Intelligent control system and method for gas-liquid separator

Abstract

The present application discloses an intelligent control system and method for a gas-liquid separator, which collects data of gas phase pressure and liquid level in real time through a sensor network, and introduces a data processing and analysis algorithm based on artificial intelligence and deep learning at the back end to analyze the coordination and time series correlation of the meteorological pressure and liquid level time series data, so as to learn and capture the fluctuation correlation relationship and interactive implicit feature information of meteorological pressure and liquid level in time series, so as to perform time series characterization of the working condition of the gas-liquid separator, so as to adaptively adjust the initial value of the valve opening. In this way, the initial value of the valve opening can be automatically corrected according to the fluctuation error of the pressure and liquid level of the gas-liquid separator, so as to realize intelligent control of the gas-liquid separator to adapt to the changing working conditions.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to an intelligent control system and method for a gas-liquid separator. Background Art

[0002] Gas-liquid separators play a vital role in industrial production, especially in the fields of petrochemicals and natural gas processing. Gas-liquid separators can separate gas and liquid in mixed fluids so that these two substances can be effectively processed and utilized in the production process.

[0003] However, traditional gas-liquid separator control schemes usually adopt fixed control strategies. This control method is usually based on fixed set values ​​or rules and lacks the ability to adapt to dynamic changes in the system. That is to say, in actual working conditions, the operating conditions of the gas-liquid separator may change constantly, resulting in the inability of traditional control schemes to be flexibly adjusted. In addition, some existing methods for regulating and controlling gas-liquid separators often rely on manual experience or simple control algorithms. This method has a low degree of intelligence and cannot be intelligently adjusted according to real-time working conditions and complex system dynamic characteristics, resulting in low control accuracy, which makes the equipment run inefficiently, causing energy waste and increased costs. With the rapid development of industrial automation and intelligence, traditional control methods can no longer meet the intelligent needs of modern industry for efficient, precise and adaptive control.

[0004] Therefore, an intelligent control system for a gas-liquid separator is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an intelligent control system and method for a gas-liquid separator, which collects data on gas phase pressure and liquid level in real time through a sensor network, and introduces a data processing and analysis algorithm based on artificial intelligence and deep learning at the back end to perform collaborative and time-series correlation analysis on the meteorological pressure and liquid level time series data, so as to learn and capture the fluctuation correlation relationship and interactive implicit feature information of meteorological pressure and liquid level in time series, so as to perform time series characterization of the working condition of the gas-liquid separator, so as to adaptively adjust the initial value of the valve opening. In this way, the initial value of the valve opening can be automatically corrected according to the fluctuation error of the pressure and liquid level of the gas-liquid separator, thereby realizing intelligent control of the gas-liquid separator to adapt to the ever-changing working conditions.

[0006] According to one aspect of the present application, an intelligent control system for a gas-liquid separator is provided, which includes: a gas phase separation data acquisition module, which is used to obtain a time series of gas phase pressure and a time series of liquid level height collected by a sensor network; a gas phase separation data time series change rate calculation module, which is used to calculate the gas phase pressure change rate and the liquid level height change rate based on the time series of the gas phase pressure and the time series of the liquid level height to obtain a time series of the gas phase pressure change rate and a time series of the liquid level height change rate; a gas phase separation data time series encoding module, which is used to respectively encode the time series of the gas phase pressure change rate and the liquid level height change rate. The time series of the height change rate is time-related encoded to obtain the gas phase pressure fluctuation time-related implicit feature vector and the liquid level height fluctuation time-related implicit feature vector; a multimodal data time-related feature threshold balance fusion module is used to input the gas phase pressure fluctuation time-related implicit feature vector and the liquid level height fluctuation time-related implicit feature vector into the balance threshold feature vector adaptive fusion module to obtain the working condition multimodal time-related representation vector; a valve opening initial value adjustment module is used to determine the control instruction based on the working condition multimodal time-related representation vector, and the control instruction is used to represent the initial value of the valve opening.

[0007] In the intelligent control system of the above-mentioned gas-liquid separator, the gas phase separation data time series change rate calculation module includes: a meteorological pressure time series change rate calculation unit, which is used to calculate the change rate of the gas phase pressure at each time point in the time series of the gas phase pressure to obtain the time series of the gas phase pressure change rate; a liquid level height time series change rate calculation unit, which is used to calculate the change rate of the liquid level height at each time point in the time series of the liquid level height to obtain the time series of the liquid level height change rate.

[0008] In the intelligent control system of the above-mentioned gas-liquid separator, the meteorological pressure time series change rate calculation unit is used to: respectively calculate the difference between the gas phase pressure at each time point in the time series of the gas phase pressure and the meteorological pressure at adjacent time points to obtain the time series of the gas phase pressure change rate.

[0009] In the intelligent control system of the above-mentioned gas-liquid separator, the gas phase separation data timing encoding module is used to: input the time series of the gas phase pressure change rate and the time series of the liquid level height change rate into a sequence encoder based on the Bi-LSTM model to obtain the gas phase pressure fluctuation timing-related implicit feature vector and the liquid level height fluctuation timing-related implicit feature vector.

[0010] In the intelligent control system of the above-mentioned gas-liquid separator, the multimodal data time series feature threshold balance fusion module includes: a meteorological pressure fluctuation-liquid level height fluctuation time series feature fusion unit, which is used to perform threshold constraints on the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector based on different linear transformations and fusion methods to obtain the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector, the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; threshold balance A semantic constraint expression unit is used to use the respective position feature values ​​in the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector as the first threshold value, use the respective position feature values ​​in the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector as the second threshold value, use the respective position feature values ​​in the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector as the third threshold value, calculate the weighted sum of the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector to obtain the multimodal time series correlation representation vector of the working condition.

[0011] In the intelligent control system of the gas-liquid separator, the meteorological pressure fluctuation-liquid level height fluctuation time series feature fusion unit is used to: cascade the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector to obtain a first gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; calculate the matrix multiplication of the first gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and the first transformation matrix, and then add the obtained feature vector to the first bias vector by position to obtain a first linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; use The activation function activates the first linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector are added by position to obtain the second gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; after calculating the matrix multiplication of the second gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector and the second transformation matrix, the obtained feature vector is added by position to the second bias vector to obtain the second linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; use The activation function activates the second linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector are multiplied by position point to obtain the third gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; after calculating the matrix multiplication of the third gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector and the third transformation matrix, the obtained feature vector is added to the third bias vector by position to obtain the third linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; use The activation function activates the third linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector.

[0012] In the intelligent control system of the above-mentioned gas-liquid separator, the threshold balance semantic constraint expression unit is used to: calculate the mean between the first threshold value, the second threshold value and the third threshold value to obtain the threshold mean; calculate the product of the eigenvalue of each position in the gas phase pressure fluctuation time series associated implicit feature vector and one minus the threshold mean to obtain the weighted gas phase pressure fluctuation time series associated implicit feature vector; calculate the product of the eigenvalue of each position in the liquid level height fluctuation time series associated implicit feature vector and the threshold mean to obtain the weighted liquid level height fluctuation time series associated implicit feature vector; add the weighted gas phase pressure fluctuation time series associated implicit feature vector and the weighted liquid level height fluctuation time series associated implicit feature vector by position to obtain the multi-modal time series associated representation vector of the working condition.

[0013] In the intelligent control system of the above-mentioned gas-liquid separator, the valve opening initial value adjustment module is used to: input the multi-modal time-series association representation vector of the operating condition into the valve opening controller based on the decoder to obtain a control instruction, and the control instruction is used to indicate the adjustment of the valve opening initial value to the decoded value.

[0014] According to another aspect of the present application, an intelligent control method for a gas-liquid separator is provided, which includes: obtaining a time series of gas phase pressure and a time series of liquid level height collected by a sensor network; based on the time series of the gas phase pressure and the time series of the liquid level height, respectively calculating the gas phase pressure change rate and the liquid level height change rate to obtain a time series of the gas phase pressure change rate and a time series of the liquid level height change rate; respectively performing time series correlation encoding on the time series of the gas phase pressure change rate and the time series of the liquid level height change rate to obtain a gas phase pressure fluctuation time series correlation implicit feature vector and a liquid level height fluctuation time series correlation implicit feature vector; inputting the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector into an adaptive fusion module of an equalization threshold feature vector to obtain a multi-modal time series correlation representation vector of an operating condition; and determining a control instruction based on the multi-modal time series correlation representation vector of an operating condition, wherein the control instruction is used to represent an initial value of a regulating valve opening.

[0015] Compared with the prior art, the intelligent control system and method of a gas-liquid separator provided by the present application monitors and collects data of gas phase pressure and liquid level in real time through a sensor network, and introduces a data processing and analysis algorithm based on artificial intelligence and deep learning at the back end to perform collaborative and time-series correlation analysis on the meteorological pressure and liquid level time series data, so as to learn and capture the fluctuation correlation relationship and interactive implicit feature information of meteorological pressure and liquid level in time series, thereby performing time series characterization of the working condition of the gas-liquid separator, so as to adaptively adjust the initial value of the valve opening. In this way, the initial value of the valve opening can be automatically corrected according to the fluctuation error of the pressure and liquid level of the gas-liquid separator, thereby realizing intelligent control of the gas-liquid separator to adapt to the changing working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 4 is a block diagram of an intelligent control system of a gas-liquid separator according to an embodiment of the present application.

[0018] Figure 2 4 is a system architecture diagram of an intelligent control system for a gas-liquid separator according to an embodiment of the present application.

[0019] Figure 3It is a block diagram of a multimodal data timing feature threshold equalization fusion module in an intelligent control system of a gas-liquid separator according to an embodiment of the present application.

[0020] Figure 4 Flow chart of an intelligent control method for a gas-liquid separator according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0022] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0023] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0024] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0025] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0026] Traditional gas-liquid separator control schemes usually adopt fixed control strategies. This control method is usually based on fixed set values ​​or rules and lacks the ability to adapt to dynamic changes in the system. That is to say, in actual working conditions, the operating conditions of the gas-liquid separator may change constantly, resulting in the inability of traditional control schemes to be flexibly adjusted. In addition, some existing methods for regulating and controlling gas-liquid separators often rely on artificial experience or simple control algorithms. This method has a low degree of intelligence and cannot be intelligently adjusted according to real-time working conditions and complex system dynamic characteristics, resulting in low control accuracy, which makes the equipment run inefficiently, causing energy waste and increased costs. With the rapid development of industrial automation and intelligence, traditional control methods can no longer meet the intelligent needs of modern industry for efficient, accurate and adaptive control. Therefore, an intelligent control system for gas-liquid separators is desired.

[0027] In the technical solution of the present application, an intelligent control system for a gas-liquid separator is proposed. Figure 1 4 is a block diagram of an intelligent control system of a gas-liquid separator according to an embodiment of the present application. Figure 2 : is a system architecture diagram of an intelligent control system for a gas-liquid separator according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the intelligent control system 300 of the gas-liquid separator includes: a gas phase separation data acquisition module 310, which is used to obtain the time series of the gas phase pressure and the time series of the liquid level height collected by the sensor network; a gas phase separation data time series change rate calculation module 320, which is used to calculate the gas phase pressure change rate and the liquid level height change rate based on the time series of the gas phase pressure and the time series of the liquid level height to obtain the time series of the gas phase pressure change rate and the time series of the liquid level height change rate; a gas phase separation data time series encoding module 330, which is used to respectively encode the time series of the gas phase pressure change rate and the liquid level height change rate The time series of the level change rate is time-correlated encoded to obtain the gas phase pressure fluctuation time-correlated implicit feature vector and the liquid level height fluctuation time-correlated implicit feature vector; the multimodal data time-series feature threshold equalization fusion module 340 is used to input the gas phase pressure fluctuation time-series associated implicit feature vector and the liquid level height fluctuation time-series associated implicit feature vector into the equalization threshold feature vector adaptive fusion module to obtain the operating condition multimodal time-series associated representation vector; the valve opening initial value adjustment module 350 is used to determine the control instruction based on the operating condition multimodal time-series associated representation vector, and the control instruction is used to represent the initial value of the valve opening.

[0028] In particular, the gas phase separation data acquisition module 310 is used to obtain the time series of gas phase pressure and liquid level height collected by the sensor network. It should be understood that the operating state and working condition of the gas-liquid separator are dynamically changing, and real-time monitoring of the change rate of pressure and liquid level can help the system capture these dynamic characteristics more accurately.

[0029] In particular, the gas phase separation data time series change rate calculation module 320 is used to calculate the gas phase pressure change rate and the liquid level change rate based on the gas phase pressure time series and the liquid level time series to obtain the gas phase pressure change rate time series and the liquid level change rate time series. Considering that in the process of automatically correcting the initial value of the valve opening by monitoring the operating state and working condition of the gas-liquid separator, the time series change rate of the gas phase pressure and the liquid level can provide information about the speed and trend of the system state change, which helps to better understand the behavior of the fluid in the separator and the working condition of the gas-liquid separator. Based on this, in the technical solution of the present application, the gas phase separation data time series change rate calculation module includes: a meteorological pressure time series change rate calculation unit, which is used to calculate the change rate of the gas phase pressure at each time point in the time series of the gas phase pressure to obtain the time series of the gas phase pressure change rate; a liquid level time series change rate calculation unit, which is used to calculate the change rate of the liquid level at each time point in the time series of the liquid level to obtain the time series of the liquid level change rate. More specifically, in a specific example of the present application, the meteorological pressure time series change rate calculation unit is used to: respectively calculate the difference between the gas phase pressure at each time point in the time series of the gas phase pressure and the meteorological pressure at the adjacent time point to obtain the time series of the gas phase pressure change rate. That is, the difference between the gas phase pressure at each time point and the liquid level at each time point and the gas phase pressure and liquid level at the adjacent time point is calculated as the change rate information. These change rates reflect the actual changes in the system operation process, which helps to understand and characterize the working conditions of the gas-liquid separator and provide a basis for the task of correcting the initial value of the valve opening.

[0030] In particular, the gas phase separation data time series encoding module 330 is used to perform time series correlation encoding on the time series of the gas phase pressure change rate and the time series of the liquid level height change rate, respectively, to obtain the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector. When analyzing the time series data of the gas phase pressure change rate and the liquid level height change rate to identify the working condition of the gas-liquid separator, it is considered that the gas phase pressure change rate and the liquid level height change rate both have a time series dynamic correlation and an implicit change trend in the time dimension. Therefore, in order to better capture the time series dynamic characteristics and implicit pattern information of the gas phase pressure change rate and the liquid level height change rate, in the technical solution of the present application, the gas phase separation data time series encoding module is used to: input the time series of the gas phase pressure change rate and the time series of the liquid level height change rate into a sequence encoder based on the Bi-LSTM model to obtain the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector. It should be understood that the Bi-LSTM model is a deep learning model suitable for processing sequence data, which can capture the long-term temporal dependencies in the data. In the operation of the gas-liquid separator, the past operating conditions have an important impact on the current and future conditions, and Bi-LSTM can use these time dependencies for more accurate analysis and prediction. Based on this, by using the sequence encoder based on the Bi-LSTM model for processing, the temporal dependencies and dynamic temporal change characteristics of the gas phase pressure change rate and the liquid level height change rate in the time dimension can be captured respectively, which helps to better analyze the actual working conditions of the gas-liquid separator, thereby correcting the initial value of the valve opening.

[0031] In particular, the multimodal data time series feature threshold balance fusion module 340 is used to input the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector into the balance threshold feature vector adaptive fusion module to obtain the working condition multimodal time series associated representation vector. It should be understood that since the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector respectively contain the time series fluctuation implicit features of the gas phase pressure and the time series fluctuation implicit features of the liquid level height, and the gas phase pressure and the liquid level height are important parameters in the operation process of the gas-liquid separator, they can each reflect the state changes of different aspects within the system. Therefore, in order to comprehensively utilize the complementary information and fusion characteristics between the gas phase pressure time series fluctuation characteristics and the liquid level height time series fluctuation characteristics to realize the multimodal comprehensive characterization of the gas-liquid separator working condition, in the technical solution of the present application, the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector are further input into the equalization threshold feature vector adaptive fusion module to obtain the multimodal time series associated representation vector of the working condition. Through the processing of the balanced threshold feature vector adaptive fusion module, the hidden correlation relationship and interaction between the gas phase pressure time series fluctuation characteristics and the liquid level height time series fluctuation characteristics of the gas-liquid separator can be learned and captured, and the importance semantics related to the gas-liquid separator working condition characterization and the valve opening initial value correction task can be learned. The expression of key features is adaptively selected through balanced threshold constraints, that is, in order to better utilize these key importance semantics in the fusion process of the implicit feature vector associated with the gas phase pressure fluctuation time series and the implicit feature vector associated with the liquid level height fluctuation time series, that is, to adjust the weights of different feature vectors according to actual conditions, so that information from different sources can be reasonably balanced in the fusion process, so as to better adapt to the complexity and diversity of the system operation state, and provide a more reliable basis for the subsequent generation of control instructions for correcting the initial value of the valve opening. In particular, in a specific example of the present application, such as Figure 3As shown, the multimodal data time series feature threshold equalization fusion module 340 includes: a meteorological pressure fluctuation-liquid level height fluctuation time series feature fusion unit 341, which is used to perform threshold constraints on the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector based on different linear transformations and fusion methods to obtain a first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector, a second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and a third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; threshold equalization semantic constraint expression Unit 342 is used to use the respective position characteristic values ​​in the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the first threshold value, use the respective position characteristic values ​​in the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the second threshold value, and use the respective position characteristic values ​​in the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the third threshold value, and calculate the weighted sum of the gas phase pressure fluctuation time series correlation implicit characteristic vector and the liquid level height fluctuation time series correlation implicit characteristic vector to obtain the multi-modal time series correlation representation vector of the operating condition.

[0032] Specifically, the meteorological pressure fluctuation-liquid level height fluctuation time series feature fusion unit 341 is used to perform threshold constraints on the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector based on different linear transformations and fusion methods to obtain the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector, the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector. In a specific example of the present application, first, the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector are cascaded to obtain the first gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; then, after calculating the matrix multiplication of the first gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and the first transformation matrix, the obtained feature vector is added to the first bias vector by position to obtain the first linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; use The activation function activates the first linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; then, the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector are added by position to obtain the second gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; after calculating the matrix multiplication of the second gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector and the second transformation matrix, the obtained feature vector is added by position to the second bias vector to obtain the second linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; use The activation function activates the second linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; then, the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector are multiplied by position point to obtain the third gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; after calculating the matrix multiplication of the third gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector and the third transformation matrix, the obtained feature vector is added to the third bias vector by position to obtain the third linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; then, use The activation function activates the third linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector.

[0033] Specifically, the threshold equalization semantic constraint expression unit 342 is used to use the respective position characteristic values ​​in the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the first threshold value, use the respective position characteristic values ​​in the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the second threshold value, and use the respective position characteristic values ​​in the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the third threshold value, calculate the weighted sum of the gas phase pressure fluctuation time series correlation implicit characteristic vector and the liquid level height fluctuation time series correlation implicit characteristic vector to obtain the multimodal time series correlation representation vector of the working condition. In a specific example of the present application, first, the mean among the first threshold value, the second threshold value and the third threshold value is calculated to obtain the threshold mean; then, the product of the eigenvalue of each position in the gas phase pressure fluctuation time series associated implicit feature vector and one minus the threshold mean is calculated to obtain the weighted item gas phase pressure fluctuation time series associated implicit feature vector; then, the product of the eigenvalue of each position in the liquid level height fluctuation time series associated implicit feature vector and the threshold mean is calculated to obtain the weighted item liquid level height fluctuation time series associated implicit feature vector; then, the weighted item gas phase pressure fluctuation time series associated implicit feature vector and the weighted item liquid level height fluctuation time series associated implicit feature vector are added by position to obtain the multi-modal time series associated representation vector of the operating condition.

[0034] In summary, in the above embodiment, the implicit feature vector associated with the timing of the gas phase pressure fluctuation and the implicit feature vector associated with the timing of the liquid level height fluctuation are input into the adaptive fusion module of the balanced threshold feature vector to obtain the multi-modal timing association representation vector of the working condition, including: inputting the implicit feature vector associated with the timing of the gas phase pressure fluctuation and the implicit feature vector associated with the timing of the liquid level height fluctuation into the adaptive fusion module of the balanced threshold feature vector to process them with the following adaptive balanced fusion formula to obtain the multi-modal timing association representation vector of the working condition; wherein the adaptive balanced fusion formula is: ;in, and are respectively the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector, and are the implicit eigenvectors of the gas phase pressure fluctuation time series correlation The implicit feature vector associated with the liquid level height fluctuation time series The eigenvalues ​​at each position in , is the characteristic value of each position in the multi-modal time series association representation vector of the working condition, represents the concatenation of vectors, , and is the transformation matrix, , and is the bias vector, is the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation eigenvector, represents vector addition by position, It means vector multiplication by position. is the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation eigenvector, is the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation eigenvector, is the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector The threshold value of each position in is the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector The threshold value of each position in is the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector The threshold value of each position in , express Activation function.

[0035] It is worth mentioning that in other specific examples of the present application, the implicit feature vector associated with the timing of the gas phase pressure fluctuation and the implicit feature vector associated with the timing of the liquid level height fluctuation can also be input into the balanced threshold feature vector adaptive fusion module in other ways to obtain the multi-modal timing association representation vector of the working condition, for example: input the implicit feature vector associated with the timing of the gas phase pressure fluctuation and the implicit feature vector associated with the timing of the liquid level height fluctuation; input the implicit feature vector associated with the timing of the gas phase pressure fluctuation and the implicit feature vector associated with the timing of the liquid level height fluctuation into the balanced threshold feature vector adaptive fusion module for fusion processing; during the fusion process, considering the multi-modality of the working condition, the module will weigh and adjust according to the importance and contribution of different features to ensure that the fused representation vector can fully characterize the characteristics of the working condition; after being processed by the balanced threshold feature vector adaptive fusion module, the multi-modal timing association representation vector of the working condition is obtained.

[0036] In particular, the valve opening initial value adjustment module 350 is used to determine the control instruction based on the working condition multimodal time series association representation vector, and the control instruction is used to represent the initial value of the valve opening. That is, in the technical solution of the present application, the valve opening initial value adjustment module is used to: input the working condition multimodal time series association representation vector into the valve opening controller based on the decoder to obtain the control instruction, and the control instruction is used to represent the initial value of the valve opening to the decoded value. In other words, the multimodal time series fusion association representation feature information of the working condition of the gas-liquid separator is used to perform decoding regression, so as to adaptively adjust the initial value of the valve opening. In this way, the initial value of the valve opening can be automatically corrected according to the fluctuation error of the pressure and liquid level of the gas-liquid separator, thereby realizing intelligent control of the gas-liquid separator to adapt to the constantly changing working conditions.

[0037] Here, considering that the implicit feature vector of the gas phase pressure fluctuation timing association and the implicit feature vector of the liquid level height fluctuation timing association respectively express the short-range-long-range bidirectional timing association characteristics of the gas phase pressure and the liquid level height, when performing balanced threshold adaptive fusion, the timing feature distribution difference caused by the source timing distribution difference will cause the balanced threshold local distribution gating difference, so that the working condition multimodal timing association representation vector has a local overflow feature, thereby affecting the decoding regression interpretability of the working condition multimodal timing association representation vector for the decoder-based valve opening controller, affecting the accuracy of the decoding result. In a preferred embodiment, when the working condition multimodal timing association representation vector is input into the decoder-based valve opening controller for decoding, the present application optimizes the working condition multimodal timing association representation vector to obtain an optimized working condition multimodal timing association representation vector.

[0038] Specifically, the operating condition multimodal time series association representation vector is optimized to obtain an optimized operating condition multimodal time series association representation vector, including the steps of: calculating the autocorrelation matrix of the operating condition multimodal time series association representation vector, that is: ;in, represents the multi-modal temporal association representation vector of the working condition, represents the transpose of a vector, represents matrix multiplication, Denotes the autocorrelation matrix.

[0039] The multi-modal temporal correlation representation vector of the working condition is multiplied by the autocorrelation matrix in a query manner to obtain an intermediate feature vector, that is: ;in, represents the multi-modal temporal association representation vector of the working condition, represents the autocorrelation matrix, represents matrix multiplication, Denotes the intermediate eigenvector.

[0040] Calculate the autocorrelation matrix Row vector and Inner product sum of row vectors distance, to obtain the first weight matrix and the second weight matrix , expressed as: ;in, and The autocorrelation matrix is ​​represented by Row vector and row vector, represents the transpose of a vector, represents matrix multiplication, represents the autocorrelation matrix, represents the first weight matrix, represents the first weight matrix The eigenvalues ​​of the positions, represents the second weight matrix, represents the second weight matrix The eigenvalues ​​of the positions, Represents the calculation vector distance.

[0041] The first weight matrix and the second weight matrix After matrix multiplication, further matrix multiplication is performed with the intermediate eigenvector to obtain the optimized working condition multi-modal time series association representation vector, that is: ;in, represents the first weight matrix, represents the second weight matrix, represents the intermediate eigenvector, represents matrix multiplication, The optimized multi-modal time series association representation vector of the working condition is represented. Finally, the optimized multi-modal time series association representation vector of the working condition is input into the valve opening controller based on the decoder to obtain the control instruction.

[0042] That is, for the bit-by-bit scattering correlation response space of the eigenvalue of the multi-modal time series correlation representation vector of the working condition, the deep embedding structure in the high-order correlation space is obtained through the joint embedding space of the vector inner product embedding space and the vector independent embedding space, and the decomposable correlation response sub-dimension of the bit-by-bit scattering correlation response space of the eigenvalue of the multi-modal time series correlation representation vector of the working condition is represented as a set of spatial sub-dimensions to further perform the dependent space mapping of the decomposable correlation response sub-dimension of the deep embedding structure of the multi-modal time series correlation representation vector of the working condition, so as to improve the interpretability of the solution space based on the correlation distribution of the decoding regression solution of the multi-modal time series correlation representation vector of the working condition through the complex dimensional representation of the correlation response, thereby improving the accuracy of the control instruction obtained by the valve opening controller based on the decoder when the multi-modal time series correlation representation vector of the working condition is input. In this way, the initial value of the valve opening can be corrected more accurately according to the fluctuation error of the pressure and liquid level of the gas-liquid separator, thereby realizing the intelligent control of the gas-liquid separator to adapt to the changing working conditions.

[0043] As described above, the intelligent control system 300 of the gas-liquid separator according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an intelligent control algorithm for the gas-liquid separator. In a possible implementation, the intelligent control system 300 of the gas-liquid separator according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the intelligent control system 300 of the gas-liquid separator can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the intelligent control system 300 of the gas-liquid separator can also be one of the many hardware modules of the wireless terminal.

[0044] Alternatively, in another example, the intelligent control system 300 of the gas-liquid separator and the wireless terminal may also be separate devices, and the intelligent control system 300 of the gas-liquid separator may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0045] Furthermore, an intelligent control method for a gas-liquid separator is also provided.

[0046] Figure 4 Flow chart of the intelligent control method of the gas-liquid separator according to the embodiment of the present application. Figure 4As shown, the intelligent control method of the gas-liquid separator according to the embodiment of the present application includes the steps of: S1, obtaining the time series of gas phase pressure and the time series of liquid level height collected by the sensor network; S2, based on the time series of the gas phase pressure and the time series of the liquid level height, respectively calculating the gas phase pressure change rate and the liquid level height change rate to obtain the time series of the gas phase pressure change rate and the time series of the liquid level height change rate; S3, respectively performing time series correlation encoding on the time series of the gas phase pressure change rate and the time series of the liquid level height change rate to obtain the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector; S4, inputting the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector into the equalization threshold feature vector adaptive fusion module to obtain the operating condition multi-modal time series correlation representation vector; S5, based on the operating condition multi-modal time series correlation representation vector, determining the control instruction, and the control instruction is used to represent the initial value of the regulating valve opening.

[0047] In summary, the intelligent control method of the gas-liquid separator according to the embodiment of the present application is explained, which collects data of gas phase pressure and liquid level in real time through a sensor network, and introduces data processing and analysis algorithms based on artificial intelligence and deep learning at the back end to conduct collaborative and time-series correlation analysis of the meteorological pressure and liquid level time series data, so as to learn and capture the fluctuation correlation relationship and interactive implicit feature information of meteorological pressure and liquid level in time series, so as to perform time series characterization of the working condition of the gas-liquid separator, so as to adaptively adjust the initial value of the valve opening. In this way, the initial value of the valve opening can be automatically corrected according to the fluctuation error of the pressure and liquid level of the gas-liquid separator, thereby realizing intelligent control of the gas-liquid separator to adapt to the ever-changing working conditions.

[0048] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent control system for a gas-liquid separator, characterized in that: include: A gas phase separation data acquisition module is used to obtain the time series of gas phase pressure and liquid level height collected by the sensor network; A gas phase separation data time series change rate calculation module is used to calculate the gas phase pressure change rate and the liquid level height change rate respectively based on the time series of the gas phase pressure and the time series of the liquid level height to obtain the time series of the gas phase pressure change rate and the time series of the liquid level height change rate; a gas phase separation data time series encoding module is used to perform time series correlation encoding on the time series of the gas phase pressure change rate and the time series of the liquid level height change rate respectively to obtain the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector; a multimodal data time series feature threshold balance fusion module is used to input the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector into the balance threshold feature vector adaptive fusion module to obtain the working condition multimodal time series correlation representation vector; a valve opening initial value adjustment module is used to determine the control instruction based on the working condition multimodal time series correlation representation vector, and the control instruction is used to represent the initial value of the valve opening.

2. The intelligent control system of the gas-liquid separator according to claim 1, characterized in that: The gas phase separation data time series change rate calculation module includes: a meteorological pressure time series change rate calculation unit, which is used to calculate the change rate of the gas phase pressure at each time point in the time series of the gas phase pressure to obtain the time series of the gas phase pressure change rate; a liquid level height time series change rate calculation unit, which is used to calculate the change rate of the liquid level height at each time point in the time series of the liquid level height to obtain the time series of the liquid level height change rate.

3. The intelligent control system of the gas-liquid separator according to claim 2, characterized in that: The meteorological pressure time series change rate calculation unit is used to respectively calculate the difference between the gas phase pressure at each time point in the time series of the gas phase pressure and the meteorological pressure at an adjacent time point to obtain the time series of the gas phase pressure change rate.

4. The intelligent control system of the gas-liquid separator according to claim 3, characterized in that: The gas phase separation data time series encoding module is used to: input the time series of the gas phase pressure change rate and the time series of the liquid level height change rate into a sequence encoder based on the Bi-LSTM model to obtain the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector.

5. The intelligent control system of the gas-liquid separator according to claim 4, characterized in that: The multimodal data time series feature threshold equalization fusion module includes: a meteorological pressure fluctuation-liquid level height fluctuation time series feature fusion unit, which is used to perform threshold constraints on the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector based on different linear transformations and fusion methods to obtain a first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector, a second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and a third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; a threshold equalization semantic constraint expression unit, It is used to use the various position characteristic values ​​in the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the first threshold value, use the various position characteristic values ​​in the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the second threshold value, and use the various position characteristic values ​​in the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation characteristic vector as the third threshold value, calculate the weighted sum of the gas phase pressure fluctuation time series correlation implicit characteristic vector and the liquid level height fluctuation time series correlation implicit characteristic vector to obtain the multi-modal time series correlation representation vector of the working condition.

6. The intelligent control system of the gas-liquid separator according to claim 5, characterized in that: The meteorological pressure fluctuation-liquid level height fluctuation time series feature fusion unit is used to: cascade the gas phase pressure fluctuation time series associated implicit feature vector and the liquid level height fluctuation time series associated implicit feature vector to obtain a first gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; calculate the matrix multiplication of the first gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector and the first transformation matrix, and then add the obtained feature vector to the first bias vector by position to obtain a first linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series associated feature vector; use The activation function activates the first linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the first activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector are added by position to obtain the second gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; after calculating the matrix multiplication of the second gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector and the second transformation matrix, the obtained feature vector is added by position to the second bias vector to obtain the second linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; use The activation function activates the second linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the second activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector are multiplied by position point to obtain the third gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; after calculating the matrix multiplication of the third gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector and the third transformation matrix, the obtained feature vector is added to the third bias vector by position to obtain the third linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector; use The activation function activates the third linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector to obtain the third activated linear transformation gas phase pressure fluctuation-liquid level height fluctuation time series correlation feature vector.

7. The intelligent control system of the gas-liquid separator according to claim 6, characterized in that: The threshold equalization semantic constraint expression unit is used to: calculate the mean between the first threshold value, the second threshold value and the third threshold value to obtain a threshold mean; calculate the product of the eigenvalue of each position in the gas phase pressure fluctuation time series associated implicit feature vector and one minus the threshold mean to obtain a weighted item gas phase pressure fluctuation time series associated implicit feature vector; calculate the product of the eigenvalue of each position in the liquid level height fluctuation time series associated implicit feature vector and the threshold mean to obtain a weighted item liquid level height fluctuation time series associated implicit feature vector; add the weighted item gas phase pressure fluctuation time series associated implicit feature vector and the weighted item liquid level height fluctuation time series associated implicit feature vector by position to obtain the multi-modal time series associated representation vector of the working condition.

8. The intelligent control system of the gas-liquid separator according to claim 7, characterized in that: The valve opening initial value adjustment module is used to: input the working condition multi-modal time series association representation vector into the valve opening controller based on the decoder to obtain a control instruction, and the control instruction is used to indicate the adjustment of the valve opening initial value to the decoded value.

9. An intelligent control method for a gas-liquid separator, characterized in that: include: Acquire the time series of gas phase pressure and the time series of liquid level height collected by the sensor network; based on the time series of gas phase pressure and the time series of liquid level height, calculate the gas phase pressure change rate and the liquid level height change rate respectively to obtain the time series of gas phase pressure change rate and the time series of liquid level height change rate; perform time series correlation encoding on the time series of gas phase pressure change rate and the time series of liquid level height change rate respectively to obtain the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector; input the gas phase pressure fluctuation time series correlation implicit feature vector and the liquid level height fluctuation time series correlation implicit feature vector into the equalization threshold feature vector adaptive fusion module to obtain the working condition multi-modal time series correlation representation vector; based on the working condition multi-modal time series correlation representation vector, determine the control instruction, and the control instruction is used to represent the initial value of the regulating valve opening.

Images