Water meter detection method and system with multi-pulse flowmeter
Through multi-pulse flowmeter and quantum noise injection technology, combined with the Ising model and generative adversarial network, the data accuracy and slow response problems of traditional flow monitoring systems are solved, and efficient flow anomaly detection and flexible flow management are achieved.
Patent Information
- Application Number
- CN202510800746.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-16
- Publication Date
- 2025-09-16
AI Technical Summary
Traditional traffic monitoring systems are susceptible to external interference, resulting in reduced data accuracy, frequent false alarms and missed alarms in anomaly detection methods, a lack of flexibility in communication protocols, and reliance on historical data leading to unreliable decision-making and slow response.
A multi-pulse flowmeter is used in combination with quantum noise injection, feature extraction and tensor construction, Ising model optimization and generative adversarial network, and the communication protocol is adjusted in real time in combination with the water-power coupling equation to achieve flow anomaly detection and management.
It improves the data accuracy and response speed of traffic monitoring, reduces the false alarm rate, enhances the system's adaptability and decision-making reliability, and optimizes the flexibility and security of traffic management.
Smart Images
Figure CN120651324A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent water meter measurement, and in particular to a water meter detection method and system with a multi-pulse flowmeter. Background Art
[0002] In the current flow monitoring field, traditional technologies rely primarily on single sensors for real-time flow monitoring. This approach has significant limitations. First, single sensors are susceptible to interference from external factors, resulting in reduced data accuracy. For example, changes in the surrounding environment, such as increased temperature or pressure, can distort sensor readings.
[0003] At the same time, many existing technologies use anomaly detection algorithms based on fixed thresholds. This approach often results in false positives or missed negatives in practical applications. The system often struggles to adapt to complex traffic fluctuations, resulting in an inability to promptly identify true anomalies. Furthermore, this approach is unable to learn and adapt to emerging anomaly patterns.
[0004] Furthermore, traffic management communication protocols often lack flexibility. Most systems are unable to quickly adjust management strategies when emergencies occur. This means that once an anomaly occurs, response times are long, potentially leading to further losses.
[0005] Finally, existing technologies often rely on historical data for predictions and decision-making. This can become highly unreliable when data is insufficient, leading to poor decision-making and, consequently, the effectiveness of overall traffic management. Such systems often perform poorly in the face of uncertainty and novel situations. Summary of the Invention
[0006] The purpose of the present invention is to provide a water meter detection method and system with a multi-pulse flowmeter, which solves the problems of low abnormality detection and management efficiency caused by inaccurate data and slow response in traditional flow monitoring.
[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A water meter detection method and system with a multi-pulse flowmeter, comprising the following steps: S1: Signal acquisition and preprocessing: Flow data is collected through three pulse flow meters with different ranges, and the flow data is preprocessed to remove electromagnetic interference and high-frequency noise to obtain high-quality multi-source signals; S2: Quantum noise injection: dynamically generating quantum noise that matches the spectrum of the high-quality multi-source signal and injecting the quantum noise into the signal; S3: Feature extraction and tensor construction: Extract features from the signal after quantum noise injection, including extracting time domain features and frequency domain features, and constructing a four-dimensional feature tensor containing these features; S4: Feature Optimization: Mapping the four-dimensional feature tensor into an Ising model and optimizing the model features using a quantum algorithm; S5: Anomaly Detection: Based on the optimized model features, a generative adversarial network is trained using an adversarial learning framework to detect and identify unknown abnormal signals and output corresponding abnormal warning information to obtain detection results; S6: Verification and Decision-making: Physically verify the detection results by using the water-electricity coupling equation, and adjust the communication protocol for flow management in real time based on the detection results.
[0008] Preferably, the pre-processing step after signal acquisition includes using a low-pass filter, the cut-off frequency of which is set to no more than 10 kHz, and a band-stop filter, the center frequency of which is set to 50 Hz.
[0009] Preferably, in the process of dynamically generating quantum noise, the calculation formula of the adaptive noise coefficient α used is: In the formula, α represents an adjustable parameter, SNR current Indicates the current signal-to-noise ratio. 30dB represents the normal signal-to-noise ratio baseline.
[0010] Preferably, the feature tensor is constructed as follows: Where T is the time step, C is the number of sensor channels, F is the feature dimension, and E is the environmental parameter.
[0011] Preferably, in step S4, the implementation of the Ising model used in the feature optimization step calculates feature weights through a quantum algorithm.
[0012] Preferably, the training of the generative adversarial network in the anomaly detection step includes identifying and classifying abnormal signals to generate corresponding early warning information.
[0013] Preferably, in step S6, the water-electricity coupling equation is used to verify the flow measurement result, and the equation is: In the formula, represents the rate of change over time t, Express Perform divergence calculation, S(x,t) represents the source term or source term distribution, λ represents a constant coefficient, σ represents a function, Represents electric and magnetic fields.
[0014] Preferably, in step S6, the dynamic protocol switching used in the traffic management optimizes the communication protocol selection according to the detection results to reduce energy consumption and response delay.
[0015] Preferably, in step S3, the feature extraction step further includes extracting frequency domain features using a short-time Fourier transform method.
[0016] A water meter detection system with a multi-pulse flowmeter, comprising: A plurality of pulse flow meters of different ranges are configured to collect flow data and connected to the preprocessing module; The pre-processing module is used to remove electromagnetic interference and noise from the collected signals and output highly clear multi-source signals to the quantum noise generation and injection module; Quantum noise generation and injection module, dynamically generates and injects quantum noise, and outputs the injected signal to the feature extraction and tensor construction module; A feature extraction and tensor construction module is used to extract features and construct feature tensors based on the feature extraction step, and output the feature tensors to the quantum optimization module for optimization; The quantum optimization module uses the Ising model to optimize the extracted features and outputs the optimized features to the anomaly detection module. The anomaly detection module uses a generative adversarial network to detect and identify unknown anomalies according to the anomaly detection steps and outputs the detection results to the decision output module. Physical verification module, which verifies the monitoring results through the hydropower coupling equation and associates them with the decision module; The decision output module adjusts the communication protocol in real time based on the monitoring results to optimize traffic management.
[0017] In summary, the present invention includes at least one of the following beneficial technical effects: 1. This invention uses a technical solution that combines water-power coupling equations with real-time data monitoring to achieve efficient verification of abnormal flow signals. This method is more comprehensive than the single-sensor monitoring method in the existing technology and solves the problem of detection delay caused by equipment failure.
[0018] 2. This invention uses generative adversarial networks to model traffic data, achieving more accurate anomaly detection. Compared with traditional threshold-based detection methods, it can effectively reduce false alarms, improve the system's signal recognition rate, and ensure the security of traffic management.
[0019] 3. The present invention realizes the real-time adjustment of the traffic management communication protocol, improves the adaptability of the system, and compared with the existing technology, the solution can quickly respond to emergencies, solves the problem of inflexible traffic regulation, and enhances the flexibility of the system.
[0020] 4. This invention combines physical models with data-driven methods to achieve better traffic management results. Compared with technical solutions that rely solely on historical data, it fully utilizes physical laws, solves the problem of model failure when data is insufficient, and improves the reliability of overall decision-making. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 Schematic diagram of the method flow of the present invention; Figure 2 Schematic diagram of the system module of the present invention. DETAILED DESCRIPTION
[0022] The following is combined with Figure 1 , the present invention is described in further detail.
[0023] The present invention provides a water meter detection method with a multi-pulse flowmeter, comprising the following steps: S1: Signal acquisition and preprocessing: Flow data is collected through three pulse flow meters with different ranges, and the flow data is preprocessed to remove electromagnetic interference and high-frequency noise to obtain high-quality multi-source signals; Specifically, the system first collects signals by setting up multiple pulse flowmeters, including low-range (0-500 L / h), mid-range (500-2000 L / h), and high-range (2000-5000 L / h) flowmeters. Each flowmeter works based on electromagnetic induction, measuring flow rate through an internal sensor and outputting a pulse signal proportional to the fluid flow rate.
[0024] After signal acquisition, a signal preprocessing model is first established. This model includes steps such as denoising, filtering, gain adjustment, and signal enhancement. Furthermore, a low-pass filter is first used with a cutoff frequency set to no greater than 10kHz to ensure that high-frequency noise above this frequency is removed. The filter's frequency response function, H(f), can be expressed as: Where H(f) represents the frequency response of the filter, f represents the signal frequency, and f f represents the cutoff frequency of the filter, and n represents the order of the filter.
[0025] Then, a band-stop filter is used with its center frequency set to 50Hz to eliminate interference caused by the power frequency. The frequency response of the band-stop filter can be expressed as: In the formula, H band-stop (f) represents the transfer function of the band-stop filter, f represents the frequency of the input signal, ω0 represents the cutoff frequency of the filter, Q represents the quality factor, j represents the imaginary unit, Indicates the response characteristics near the frequency f, It represents the combination of frequency f, imaginary part and cutoff frequency.
[0026] After filtering, the signals output by each flow meter are combined into a multi-source signal to form a composite signal S, which can be expressed as: S(t)=S1(t)+S2(t)+S3(t); Where S(t) represents the synthetic signal, S1(t), S2(t) and S3(t) are the output signals of the low, medium and high range flowmeters respectively.
[0027] After signal synthesis, the synthesized signal S(t) is further denoised to improve signal quality. The power spectrum density P(f) of the signal is calculated through frequency domain analysis, and the signal-to-noise ratio (SNR) is defined to evaluate the signal quality: In the formula, P signal Represents the power spectral density of the signal, P noise Represents the power spectral density of the noise.
[0028] The power spectral density P(f) can be obtained by Fourier transform of the signal: P(f)=|S(f)| 2 ; Where S(f) represents the discrete Fourier transform (DFT) of the signal, which is expressed as: In the formula, S(f) represents the frequency domain signal or spectrum, represents the sum of sample points n from 0 to N-1, S(n) represents the amplitude or value of the discrete time signal at the nth sample point, e -j2πfn / N represents the frequency component of the signal, j is an imaginary unit representing the phase, 2π is a constant used to convert frequency to radians, f is the frequency variable obtained after Fourier transform, n is the sample index of the discrete time series, and N is the total number of samples of the signal, which is used to normalize the frequency range.
[0029] When calculating S(n), the discretization process of the flowmeter output signal should follow the sampling theorem to ensure that the sampling frequency f s Greater than twice the maximum signal frequency.
[0030] After the signal denoising process is completed, the gain of the processed signal is adjusted by the set gain adjustment algorithm to ensure the signal uniformity.
[0031] The adjustment formula example can be expressed as follows: Sadjusted (t) = G·S(t); In the formula, S adjusted (t) represents the gain-adjusted signal, and G represents the gain factor.
[0032] Finally, the preprocessed signal will be used in the subsequent feature extraction and analysis stages. The resulting high-quality multi-source signal will lay an effective foundation for the subsequent injection of quantum noise and extraction of flow characteristics. It is important to emphasize that by establishing a detailed signal acquisition and preprocessing process, the technical solution effectively connects the signal acquisition, signal denoising, and signal enhancement operations of multiple flow meters, ensuring data accuracy and reliability and meeting all requirements for technical implementation. The physical and logical connections of this process are reflected in each step of flow signal acquisition, processing, and subsequent application, providing a solid data foundation for the entire detection system.
[0033] S2: Quantum noise injection: dynamically generating quantum noise that matches the spectrum of the high-quality multi-source signal and injecting the quantum noise into the signal; Specifically, first, based on the high-quality multi-source signal S obtained in step S1 combined (t), which contains pre-processed traffic information. Quantum noise can be considered to have random and statistical characteristics, and its amplitude can be expressed by the following formula: Where N(t) represents the quantum noise amplitude at time t, α represents the adaptive noise coefficient, and Z(t) represents a Gaussian white noise process with zero mean and unit variance.
[0034] The calculation formula of adaptive noise coefficient α is: In the formula, α represents an adjustable parameter, SNR current Indicates the current signal-to-noise ratio, and 30dB represents the baseline signal-to-noise ratio under normal conditions. In the process of generating quantum noise, first, S combined (t) Perform Fourier transform to obtain the frequency domain, and then calculate the power spectrum density P of the signal signal (f), this process is implemented by Fast Fourier Transform (FFT) to improve computational efficiency. The power spectral density describes the power distribution of the signal at each frequency.
[0035] Next, to ensure that the spectrum of the quantum noise matches the signal, the power spectral density of the quantum noise is set to: In the formula, P noise (f) represents the power spectral density of quantum noise, P signal(f) represents the power spectral density of the signal, and β represents the noise gain factor.
[0036] After the quantum noise is generated, N(t) is injected into the original signal to construct the synthetic signal S final (t), the expression is: S final (t) = S combined (t)+N(t); In the formula, S final (t) represents the time domain representation of the final signal, S combined (t) represents the integrated signal in the time domain, and N(t) represents the noise present in the signal.
[0037] During this process, the injection of quantum noise not only enhances the signal's ability to resist interference but also makes the measurement more accurate by matching it with the signal spectrum. To achieve dynamic generation of quantum noise, the process can also form a closed-loop feedback loop by correlating the flowmeter's data volume with the current noise level in real time. This feedback mechanism can be expressed as follows: Where β(t) represents the dynamic noise gain factor at time t, and k represents the gain adjustment constant.
[0038] The entire implementation process collects signals through the flow meter and undergoes dynamic generation and injection of quantum noise, forming a complete process from signal preprocessing to quantum noise addition, ensuring the logical connection between the generation of quantum noise and signal characteristics, and achieving the best signal processing effect.
[0039] S3: Feature extraction and tensor construction: Extract features from the signal after quantum noise injection, including extracting time domain features and frequency domain features, and constructing a four-dimensional feature tensor containing these features; Specifically, first, based on the synthetic signal S obtained in step S2 final (t), construct a feature extraction model. Feature extraction includes two parts: time domain features and frequency domain features. Key statistics for time domain feature extraction include basic statistics such as the signal's mean, variance, peak value, root mean square value, and instantaneous power.
[0040] The mean μ of the signal can be calculated using the following formula: In the formula, μ represents the signal S final (t) is the average value in the time interval [0,T], Represents the inverse of the time interval, represents the integral from 0 to T, S final (t) represents the final signal at time t.
[0041] The variance of the signal σ 2 Defined as: In the formula, σ 2 Indicates signal S final The variance of (t), (S final (t)-μ) 2 It represents the square of the difference between the value of the signal at time t and its mean μ.
[0042] The peak value P of the signal is defined as: P=max(S final (t)); Where P represents the signal S final The peak value of (t), max(S final (t)) represents all possible S at time t. final (t)The maximum value among the values.
[0043] The formula for calculating the root mean square value R is: Where R represents the root mean square value of the signal.
[0044] Instantaneous power P inst (t) is expressed as: P inst (t) = S final (t) 2 ; In the formula, P inst (t) represents the instantaneous power at time t.
[0045] Frequency domain feature extraction uses fast Fourier transform (FFT) to convert the signal into a frequency domain representation. Frequency domain features include maximum frequency, average frequency, and the energy of specific frequency components in the spectrum. Its spectrum S(f) can be expressed as: In the formula, S(f) represents the signal in the frequency domain, f represents the frequency component of the signal component, represents the integral of time t from negative infinity to positive infinity, S final (t) represents the time domain expression of the final signal, e -j2πft It represents the frequency component of the signal, j is an imaginary unit used to introduce phase information, 2π is a constant used to convert the frequency f into radians, f is the frequency variable obtained by Fourier transform, usually in Hertz, and t is the time variable, representing the instantaneous value of the signal.
[0046] Power spectral density P signal (f) is calculated by the following formula: Psignal (f)=|S(f)| 2 ; In the formula, P signal (f) Signal power distribution in the frequency domain.
[0047] Maximum frequency f in the frequency domain max Defined as: f max =argmax(|S(f)|); In the formula, f max It represents the main frequency of the signal in the frequency domain, and argmax(|S(f)|) refers to finding the parameter that makes the function take the maximum value.
[0048] Average frequency f avg The calculation method is: In the formula, f avg represents the weighted average frequency of the signal, represents the integral from frequency 0 to the maximum frequency F, f·P signal (f) This is the integrand, P signal (f) represents the power spectrum density of the signal at frequency f, Indicates the overall energy of the calculated signal power spectrum.
[0049] After all these time domain and frequency domain feature extractions are completed, they are constructed into a four-dimensional feature tensor The data structure of this tensor is as follows: Where T is the time step, C is the number of sensor channels, F is the feature dimension, and E is the environmental parameter.
[0050] The feature tensor construction process involves gathering the extracted time and frequency domain features and organizing them into arrays that fit within the tensor's dimensions. Each time step corresponds to the flow meter signal and the extracted features, creating a consistent input format that can meet the needs of machine learning models or other algorithms.
[0051] Throughout the implementation process, the physical and logical connections between various technical elements are clearly defined, forming a complete technical chain from signal acquisition and processing to feature extraction and tensor construction. This process ensures comprehensive feature extraction and provides a solid foundation for subsequent analysis and utilization.
[0052] S4: Feature optimization: mapping the four-dimensional feature tensor into an Ising model and optimizing the model features using a quantum algorithm; Specifically, first, based on the four-dimensional feature tensor constructed in step S3 Establish the Ising model. The Ising model is a mathematical model used to describe the state of a system, commonly used in statistical physics. Its mathematical expression is: H=-∑ i,j J ij σ i σ j -∑ i h i σ i ; In the formula, H represents the Hamiltonian of the system, J represents the total energy of the system, ij represents the interaction strength between the i-th and j-th states, σ i represents the i-th spin state, σ j represents the jth spin state, h i represents the influence of the external magnetic field on the i-th spin state.
[0053] The four-dimensional feature tensor is mapped to a spin variable to facilitate its application in the Ising model. This mapping process can be expressed by the following formula: σ i =f(T ijkl ); In the formula, T ijkl is the index of the four-dimensional feature tensor, indicating the specific value of the feature data, and f represents the feature mapping function.
[0054] After the features are mapped to spin states through a specific algorithm, the Ising model is optimized using a quantum algorithm. The core steps of the quantum algorithm include: First, initialize the quantum state. The spin state is represented by quantum bits (qubits), and each spin state corresponds to a quantum bit. The state can be expressed as: |ψ>=∑ i c i |i>; In the formula, |ψ> represents the quantum state wave function, c i represents the probability amplitude of each state, and |i> represents the ground state of the quantum state.
[0055] Secondly, the Ising model is solved by the variational quantum eigensolver, with the lowest energy state as the optimization target. The energy expectation value is expressed as: E(θ)=<ψ(θ)|H|ψ(θ)>; In the formula, E(θ) represents the energy expectation value that depends on the parameter θ, represents the expectation value of the configuration at this energy state, H represents an operator, and |ψ(θ)> represents the parameterized quantum state wave function.
[0056] Finally, an optimization algorithm (such as gradient descent or genetic algorithm) is used to update the quantum state parameters. The optimized model features are extracted, and finally the optimized state variables are obtained, and the four-dimensional feature tensor τ′ is updated, which is expressed as: τ′=g(σ i ); Where: τ′ represents the optimized four-dimensional feature tensor, and g represents the state reconstruction function.
[0057] Throughout the implementation, the four-dimensional feature tensor The establishment, mapping, and optimization of the network have clear physical and logical connections. The feature tensor provides input, the Ising model establishes the coupling relationship between features, and the quantum algorithm improves the overall model performance through optimization.
[0058] S5: Anomaly Detection: Based on the optimized model features, the generative adversarial network of the adversarial learning framework is trained to detect and identify unknown abnormal signals and output corresponding abnormal warning information to obtain detection results; Specifically, first, based on the optimized four-dimensional feature tensor τ′ obtained in step S4, a generative adversarial network model is established. The generative adversarial network consists of a generator G and a discriminator D. The goal of the generator is to generate samples that are similar to the real data distribution, while the goal of the discriminator is to distinguish between the generated samples and the real samples. The output of the generator can be defined by the following formula: G(z)=Generator(z;θ G ); In the formula, G(z) represents the generated feature data, Generator(z; θ G ) represents the expression for calling the generator function, z is the input noise vector, θ G represents the weights of the generator network.
[0059] The discriminator is used to determine whether the input sample is a real sample, and its output is expressed as: D(x)=Discriminator(x;θ D ); In the formula, D(x) represents the authenticity judgment of the input sample x, x is the input feature data, θ D Denotes the weight of the discriminator network, Discriminator(x;θ D ) represents an expression that calls the discriminator function.
[0060] During the training process, the loss function L is used to measure the learning effect of the generator and the discriminator. It is usually expressed as a minimum-maximum loss function: In the formula, L represents the value of the loss function, Represents the expected value of the true sample x, P data Represents the distribution of real data, D(x) is the probability that the discriminator classifies the real sample x, logD(x) is the confidence value of the discriminator for the real sample, represents the expected value of the generated sample, G(z) is the sample generated by the generator, D(G(z)) is the judgment probability of the discriminator on the generated sample, and log(1-D(G(z))) reflects the discriminator's negative confidence in the generated sample.
[0061] This loss function represents the discriminator's ability to detect real samples and the generator's ability to generate fake samples. The process of minimizing the loss includes the following steps: First, fix the generator parameters θ G , by optimizing the parameters θ of the discriminator D D To minimize the loss function: In the formula, represents the optimal parameters of the discriminator, This is a mathematical operation that refers to finding the parameters θ that maximize the loss function L. D .
[0062] Secondly, fix the discriminator parameters θ D , update the generator parameters θ G , so that the generated samples are judged as real samples by the discriminator as much as possible, which can be expressed as: In the formula, represents the optimal parameters of the generator, This is a mathematical operation that means finding the parameters θ that minimize the loss function L. G .
[0063] Optimize by following these steps: Randomly initialize the parameters of the generator and discriminator; In each iteration, we first provide the discriminator with real data and generated data, calculate the loss and update θ D ; Then update the parameters θ of the generator G To optimize the generated samples to make them closer to the real data.
[0064] After multiple iterations of training, the samples generated by the generator G are very similar to the real data, and the discriminator D can also better judge the authenticity of the samples. After the training is completed, the trained generator is used to generate feature data for identifying unknown abnormal signals. new, first extract its features and map them into a four-dimensional feature tensor T new . Then the discriminator D is used to evaluate and calculate the discriminant loss value D(S new ). If D(S new )<∈; In the formula, D(S new ) represents the output of the discriminator D for the new sample S new The predicted probability of , ε represents the acceptance of the discriminator to the new sample.
[0065] Where ∈ is the set abnormal threshold, and the signal is marked as an abnormal signal.
[0066] Finally, the system will generate corresponding abnormal warning information and report the detection results to facilitate subsequent processing or decision-making.
[0067] Throughout the implementation process, the four-dimensional feature tensor τ′ serves as input, and a generative adversarial network provides the model framework. Through adversarial training, the model's ability to detect abnormal signals is optimized. This effective combination of signal processing, model optimization, and anomaly identification techniques provides a highly efficient method for traffic monitoring.
[0068] S6: Verification and Decision-making: Physically verify the detection results by using the water-electricity coupling equation and adjust the communication protocol of flow management in real time according to the detection results; Specifically, first, the abnormal signal is verified using the hydroelectric coupling equation, which is mathematically expressed as: In the formula, represents the rate of change over time t, Express Perform divergence calculation, S(x,t) represents the source term or source term distribution, λ represents a constant coefficient, σ represents a function, Represents electric and magnetic fields.
[0069] During the verification phase, the flow signal and water level data are monitored in real time. These measured data are obtained through the flow meter and water level sensor. By solving the water-electricity coupling equation, the monitored flow Q and flow velocity v are substituted into the equation in the following steps: Calculate the current values of flow rate Q and flow velocity v based on the measured data; Combining velocity and flow rate, we can solve Computes the current input influence of the source term S(x,t).
[0070] Next, we define the residual R between the model prediction value and the actual measurement value, which is expressed as: R=Q measured -Q predicted ; In the formula, Qmeasured Indicates the flow value measured by the sensor, Q predicted It represents the flow prediction value calculated by the hydropower coupling equation.
[0071] The residual R is evaluated. When R exceeds the pre-set threshold ∈, the system will identify it as an abnormal traffic situation. In this case, the system will adjust the traffic management communication protocol in real time. The traffic management communication protocol can be expressed as a set of protocols: P = {p1, p2, ..., p n}; In the formula, P represents the set of traffic management protocols, including different traffic monitoring and regulation strategies, n Indicates the characteristics of the nth protocol.
[0072] When abnormal traffic is detected, the following logic is used to select the appropriate communication protocol: If R>∈, switch to p i Here, ∈ is a threshold value that defines the acceptable residual range. The specific value can be set based on historical data statistics or experience. After switching to the new protocol, the system can achieve higher flexibility and security in the face of emergencies.
[0073] Throughout the implementation process, the physical and logical connections between various technical elements are clearly defined. Real-time data collected by sensors is then input into the hydropower coupling model for verification. If anomalies are detected, the communication protocol is adaptively adjusted based on the verification results, enabling the flow management system to strike a balance between safety and accuracy.
[0074] Please see the attached Figure 2 The present invention also provides a water meter detection system with a multi-pulse flowmeter, comprising: A plurality of pulse flow meters of different ranges are configured to collect flow data and connected to the preprocessing module; The pre-processing module is used to remove electromagnetic interference and noise from the collected signals and output highly clear multi-source signals to the quantum noise generation and injection module; Quantum noise generation and injection module, dynamically generates and injects quantum noise, and outputs the injected signal to the feature extraction and tensor construction module; A feature extraction and tensor construction module is used to extract features and construct feature tensors based on the feature extraction step, and output the feature tensors to the quantum optimization module for optimization; The quantum optimization module uses the Ising model to optimize the extracted features and outputs the optimized features to the anomaly detection module. The anomaly detection module uses a generative adversarial network to detect and identify unknown anomalies according to the anomaly detection steps and outputs the detection results to the decision output module. Physical verification module, which verifies the monitoring results through the hydropower coupling equation and associates them with the decision module; The decision output module adjusts the communication protocol in real time based on the monitoring results to optimize traffic management.
[0075] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.
Claims
1. A water meter detection method and system with a multi-pulse flowmeter, characterized in that: The following steps are involved: S1: Signal acquisition and preprocessing: Flow data is collected through three pulse flow meters with different ranges, and the flow data is preprocessed to remove electromagnetic interference and high-frequency noise to obtain high-quality multi-source signals; S2: Quantum noise injection: dynamically generating quantum noise that matches the spectrum of the high-quality multi-source signal and injecting the quantum noise into the signal; S3: Feature extraction and tensor construction: Extract features from the signal after quantum noise injection, including extracting time domain features and frequency domain features, and constructing a four-dimensional feature tensor containing these features; S4: Feature optimization: mapping the four-dimensional feature tensor into an Ising model and optimizing the model features using a quantum algorithm; S5: Anomaly Detection: Based on the optimized model features, the generative adversarial network of the adversarial learning framework is trained to detect and identify unknown abnormal signals and output corresponding abnormal warning information to obtain detection results; S6: Verification and Decision-making: Physically verify the detection results by using the water-electricity coupling equation, and adjust the communication protocol for flow management in real time based on the detection results.
2. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: The pre-processing step after signal acquisition includes using a low-pass filter with a cut-off frequency set to no more than 10 kHz and a band-stop filter with a center frequency set to 50 Hz.
3. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: In the process of dynamically generating quantum noise, the calculation formula of the adaptive noise coefficient α used is: In the formula, α represents an adjustable parameter, SNR current Indicates the current signal-to-noise ratio. 30dB represents the normal signal-to-noise ratio baseline.
4. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: The feature tensor is constructed as follows: Where T is the time step, C is the number of sensor channels, F is the feature dimension, and E is the environmental parameter.
5. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: In step S4, the feature optimization step uses an implementation of the Ising model to calculate feature weights using a quantum algorithm.
6. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: The training of the generative adversarial network in the anomaly detection step includes identifying and classifying abnormal signals to generate corresponding early warning information.
7. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: In step S6, the water-electricity coupling equation is used to verify the flow measurement results, and the equation is: In the formula, represents the rate of change over time t, Express Perform divergence calculation, S(x,t) represents the source term or source term distribution, λ represents a constant coefficient, σ represents a function, Represents electric and magnetic fields.
8. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: In step S6, the dynamic protocol switching used in the traffic management optimizes the communication protocol selection according to the detection results to reduce energy consumption and response delay.
9. A water meter detection method and system with a multi-pulse flowmeter according to claim 1, characterized in that: In step S3, the feature extraction step further includes extracting frequency domain features using a short-time Fourier transform method.
10. A water meter detection system with a multi-pulse flowmeter, applied to a water meter detection method with a multi-pulse flowmeter according to any one of claims 1 to 9, characterized in that: include: A plurality of pulse flow meters of different ranges are configured to collect flow data and connected to the preprocessing module; The pre-processing module is used to remove electromagnetic interference and noise from the collected signals and output highly clear multi-source signals to the quantum noise generation and injection module; Quantum noise generation and injection module, dynamically generates and injects quantum noise, and outputs the injected signal to the feature extraction and tensor construction module; A feature extraction and tensor construction module is used to extract features and construct feature tensors based on the feature extraction step, and output the feature tensors to the quantum optimization module for optimization; The quantum optimization module uses the Ising model to optimize the extracted features and outputs the optimized features to the anomaly detection module; An anomaly detection module, which uses a generative adversarial network to detect and identify unknown anomalies according to the anomaly detection step, and outputs the detection results to the decision output module; Physical verification module, which verifies the monitoring results through the hydropower coupling equation and associates them with the decision module; The decision output module adjusts the communication protocol in real time based on the monitoring results to optimize traffic management.