Fluid flow prediction method of low-power-consumption flowmeter

Abstract

The invention relates to the field of intelligent flow measurement, and discloses a fluid flow prediction method for a low-power-consumption flowmeter, which is used for solving the problems of fluctuation and error of flow data caused by various factors such as change of fluid flow state and influence of impurities in a pipe section when the flowmeter is used. A flow prediction algorithm is used for enabling the flow meter to predict and correct the fluid flow within the interval time of intermittent power supply excitation metering of the battery, so that the use scheduling of the battery of the flow meter is reduced, and the service life of the flow meter is prolonged; a long-short-term memory machine learning algorithm is used for performing fitting calculation on historical flow data, early warning is performed on abnormal flow possibly generated in future time in combination with the historical data, the accuracy of the flow meter is effectively improved, and waste of resources is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent flow metering, and in particular to a fluid flow prediction method of a low-power consumption flow meter. Background Art

[0002] A flow meter is an instrument used to measure the amount of fluid (such as liquid or gas) flowing. It can accurately monitor the flow rate or flow rate of the fluid and convert this data into easy-to-read numerical values. This data is critical for applications such as industrial process control, resource management and safety monitoring. There are many types of flow meters, including mechanical, electronic and ultrasonic, each type has its own unique working principle and scope of application.

[0003] When measuring fluid flow, traditional flow meters usually rely on sensors and converters to read data directly. Flow meters powered by mains electricity use power frequency signal excitation, which has large interference noise. The installation environment of the flow meter is relatively remote, and it is difficult to supply mains electricity. It is powered by batteries, but due to the problem of battery capacity, in order to ensure the service life of the flow meter, continuous excitation is impossible. Therefore, intermittent excitation can only be used for excitation measurement. In actual applications, due to various factors such as changes in fluid flow state and the influence of impurities in the pipe section, flow data fluctuates and errors exist. Therefore, how to increase the service life of the flow meter on the basis of ensuring stability has become an urgent problem to be solved in this field. In addition, as a measuring instrument for the pipeline network, the flow meter is characterized by a particularly large caliber flow rate. Once a problem occurs, it will cause disasters and waste of resources in the surrounding area. Therefore, it is very important to use a prediction algorithm to warn of abnormal flow that may occur in the future to reduce harm and waste of resources. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a fluid flow prediction method of a low-power flow meter to solve the problems existing in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for predicting fluid flow rate of a low-power flow meter includes obtaining a pre-measured fluid flow rate corresponding to the flow meter;

[0007] Continuously calibrate the fluid flow of the flow meter and establish a historical database;

[0008] Calculate characteristic values ​​according to data in the historical database, perform autocorrelation analysis on historical fluid flow data according to the autocorrelation function to obtain an autocorrelation coefficient, and obtain a partial autocorrelation function according to the autocorrelation coefficient;

[0009] Constructing an ARIMA autoregressive integrated moving average model according to the partial autocorrelation function, and training the ARIMA autoregressive integrated moving average model;

[0010] Predicting the fluid flow of the flow meter according to the trained ARIMA autoregressive integrated moving average model;

[0011] Introducing a long short-term memory network into the ARIMA autoregressive integrated moving average model for prediction;

[0012] Time series traffic data and factors affecting traffic mutation are collected as original data sets for the training model of the long short-term memory network.

[0013] Preferably, the method of obtaining the fluid flow rate corresponding to the pre-measured flow meter is specifically:

[0014] Measuring the inner diameter of the circular pipe of the flow meter and measuring the average flow rate per unit time, and calculating the fluid flow rate passing through the circular pipe according to the inner diameter of the circular pipe and the average flow rate;

[0015] The formula for calculating fluid flow is:

[0016] The formula for calculating the induced electromotive force is:

[0017] Combining (1) and (2), the relationship between the fluid flow rate and the induced electromotive force is:

[0018] Among them, Q represents the fluid flow rate, B represents the magnetic induction intensity, which is obtained by the magnetic field measuring instrument, and D represents the inner diameter of the circular pipe. Expressed as average flow velocity.

[0019] Preferably, the characteristic value is calculated based on the data in the historical database as follows:

[0020] The average value of historical traffic data is calculated based on the data in the historical database. The calculation formula is: Where S n Expressed as the average value of historical traffic data, S i represents the i-th historical traffic data, and n is the number of historical traffic data;

[0021] The standard deviation of historical traffic data is calculated based on the data in the historical database. The calculation formula is: Where σ represents the standard deviation, S i represents the i-th historical traffic data, S n Expressed as an average of historical traffic data.

[0022] Preferably, the autocorrelation coefficient obtained by performing autocorrelation analysis on the historical fluid flow data according to the autocorrelation function is specifically:

[0023] The autocorrelation function ACF is used to measure the correlation of time series data at different lag times. The time series of historical traffic data is y t , where t is the time point, then the ACF calculation formula at lag K is where ρ k Expressed as the autocorrelation coefficient, Cov(y t ,y t-k ) is represented by y t and t-k The covariance of Var(y t ) and Var(y t-k ) are y t and t-k The variance of .

[0024] Preferably, the partial autocorrelation function obtained according to the autocorrelation coefficient is:

[0025] By linear regression method, the time series X t To build a model, for a given lag order k, linearly regress the k-order lagged data of the time series itself with the first k-1 lagged data of the time series;

[0026] The residual ε obtained from the linear regression model t Represents the time series X after considering the effect of lag k-1 t In the independent part at lag k;

[0027] Calculate the time series X t The autocorrelation coefficient ρ at lag k k ;

[0028] Compute the autocorrelation coefficient at lag k, controlling for lags 1 to k-1.

[0029] Preferably, the fluid flow of the flow meter is predicted according to the trained ARIMA autoregressive integrated moving average model as follows:

[0030] The AP model is constructed using the delayed values ​​of the previous time series, which represents the linear combination of the current value and the previous p times. Its expression is AR(p):y t =c+φ1y t-1 +φ2y t-2 +...+φ p y t-p +ε t , where yt is the current value of the time series, φ1, φ2…φ pis the parameter of the model, c is a constant term, ε t is the error term;

[0031] Construct the difference part, which is expressed as Difference (d): Δ d y t =(1-L) d y t ;

[0032] Construct a moving average model, whose expression is MA(q): c+θ1ε t-1 +θ2ε t-2 +...+θ q ε t-q +ε t , where θ is the parameter of the model, ε t is the random error term at the current moment, and c is the constant term;

[0033] Through modeling training for continuous time series historical data, the calibrated historical data is used to determine the q, d, and p parameters, and to check whether the time series is stable;

[0034] By operating on historical flow data, the flow data of the excitation interval is predicted in the short term by relying on the autocorrelation with the time series flow changes.

[0035] Preferably, the introduction of a long short-term memory network into the ARIMA autoregressive integrated moving average model for prediction is specifically as follows:

[0036] The processed residual sequence is divided into training set and test set, and the input sequence and target sequence are created using sliding window technology, and then organized into an input form suitable for the LSTM network model;

[0037] Designing an LSTM network model structure, wherein the LSTM network model structure includes an LSTM layer and a fully connected layer;

[0038] Use the training data to train the LSTM network model, select the loss function and optimization algorithm to optimize the model parameters, and adjust the model's hyperparameters based on the performance of the validation set;

[0039] The prediction results of the ARIMA autoregressive integrated moving average model and the LSTM network model are combined, the prediction results of the ARIMA autoregressive integrated moving average model are added to the residual prediction results of the LSTM network model to obtain the final prediction value, the trained LSTM network model is used to predict the future residuals, and the future values ​​are generated in combination with the prediction results of the ARIMA autoregressive integrated moving average model.

[0040] Compared with the prior art, the fluid flow prediction method of a low-power flow meter provided by the present invention has at least the following beneficial effects:

[0041] The ARIMA autoregressive integrated moving average model is used to allow the flow meter to predict and correct the fluid flow during the interval of intermittent battery power excitation metering, thereby reducing the use scheduling of the flow meter battery and increasing the service life of the flow meter. At the same time, a long short-term memory network machine learning algorithm is used to fit and extrapolate historical flow data, and combined with historical flow data, early warning of abnormal flow that may occur in the future is issued, which effectively improves the accuracy of the flow meter and reduces resource waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without inventiveness.

[0043] Figure 1 A flow chart of time series modeling provided by the present invention;

[0044] Figure 2 Schematic diagram of the LSTM network model provided by the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are merely illustrative. The fluid flow prediction method of a low-power flow meter involved in the present invention is not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0046] The present invention provides a method for predicting fluid flow rate of a low-power flow meter, comprising the following steps:

[0047] First, obtain the fluid flow rate corresponding to the pre-measured flowmeter, and use the flowmeter to obtain the induced electromotive force generated when the fluid flows through a uniform magnetic field based on Faraday's law of electromagnetic induction, and calculate the fluid flow rate based on the induced electromotive force, and then calculate the fluid flow rate. During actual measurement, the measurement principle of the flowmeter is based on Faraday's law of electromagnetic induction. When the fluid flows through a uniform magnetic field, an induced electromotive force will be generated. In the conductive fluid, the fluid, electrode and wire form a loop.

[0048] In this embodiment, it should be specifically explained that the specific steps of obtaining the fluid flow corresponding to the pre-measured flow meter are:

[0049] Measure the inner diameter of the circular pipe, measure the magnetic average flow velocity per unit time, and calculate the volume flow of the fluid passing through the measured pipe section based on the measured inner diameter and the average flow velocity. The calculation formula is: Where Q represents the fluid volume flow rate, D represents the measuring inner diameter, Expressed as average flow velocity;

[0050] According to the induced electromotive force calculation formula Where B represents the magnetic induction intensity, which is obtained by a magnetic field measuring instrument, and D represents the measured inner diameter. Expressed as the average flow rate, substituting the fluid volume flow rate Q calculation formula into the fluid flow rate and the induced electromotive force is a linear relationship, and its expression is: Where B represents the magnetic induction intensity, D represents the measured inner diameter. After the structural design and excitation current are determined, is a constant K.

[0051] Secondly, the fluid flow of the flow meter is continuously calibrated to establish a historical database. The intermittent excitation sampling method is used to set the excitation cycle of the electromagnetic flow meter so that the electromagnet works within a predetermined time interval and is in a dormant state for the rest of the time, thereby reducing power consumption. The excitation cycle T on With sleep cycle T off The selection is optimized according to the battery capacity and actual application requirements;

[0052] The intermittent excitation sampling method is a technology that reduces the power consumption of electromagnetic flowmeters by periodically turning on and off the excitation current. The electromagnetic flowmeter is powered on during the set excitation cycle to generate a magnetic field and collect flow data, and is powered off during the dormant cycle to reduce power consumption. This intermittent excitation method can effectively extend the battery life while maintaining the accuracy of flow measurement.

[0053] By periodically turning the excitation current on and off, the power consumption of the device during non-measurement periods is reduced. The electromagnetic flowmeter does not consume energy during the sleep cycle, which significantly reduces the overall energy consumption. For battery-powered devices, reducing power consumption directly extends the battery life and reduces the frequency of battery replacement and maintenance costs. By precisely controlling the excitation and sleep cycles, the energy efficiency of the device can be adjusted according to actual needs to ensure sufficient performance when measurement is required, while reducing energy consumption when measurement is not required. Intermittent excitation sampling can maintain sufficient measurement accuracy through appropriate excitation cycles, especially when the fluid flow rate does not change drastically. Regularly turning on the excitation ensures the accuracy and consistency of the data. During the sleep cycle, the prediction algorithm can be used to predict and compensate for the flow to ensure that continuous flow data is not affected by intermittent sampling.

[0054] In actual operation, the flow data is collected during the excitation cycle, and the timestamp and the corresponding flow value are recorded. According to the timestamp and the corresponding flow value and the boundary value of each flow point, and considering the influence of various environmental factors on the flow, the flow meter is continuously calibrated under the influence of various parameters such as pressure and water temperature. The calibration formula is: Where Q' represents the flow rate, V is the induced electromotive force, k is a constant, B is the magnetic induction intensity, and D is the pipe diameter. The flow data at each flow point is collected to establish a historical database.

[0055] The third step is to calculate the characteristic value based on the data in the historical database, perform autocorrelation analysis on the historical fluid flow data based on the autocorrelation function to obtain the autocorrelation coefficient, and obtain the partial autocorrelation function based on the autocorrelation coefficient; autocorrelation analysis is a method to study the correlation between a time series data and itself at different time points. It measures the degree of correlation between the same event in two different periods.

[0056] Eigenvalues ​​are the main statistics used to describe a data set, usually including mean, median, mode, variance, and standard deviation. The mean represents the center of the data; the median is the middle position after the data is sorted; the mode is the most frequent value in the data; the variance and standard deviation measure the degree of dispersion or volatility of the data. These statistical characteristics help understand the overall distribution and variability of the data and are the basis of data analysis and decision-making.

[0057] The autocorrelation function is a statistical tool used to measure the autocorrelation of time series data at different time lags. It calculates the degree of correlation between the data values ​​at one time point in the series and the data values ​​at other time points, thereby revealing the periodicity and trend of the data. The autocorrelation function helps identify patterns and dependencies in the data by comparing the correlation between a data point and its lagged data points. The autocorrelation coefficient ACF value ranges from -1 to 1, with values ​​close to 1 or -1 indicating a strong positive or negative correlation, and values ​​close to 0 indicating a weak correlation.

[0058] The time series traffic data and the factors that affect traffic mutation are collected as the original data set for the training model of the long short-term memory network. Autocorrelation function is a function of the degree of correlation between random signals or time series data at different times. It is equivalent to "cross-correlating" the signal or data sequence itself to measure the similarity or correlation of the same sequence at different time points.

[0059] The partial autocorrelation function is a statistical tool used to measure the autocorrelation of time series data at a given lag while controlling for the effects of shorter lags. Unlike the autocorrelation function, the partial autocorrelation function focuses on the direct correlation between the data at a specific lag and the series itself after excluding the effects of the intermediate lags. This helps identify the true independent relationship in the time series and is often used to determine the order of the autoregressive model, thereby providing a more accurate basis for model construction in time series analysis.

[0060] Fourth, an ARIMA autoregressive integrated moving average model is constructed based on the partial autocorrelation function, and the ARIMA autoregressive integrated moving average model is trained; the fluid flow of the flow meter is predicted based on the trained ARIMA autoregressive integrated moving average model;

[0061] The ARIMA autoregressive integrated moving average model is a statistical model widely used for time series forecasting. It combines the autoregressive (AR) part, the difference (I) part, and the moving average (MA) part. The AR part captures the linear relationship in the time series, the MA part handles the random noise in the data, and the I part eliminates the trend and seasonality in the data through the difference operation to achieve stationarity. The ARIMA model fits historical data by adjusting the parameters of these components and predicts future values ​​based on these fitting results. It is suitable for processing non-stationary time series data.

[0062] In this embodiment, it should be specifically explained that the fluid flow of the flow meter is predicted according to the trained ARIMA autoregressive integrated moving average model as follows:

[0063] The AP(p) model is constructed using the delayed values ​​of the previous time series, which represents the linear combination of the current value and the previous p times. Its expression is AR(p):y t =c+φ1y t-1 +φ2y t-2 +...+φ p y t-p +ε t , where yt is the current value of the time series, φ1, φ2…φ p is the parameter of the model, c is a constant term, ε t is the error term;

[0064] Construct the difference (I) part. Difference is to perform subtraction on the original time series to make it a stationary time series. The characteristic of a stationary time series is that its statistical characteristics remain unchanged over time. Its expression is difference (d): Δ d y t =(1-L) d y t ;

[0065] Construct a moving average (MA) model. The moving average model represents the linear combination of the current observation value and the random error term and the past random error term. The MA(q) model is used to protect q lagged error terms. Its expression is MA(q): c+θ1ε t-1 +θ2ε t-2 +...+θ q ε t-q +ε t , where θ is the parameter of the model, ε t is the random error term at the current moment, and c is the constant term;

[0066] Through modeling training for continuous time series historical data, the q, d, and p parameters are determined using calibrated historical data, and whether the time series is stable is checked; through a series of operations on historical flow data, the flow data of the excitation interval can be predicted in a short period of time by relying on the autocorrelation with the time series flow changes.

[0067] Finally, the long short-term memory network (LSTM) is introduced into the ARIMA autoregressive integrated moving average model for prediction, so that the model can autonomously select the information to be retained and forgotten, so as to better capture the long-term dependencies in the sequence.

[0068] Long short-term memory network (LSTM) is a special type of recurrent neural network. LSTM network controls the preservation and forgetting of information by introducing gating mechanisms (including input gate, forget gate and output gate), so that the network can selectively retain important long-term information and dynamically forget information that is no longer relevant. This mechanism makes LSTM particularly suitable for processing and predicting sequence data with complex time dependencies, such as time series prediction, natural language processing and other application scenarios.

[0069] Most of the flow data of the flow meter during operation is stable and normal. This part of the data is no longer of reference value after being saved as cumulative flow, but the flow data in the period before the large flow, overload flow, abnormal flow mutation and these flows are of great reference value. Using LSTM can effectively transmit and express information in long time series without causing useful information from a long time ago to be ignored. Only part of the information needs to be remembered for a long time. At the same time, we also need a mechanism to dynamically process the "memory" of the neural network, because some information may be very valuable at the beginning, and then the value gradually decreases. At this time, we also need to let the neural network learn to "forget" specific information. With the help of its recording and forgetting characteristics, we input the fusion features into the LSTM module with powerful time series prediction capabilities.

[0070] In this embodiment, it should be specifically explained that the specific steps of introducing the long short-term memory network into the ARIMA autoregressive integrated moving average model for prediction are as follows:

[0071] Arrange the processed data (such as residual sequences and other features) into a form suitable for LSTM input. Usually, the data needs to be split into training sets and test sets, and the sliding window technique is used to create input sequences and target sequences;

[0072] Design the LSTM network structure, such as Figure 2 As shown, it includes LSTM layer and fully connected layer. LSTM layer is used to capture long-term dependencies in data, and the output layer generates prediction results;

[0073] Use the training data to train the LSTM model, select appropriate loss functions and optimization algorithms to optimize model parameters, and adjust the model's hyperparameters, such as the number of units in the LSTM layer, learning rate, batch size, etc., based on the performance of the validation set;

[0074] Combine the prediction results of the ARIMA model and the LSTM model, add the prediction results of the ARIMA model to the residual prediction results of the LSTM model to get the final prediction value, use the trained LSTM model to predict the future residuals, and generate future values ​​in combination with the prediction results of the ARIMA model.

[0075] In actual use, the time series traffic data and the mutation factors that affect the traffic are collected as the original data set for the training model of the long short-term memory network.

[0076] The training implementation code of the LSTM module is as follows:

[0077]

[0078]

[0079] The LSTM model is trained according to the above training method. In the long-term operation environment of the flow meter, the long-term memory of abnormal flow and the dependency relationship between parameters of various time series flows are captured to predict abnormal flow events in the future.

[0080] In this embodiment, it should be specifically explained that the characteristic value is calculated based on the data in the historical database as follows:

[0081] The average value of historical traffic data is calculated based on the data in the historical database to reflect the typical level of traffic. The calculation formula is: Where S n Expressed as the average value of historical traffic data, S i represents the i-th historical traffic data, and n is the number of historical traffic data;

[0082] The standard deviation of historical traffic data is calculated based on the data in the historical database to measure the volatility or variation of traffic data. The calculation formula is: Where σ represents the standard deviation, S i represents the i-th historical traffic data, S n Expressed as an average of historical traffic data.

[0083] In this embodiment, it should be specifically explained that the steps of using the autocorrelation function to perform autocorrelation analysis on historical traffic data to obtain the autocorrelation coefficient are as follows:

[0084] The autocorrelation function ACF is used to measure the correlation of time series data at different lag times. The time series of historical traffic data is y t , where t is the time point, then the ACF calculation formula at lag K is where ρ k Expressed as the autocorrelation coefficient, Cov(y t ,y t-k ) is represented by y t and t-k The covariance of Var(y t ) and Var(y t-k ) are y t and t-k The variance of .

[0085] In this embodiment, it should be specifically explained that the steps of obtaining the partial autocorrelation function PACF according to the autocorrelation coefficient are:

[0086] By linear regression method, the time series X t For a given lag order k, linear regression is performed on the k-order lagged data of the time series itself and the first k-1 lagged data of the time series;

[0087] The residual ε obtained from the linear regression model t Represents the time series X after considering the effect of lag k-1 t In the independent part at lag k;

[0088] Calculate the time series X t The autocorrelation coefficient ρ at lag k k ;

[0089] Calculate the autocorrelation coefficient of lag k under the condition of controlling lag 1 to k-1. Specifically, use the regression coefficient in the linear regression model to estimate, and its expression is where φ kk Expressed as the partial autocorrelation function of lag k, ρ kExpressed as the autocorrelation coefficient, φ jk is the partial autocorrelation coefficient of lag j, ρ k-j is the autocorrelation coefficient of lag kj.

[0090] The fluid flow prediction method of a low-power flow meter provided in this application, in actual use, in the calculation and training of the two-layer model, by calibrating the original flow data and collecting the operation data, the flow data within the interval excitation time can be predicted through the training method of autocorrelation regression, so as to achieve the purpose of reducing the use of battery scheduling, thereby increasing the service life of the flow meter. It can also warn of possible flow anomalies in the future during long-term operation, thereby reducing the possibility of disasters and waste of resources.

[0091] Finally: The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0092] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting fluid flow rate of a low power consumption flow meter, characterized in that: include: Obtaining the fluid flow rate corresponding to the pre-measured flow meter; Continuously calibrate the fluid flow of the flow meter and establish a historical database; Calculate characteristic values ​​according to data in the historical database, perform autocorrelation analysis on historical fluid flow data according to the autocorrelation function to obtain an autocorrelation coefficient, and obtain a partial autocorrelation function according to the autocorrelation coefficient; Constructing an ARIMA autoregressive integrated moving average model according to the partial autocorrelation function, and training the ARIMA autoregressive integrated moving average model; Predicting the fluid flow of the flow meter according to the trained ARIMA autoregressive integrated moving average model; Introducing a long short-term memory network into the ARIMA autoregressive integrated moving average model for prediction; Time series traffic data and factors affecting traffic mutation are collected as original data sets for the training model of the long short-term memory network.

2. The method for predicting fluid flow rate of a low power consumption flow meter according to claim 1, characterized in that: The method of obtaining the fluid flow rate corresponding to the pre-measured flow meter is specifically: Measuring the inner diameter of the circular pipe of the flow meter and measuring the average flow rate per unit time, and calculating the fluid flow rate passing through the circular pipe according to the inner diameter of the circular pipe and the average flow rate; The formula for calculating fluid flow is: The formula for calculating the induced electromotive force is: Combining (1) and (2), the relationship between the fluid flow rate and the induced electromotive force is: Among them, Q represents the fluid flow rate, B represents the magnetic induction intensity, which is obtained by the magnetic field measuring instrument, D represents the inner diameter of the circular pipe, and v represents the average flow velocity.

3. The method for predicting fluid flow rate of a low power consumption flow meter according to claim 1, characterized in that: The specific calculation of the characteristic value based on the data in the historical database is as follows: The average value of historical traffic data is calculated based on the data in the historical database. The calculation formula is: Where S n Expressed as the average value of historical traffic data, S i represents the i-th historical traffic data, and n is the number of historical traffic data; The standard deviation of historical traffic data is calculated based on the data in the historical database. The calculation formula is: Where σ represents the standard deviation, S i represents the i-th historical traffic data, S n Expressed as an average of historical traffic data.

4. The method for predicting fluid flow rate of a low power consumption flow meter according to claim 1, characterized in that: The autocorrelation coefficient obtained by performing autocorrelation analysis on the historical fluid flow data according to the autocorrelation function is specifically: The autocorrelation function ACF is used to measure the correlation of time series data at different lag times. The time series of historical traffic data is y t , where t is the time point, then the ACF calculation formula at lag K is where ρ k Expressed as the autocorrelation coefficient, Cov(y t ,y t-k ) is represented by y t and t-k The covariance of Var(y t ) and Var(y t-k ) are y t and t-k The variance of .

5. The method for predicting fluid flow rate of a low power consumption flow meter according to claim 4, characterized in that: The partial autocorrelation function obtained according to the autocorrelation coefficient is: By linear regression method, the time series X t To build a model, for a given lag order k, linearly regress the k-order lagged data of the time series itself with the first k-1 lagged data of the time series; The residual ε obtained from the linear regression model t Represents the time series X after considering the effect of lag k-1 t In the independent part at lag k; Calculate the time series X t The autocorrelation coefficient ρ at lag k k ; Compute the autocorrelation coefficient at lag k, controlling for lags 1 to k-1.

6. The method for predicting fluid flow rate of a low power consumption flow meter according to claim 1, characterized in that: The fluid flow of the flow meter is predicted according to the trained ARIMA autoregressive integrated moving average model as follows: The AP model is constructed using the delayed values ​​of the previous time series, which represents the linear combination of the current value and the previous p times. Its expression is AR(p):y t =c+φ1y t-1 +φ2y t-2 +...+φ p y t-p +ε t , where yt is the current value of the time series, φ1, φ2…φ p is the parameter of the model, c is a constant term, ε t is the error term; Construct the difference part, which is expressed as Difference (d): Δ d y t =(1-L) d y t ; Construct a moving average model, whose expression is MA(q): c+θ1ε t-1 +θ2ε t-2 +…+θ q ε t-q +ε t , where θ is the parameter of the model, ε t is the random error term at the current moment, and c is the constant term; Through modeling training for continuous time series historical data, the calibrated historical data is used to determine the q, d, and p parameters, and to check whether the time series is stable; By operating on historical flow data, the flow data of the excitation interval is predicted in the short term by relying on the autocorrelation with the time series flow changes.

7. The method for predicting fluid flow rate of a low power consumption flow meter according to claim 1, characterized in that: The introduction of long short-term memory network into the ARIMA autoregressive integrated moving average model for prediction is specifically as follows: The processed residual sequence is divided into training set and test set, and the input sequence and target sequence are created using sliding window technology, and then organized into an input form suitable for the LSTM network model; Designing an LSTM network model structure, wherein the LSTM network model structure includes an LSTM layer and a fully connected layer; Use the training data to train the LSTM network model, select the loss function and optimization algorithm to optimize the model parameters, and adjust the model's hyperparameters based on the performance of the validation set; The prediction results of the ARIMA autoregressive integrated moving average model and the LSTM network model are combined, the prediction results of the ARIMA autoregressive integrated moving average model are added to the residual prediction results of the LSTM network model to obtain the final prediction value, the trained LSTM network model is used to predict the future residuals, and the future values ​​are generated in combination with the prediction results of the ARIMA autoregressive integrated moving average model.

Images