2021-03-16 中国石油大学(华东) 基于塔板温度的内部热耦合精馏塔全局状态观测器及方法
CN112295255B (zh) * 2020-10-24 2021-11-02 四川泸天化创新研究院有限公司 一种甲醇精馏装置智能控制系统及控制方法
CN112365045A (zh) * 2020-11-09 2021-02-12 上海明华电力科技有限公司 一种基于大数据的主蒸汽温度智能预测方法
CN112486686A (zh) * 2020-11-30 2021-03-12 之江实验室 基于云边协同的定制化深度神经网络模型压缩方法及系统
CN113783527A (zh) * 2021-09-26 2021-12-10 中机国能电力工程有限公司 一种基于光伏组件温度和辐照度校验的数据挖掘方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07182009A (ja) * 1993-12-22 1995-07-21 Kobe Steel Ltd 制御装置
CN106882819A (zh) * 2017-03-29 2017-06-23 东营石大胜华新能源有限公司 一种六氟磷酸锂的制备纯化方法
CN113139436A (zh) * 2021-03-31 2021-07-20 哈尔滨端点科技发展有限公司 基于图像识别的火锅温度自动控制方法、系统及介质
CN113856235A (zh) * 2021-09-29 2021-12-31 浙江大华技术股份有限公司 降温结晶控制方法、装置、电子设备和系统
CN113731321A (zh) * 2021-10-08 2021-12-03 广西埃索凯循环科技有限公司 一种基于多传感技术的硫酸锌的生产系统
CN114548246A (zh) * 2022-02-11 2022-05-27 杭州吾好科技有限公司 智能设备的控制方法、系统和电子设备
CN114475350A (zh) * 2022-02-14 2022-05-13 杭州鸽然科技有限公司 智能充电系统及其工作方法
CN114870416A (zh) * 2022-04-28 2022-08-09 福建德尔科技股份有限公司 用于电子级一氟甲烷制备的精馏控制系统及精馏控制方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117669838A (zh) * 2024-01-31 2024-03-08 江西荧光磁业有限公司 钕铁硼磁铁生产的优化控制系统及方法
CN117669838B (zh) * 2024-01-31 2024-05-10 江西荧光磁业有限公司 钕铁硼磁铁生产的优化控制系统及方法

Also Published As

Publication number Publication date
CN115079572A (zh) 2022-09-20
CN115079572B (zh) 2023-02-03

Similar Documents

Publication Publication Date Title
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
CN115453990B (zh) 用于氟化铵生产的生产管理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
CN115586755B (zh) 用于六氟磷酸锂制备的生产管理控制系统及其方法
Alipour et al. A review on temperature-dependent electrochemical properties, aging, and performance of lithium-ion cells
WO2022151819A1 (zh) 一种基于聚类分析的电池系统在线故障诊断方法和系统
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
CN114475350A (zh) 智能充电系统及其工作方法
CN115896817A (zh) 氟氮混合气的生产方法及其系统
Park et al. Probabilistic representations for video contrastive learning
WO2024007445A1 (zh) 用于氟化锂制备的能源管理控制系统及其控制方法
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
Fu et al. Comparison of lithium-ion anode materials using an experimentally verified physics-based electrochemical model
CN115754107B (zh) 用于六氟磷酸锂制备的自动化采样分析系统及其方法
CN111680644B (zh) 一种基于深度空时特征学习的视频行为聚类方法
CN110826500A (zh) 基于运动链接空间的对抗性网络估计3d人体姿态的方法
Peng et al. Pixelwise adaptive discretization with uncertainty sampling for depth completion
Yang et al. Study of algorithm for aerial target detection based on lightweight neural network
Xu et al. Temperature field prediction of lithium-ion batteries using improved local tangent space alignment
CN117455451A (zh) 一种草甘膦水解溶剂的回收方法
WO2023226236A1 (zh) 用于电子级氢氟酸制备的能源管理控制系统及其控制方法
WO2024020994A1 (zh) 用于电芯的缺陷检测模型的训练方法和训练装置
CN107330512B (zh) 预测蛋白质序列的远同源性关系的系统及方法
CN114817847A (zh) 一种基于多层前馈神经网络的储能电站智能监测方法
CN115201685A (zh) 基于自编码器和对抗学习的可迁移电池热失控风险评估系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22948890

Country of ref document: EP

Kind code of ref document: A1