Search

CN-121997704-A - LSTM-based pumped storage ultrasonic flowmeter temperature compensation model and method

CN121997704ACN 121997704 ACN121997704 ACN 121997704ACN-121997704-A

Abstract

Aiming at the defect that nonlinear and dynamic temperature influences are difficult to process under the complex working condition of a pumped storage power plant in the conventional temperature compensation method, the invention constructs and trains a neural network model comprising an input layer, an LSTM layer, a full-connection layer and an output layer after cleaning and standardized pretreatment by collecting temperature, ultrasonic propagation time and flow data under the actual working condition of laboratory simulation and field. The model can learn the long-term time sequence dependency relationship between the temperature and the flow error, output a high-precision error compensation value, and is deployed on a main control chip of the flowmeter to realize real-time dynamic compensation. The invention obviously improves the measurement precision and adaptability of the ultrasonic flowmeter under the condition of wide temperature range and provides reliable support for the accurate operation of the pumped storage power plant.

Inventors

  • PENG LI
  • WEI CHUAN
  • ZHAO YUAN

Assignees

  • 长电新能有限责任公司

Dates

Publication Date
20260508
Application Date
20251217

Claims (10)

  1. 1. An LSTM based pumped storage ultrasonic flowmeter temperature compensation model, comprising: The input layer is used for receiving the preprocessed temperature, ultrasonic wave propagation time and flow data and converting the preprocessed temperature, ultrasonic wave propagation time and flow data into tensor forms; an LSTM layer for learning a long-term dependence between temperature and flow measurement error; the full-connection layer is used for integrating the characteristics extracted by the LSTM layer; and the output layer is used for outputting the flow measurement error predicted value after temperature compensation.
  2. 2. The LSTM based pumped storage ultrasonic flow meter temperature compensation model of claim 1 wherein the data received by the input layer is temperature, ultrasonic travel time and flow data for the past 10 time steps and converted to tensor form.
  3. 3. The model of claim 1, wherein the LSTM layer comprises 2-3 LSTM cells, each layer contains 64-128 neurons, and long-term dependence between temperature and flow measurement errors is learned by retention and update of forgetting gate, input gate and output gate control information.
  4. 4. The LSTM-based pumped storage ultrasonic flowmeter temperature compensation model of claim 1, wherein the number of neurons in the output layer is 1, and a linear activation function is used to output a continuous temperature compensated flow measurement error prediction value.
  5. 5. The LSTM based pumped storage ultrasonic flow meter temperature compensation model of claim 1, wherein the model is trained using a mean square error loss function and Adam optimizer using data inputs of batch size 32-64.
  6. 6. A method of providing a LSTM based pumped storage ultrasonic flow meter temperature compensation model according to any one of claims 1 to 5, comprising the steps of: s1, acquiring temperature data, ultrasonic propagation time data and flow data of an ultrasonic flowmeter in a pumped storage power plant; S2, preprocessing the collected data, including data cleaning, outlier processing and data standardization; S3, constructing a temperature compensation model based on LSTM, wherein the temperature compensation model comprises an input layer, an LSTM layer, a full connection layer and an output layer; S4, training the temperature compensation model by using the preprocessed data, and updating parameters based on a loss function and an optimizer; s5, evaluating the performance of the model, and if the performance of the model does not reach a preset index, performing model optimization; and S6, deploying the optimized model into a main control chip of the ultrasonic flowmeter to realize real-time temperature compensation.
  7. 7. The method of LSTM based pumped storage ultrasonic flow meter temperature compensation model of claim 6, wherein data acquisition in step S1 comprises: simulating high-low temperature working conditions of a pumped storage power plant in a laboratory, and acquiring multi-station data at intervals of-50 ℃ to 150 ℃ and 5 ℃; Data are continuously collected in an actual power plant for at least 6 months, and the temperature, the ultrasonic wave propagation time, the actual flow and the unit state are recorded.
  8. 8. The method of LSTM based pumped storage ultrasonic flow meter temperature compensation model of claim 6, wherein the data preprocessing in step S2 comprises: Using Identifying and deleting abnormal values in principle, and filling missing data by adopting a linear interpolation or time sequence prediction method; Temperature, ultrasonic travel time, and flow data were normalized to the [0,1] interval: 。
  9. 9. The method of LSTM based pumped storage ultrasonic flow meter temperature compensation model of claim 6, wherein model training in step S4 comprises: training using Adam optimizer with mean square error as loss function, initial learning rate of 0.001: ; dividing the data into a training set and a testing set in proportion, inputting a model in batches, and updating parameters through forward propagation and backward propagation; training is stopped when the training loss is no longer significantly reduced for 10-15 consecutive rounds.
  10. 10. The method of LSTM based pumped storage ultrasonic flow meter temperature compensation model of claim 6, wherein model deployment in step S6 comprises: the main control chip collects temperature, ultrasonic wave propagation time and flow data in real time, and inputs the data into the model after pretreatment; The model outputs a flow measurement error compensation value, and the main control chip corrects the measurement result of the flowmeter according to the value.

Description

LSTM-based pumped storage ultrasonic flowmeter temperature compensation model and method Technical Field The invention belongs to the technical field of pumped storage, and particularly relates to a temperature compensation model and a temperature compensation method of a pumped storage ultrasonic flowmeter based on LSTM. Background In pumped storage power plants, ultrasonic flow meters play a role in draft, suction and water pipelines and the like, and are important as key flow measuring devices. But the propagation speed of ultrasonic waves in the fluid is greatly affected by temperature, which in turn leads to flow meter measurement errors. The existing temperature compensation method mostly adopts linear compensation or a simple empirical formula, and is difficult to accurately adapt to complex and changeable high-low temperature working conditions and nonlinear relations between temperature and measurement errors of the pumped storage power plant. In addition, the traditional method cannot effectively utilize dynamic information of temperature change along with time, so that the compensation effect is poor, the measurement reliability of the ultrasonic flowmeter in an extreme temperature environment is seriously influenced, and the accurate operation and management of the pumped storage power plant are restricted. Therefore, the patent provides a temperature compensation technology capable of effectively processing time series data and accurately capturing nonlinear relation between temperature and measurement error. Therefore, there is a need to design a pumped storage ultrasonic flowmeter temperature compensation model and method based on LSTM to solve the above-mentioned problems. Disclosure of Invention The invention aims to solve the technical problems of low compensation precision and poor adaptability of the traditional temperature compensation method in a high-low temperature environment of a pumped storage power plant by constructing a neural network model based on LSTM and combining a specific data processing and training strategy. In order to solve the technical problems, the invention adopts the following technical scheme: An LSTM based pumped storage ultrasonic flow meter temperature compensation model comprising: The input layer is used for receiving the preprocessed temperature, ultrasonic wave propagation time and flow data and converting the preprocessed temperature, ultrasonic wave propagation time and flow data into tensor forms; an LSTM layer for learning a long-term dependence between temperature and flow measurement error; the full-connection layer is used for integrating the characteristics extracted by the LSTM layer; and the output layer is used for outputting the flow measurement error predicted value after temperature compensation. Preferably, the data received by the input layer are temperature, ultrasonic propagation time and flow data of the past 10 time steps, and are converted into tensor form. Preferably, the LSTM layer comprises 2-3 layers of LSTM units, each layer comprises 64-128 neurons, and long-term dependence between temperature and flow measurement errors is learned through retention and updating of forgetting gate, input gate and output gate control information. Preferably, the number of neurons of the output layer is 1, and a linear activation function is adopted to output a continuous temperature compensated flow measurement error prediction value. Preferably, the model is trained by means of a mean square error loss function and Adam optimizer using data inputs of batch size 32-64. Preferably, a method of a LSTM based pumped storage ultrasonic flow meter temperature compensation model includes the steps of: s1, acquiring temperature data, ultrasonic propagation time data and flow data of an ultrasonic flowmeter in a pumped storage power plant; S2, preprocessing the collected data, including data cleaning, outlier processing and data standardization; S3, constructing a temperature compensation model based on LSTM, wherein the temperature compensation model comprises an input layer, an LSTM layer, a full connection layer and an output layer; S4, training the temperature compensation model by using the preprocessed data, and updating parameters based on a loss function and an optimizer; s5, evaluating the performance of the model, and if the performance of the model does not reach a preset index, performing model optimization; and S6, deploying the optimized model into a main control chip of the ultrasonic flowmeter to realize real-time temperature compensation. Preferably, the data acquisition in step S1 includes: simulating high-low temperature working conditions of a pumped storage power plant in a laboratory, and acquiring multi-station data at intervals of-50 ℃ to 150 ℃ and 5 ℃; Data are continuously collected in an actual power plant for at least 6 months, and the temperature, the ultrasonic wave propagation time, the actual flow and the unit state are re