CN-121980248-A - Valve flow monitoring method and system

CN121980248ACN 121980248 ACN121980248 ACN 121980248ACN-121980248-A

Abstract

The invention discloses a valve flow monitoring method and a valve flow monitoring system, which belong to the technical field of flow monitoring and have the technical scheme that characteristic data corresponding to the current moment are obtained, the characteristic data comprise a temperature signal and a pressure signal, the characteristic data are preprocessed to obtain standardized characteristics, viscous pseudo-pressure difference, viscous temperature sensitive error and net temperature drift error are obtained according to the standardized characteristics and a preset model unit, effective pressure difference is obtained according to the viscous pseudo-pressure difference and the net temperature drift error, and flow corresponding to the current moment is obtained according to the effective pressure difference and the current viscosity.

Inventors

CHI ZHENPENG
MENG GUANGQUAN
ZHANG ZHENXUAN
FENG YUHUI
MA QIANG
LIANG SHULONG
MENG ZHAOCHENG
Lv Zhongfeng

Assignees

大连津利流体设备开发有限公司

Dates

Publication Date: 20260505
Application Date: 20260408

Claims (9)

1. A method of monitoring valve flow, comprising: Acquiring characteristic data corresponding to the current moment, wherein the characteristic data comprises a temperature signal and a pressure signal; preprocessing the feature data to obtain standardized features; obtaining viscous pseudo-pressure difference, viscous temperature sensitive error and net temperature drift error according to the standardized characteristics and a preset model unit; Obtaining an effective pressure difference according to the viscous pseudo pressure difference and the net temperature drift error; Obtaining the flow corresponding to the current moment according to the effective pressure difference and the current viscosity; The preset model unit comprises an LSTM model, a XGBoost model, a GPR model and an attention network model, and the viscous pseudo-pressure difference, the viscous temperature sensitivity error and the net temperature drift error are obtained according to the standardized characteristics and the preset model unit, and the method comprises the following steps: Obtaining the viscous pseudo-differential pressure according to the standardized characteristics, the LSTM model and the XGBoost model; obtaining a total temperature drift prediction result according to the standardized characteristics and the GPR model; Obtaining a viscous temperature sensitive error according to the standardized characteristics, the viscous pseudo-differential pressure and the attention network model; and obtaining the net temperature drift error according to the viscous temperature sensitivity error and the total temperature drift prediction result.
2. The method of claim 1, wherein the preprocessing the feature data to obtain standardized features comprises: Performing wavelet decomposition on the characteristic data to obtain a denoising signal; Obtaining the current viscosity according to the denoising signal and an Arrhenius formula; and obtaining a plurality of standardized features according to the denoising signals, wherein each standardized feature corresponds to one model in the preset model unit.
3. The method of claim 1, wherein said deriving said viscous pseudo-differential pressure based on said normalized features, said LSTM model and said XGBoost model comprises: obtaining a first error standard deviation, a second error standard deviation, a first predicted differential pressure and a second predicted differential pressure according to the standardized features, the LSTM model and the XGBoost model; obtaining posterior weights according to the first error standard deviation, the second error standard deviation and the Bayes theorem; and obtaining the viscous pseudo-differential pressure according to the posterior weight, the first predicted differential pressure and the second predicted differential pressure.
4. The method of claim 1, wherein the obtaining a total temperature drift prediction result according to the normalized feature and the GPR model comprises: obtaining a similarity vector according to the standardized features and the training set samples; obtaining labels corresponding to the training set samples and a nuclear matrix between the training set samples; and weighting the label according to the kernel matrix and the similarity vector to obtain the total temperature drift prediction result.
5. A method of valve flow monitoring according to claim 1, wherein said deriving a viscous temperature-sensitive error from said normalized characteristics, said viscous pseudo-differential pressure, and said attention network model comprises: Obtaining attention weight according to the standardized characteristics and the attention network model; And obtaining the viscous temperature sensitive error according to the attention weight and the viscous pseudo pressure difference.
6. The method of claim 1, wherein said deriving an effective differential pressure based on said viscous pseudo-differential pressure and said net temperature drift error comprises: Performing wavelet decomposition on the characteristic data to obtain a denoising signal; obtaining an initial pressure difference according to the denoising signal, the viscous pseudo pressure difference and the clear temperature drift error; And obtaining the effective pressure difference according to the initial pressure difference and a preset validity verification rule.
7. The method for monitoring the valve flow according to claim 1, wherein the obtaining the flow corresponding to the current moment according to the effective pressure difference and the current viscosity comprises: obtaining a corrected outflow coefficient according to the current viscosity and the correction coefficient; And obtaining the flow corresponding to the current moment according to the corrected outflow coefficient, the pore plate opening area and the effective pressure difference.
8. The method of valve flow monitoring of claim 4, wherein the step of training the GPR model comprises: Constructing a kernel function, wherein the kernel function comprises a square index kernel, a Matern kernel, a period kernel and a white noise kernel; screening the high-confidence pseudo tag according to the original marking data and the original GPR model to obtain an extended training set; obtaining super parameters according to a Gaussian process and a preset acquisition function, wherein the preset acquisition function is determined according to expected improvement; and obtaining a trained GPR model according to the kernel function, the extended training set and the super parameters.
9. A valve flow monitoring system for implementing a valve flow monitoring method according to any one of claims 1-8, comprising: The acquisition module is used for acquiring characteristic data corresponding to the current moment, wherein the characteristic data comprises a temperature signal and a pressure signal; The preprocessing module is used for preprocessing the characteristic data to obtain standardized characteristics; The error calculation module is used for obtaining viscous pseudo-pressure difference, viscous temperature sensitive error and net temperature drift error according to the standardized characteristics and a preset model unit; The differential pressure calculation module is used for obtaining an effective differential pressure according to the viscous pseudo differential pressure and the net temperature drift error; and the flow calculation module is used for obtaining the flow corresponding to the current moment according to the effective pressure difference and the current viscosity.

Description

Valve flow monitoring method and system Technical Field The invention relates to the technical field of flow monitoring, in particular to a valve flow monitoring method and a valve flow monitoring system. Background In the industrial fields of petrochemical industry, coal chemical industry and the like, pipeline conveying flow monitoring of high-viscosity media (such as petroleum), natural gas, steam and the like is one of the key links of production process regulation and control, energy consumption metering and process optimization. The basic principle of the flow monitoring device is that the throttling effect of fluid passing through a pore plate is utilized to generate pressure difference, and the actual flow is deduced through the association relation between the pressure difference and the flow. However, the high viscosity medium has the physical characteristics of strong viscosity and poor fluidity, a residual liquid film is easy to form on the valve cavity and the sensor surface, a small amount of impurities such as heavy hydrocarbon, moisture and sulfide contained in the natural gas are easy to be adsorbed or condensed on the dead angle of the valve cavity and the surface of the transmitter diaphragm to form a micro residual layer, steam and hot water are easy to be condensed due to Wen Bianchan, the problems can be treated only by manual dismantling and washing, periodic purging or fixed threshold compensation in the prior art, the interference caused by the problems cannot be accurately predicted and eliminated, the flow measurement result is finally distorted, and the accurate flow value is difficult to obtain, so the prior art has the defects. Disclosure of Invention Aiming at the defects existing in the prior art, the invention aims to provide a valve flow monitoring method and a system, which are used for capturing the coupling relation between dynamic time sequence dependence and static multifactor through an LSTM model and a XGBoost model, improving the prediction precision of pseudo pressure difference, fitting the total temperature drift through a GPR model, automatically learning the contribution weight of temperature to viscous errors by combining with a attention network model, realizing independent separation of the two types of errors, and finally obtaining an accurate flow value. In order to achieve the above purpose, the present invention provides the following technical solutions: The invention provides a valve flow monitoring method, which comprises the following steps: Acquiring characteristic data corresponding to the current moment, wherein the characteristic data comprises a temperature signal and a pressure signal; preprocessing the feature data to obtain standardized features; obtaining viscous pseudo-pressure difference, viscous temperature sensitive error and net temperature drift error according to the standardized characteristics and a preset model unit; Obtaining an effective pressure difference according to the viscous pseudo pressure difference and the net temperature drift error; And obtaining the flow corresponding to the current moment according to the effective pressure difference and the current viscosity. As a further improvement of the present invention, the preprocessing the feature data to obtain standardized features includes: Performing wavelet decomposition on the characteristic data to obtain a denoising signal; Obtaining the current viscosity according to the denoising signal and an Arrhenius formula; and obtaining a plurality of standardized features according to the denoising signals, wherein each standardized feature corresponds to one model in the preset model unit. As a further improvement of the present invention, the preset model unit includes an LSTM model, a XGBoost model, a GPR model, and an attention network model, and the obtaining, according to the standardized feature and the preset model unit, a viscous pseudo-pressure difference, a viscous temperature sensitivity error, and a net temperature drift error includes: Obtaining the viscous pseudo-differential pressure according to the standardized characteristics, the LSTM model and the XGBoost model; obtaining a total temperature drift prediction result according to the standardized characteristics and the GPR model; Obtaining a viscous temperature sensitive error according to the standardized characteristics, the viscous pseudo-differential pressure and the attention network model; and obtaining the net temperature drift error according to the viscous temperature sensitivity error and the total temperature drift prediction result. As a further improvement of the present invention, said deriving said viscous pseudo-differential pressure from said normalized features, said LSTM model and said XGBoost model comprises: obtaining a first error standard deviation, a second error standard deviation, a first predicted differential pressure and a second predicted differential pressure according to