Intelligent water meter battery capacity dynamic monitoring method and system based on multiple parameters

Abstract

The invention discloses an intelligent water meter battery capacity dynamic monitoring method and system based on multiple parameters, and relates to the technical field of intelligent water meter battery monitoring. According to the method, multi-dimensional operation parameters are collected, feature extraction quality is optimized in cooperation with a temperature change model and a mean filtering algorithm, the influence of the temperature on self-discharge is accurately quantified in combination with an Arrhenius equation and linear regression, nonlinear changes of the battery capacity are captured with the help of the time sequence modeling capacity of an LSTM neural network, and the self-discharge performance of the battery is improved. Dynamic compensation of the self-discharge effect is achieved, the accuracy and the fitting actual degree of battery capacity prediction are greatly improved, meanwhile, the intelligent water meter outage risk caused by battery faults is avoided in advance through a three-level early warning mechanism, the equipment power supply stability is effectively guaranteed, and the problems that traditional battery monitoring is single in dimension, only depends on voltage parameters, and the monitoring precision is low are solved. The obvious influence of the environment temperature, the energy consumption of the communication module and the like on the battery capacity is ignored, so that the state evaluation is one-sided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent water meter battery monitoring, and in particular to an intelligent water meter battery capacity dynamic monitoring method and system based on multiple parameters. BACKGROUND

[0002] Intelligent water meters are usually powered by batteries, and the stability of the battery capacity directly affects the normal operation of the water meter. Once the battery capacity is insufficient, it may cause problems such as interruption of water meter data collection and communication failure, affecting water metering and management efficiency. According to the provisions of "Drinking Cold Water Meters and Hot Water Meters" Part 1, manufacturers need to set a low battery power indicator or replacement date prompt on the water meter, and the time from the prompt to the complete depletion of the battery should not be less than 180 days. However, the existing battery capacity monitoring has the following problems:

[0003] 1. High false positive rate of battery voltage monitoring. Intelligent water meters often use lithium sub-batteries as power sources, and their discharge curves exhibit nonlinear characteristics. Battery voltage is affected by factors such as temperature and load current, resulting in a high false positive rate.

[0004] 2. Lack of dynamic fusion and early warning mechanism for historical data and real-time working conditions. Battery failure is discovered passively, which requires maintenance personnel to frequently inspect and troubleshoot, increasing maintenance costs and potentially causing metering data loss due to sudden battery failure. SUMMARY

[0005] The present application aims to provide an intelligent water meter battery capacity dynamic monitoring method and system based on multiple parameters to improve the above technical problems.

[0006] To achieve the above application purpose, the embodiments of the present application provide the following technical solutions:

[0007] An intelligent water meter battery capacity dynamic monitoring method based on multiple parameters includes:

[0008] Timely collect multiple dimension running parameters of the intelligent water meter battery through sensors; the multiple dimension running parameters include battery voltage, working current, ambient temperature, reference signal received power and transmit power;

[0009] Construct a temperature variation model, combine the mean filtering algorithm to extract features of the multiple dimension running parameters, and generate a multiple dimension running feature vector;

[0010] Based on the Arrhenius equation and linear regression method, calculate the self-discharge coefficient; based on the LSTM neural network, construct a battery capacity prediction model;

[0011] Based on the self-discharge coefficient, predict the multiple dimension running feature vector through the battery capacity prediction model, and generate a battery capacity prediction result;

[0012] The battery of the intelligent water meter is monitored based on the battery capacity prediction result and the battery voltage in combination with a three-level early warning mechanism.

[0013] In the above scheme, the application collects multi-dimensional operation parameters, optimizes the feature extraction quality by combining a temperature change model and a mean filtering algorithm, accurately quantifies the influence of temperature on self-discharge by means of the Arrhenius equation and linear regression, captures the nonlinear change of battery capacity by means of the time series modeling capability of the LSTM neural network, realizes dynamic compensation of the self-discharge effect, greatly improves the accuracy and actual fitting degree of battery capacity prediction, and effectively guarantees the stability of equipment power supply by avoiding the risk of intelligent water meter shutdown caused by battery failure in advance through a three-level early warning mechanism. The application solves the problem of single monitoring dimension of traditional batteries, which only relies on voltage parameters and ignores the significant influence of environmental temperature, communication module energy consumption and the like on battery capacity, resulting in one-sided state evaluation. The application realizes accurate quantification of temperature and self-discharge effect and corrects the capacity prediction deviation caused by the large difference in battery self-discharge rate at different temperatures.

[0014] Further, the generating a multi-dimensional operation feature vector comprises:

[0015] The historical environmental temperature of the intelligent water meter battery under various working conditions is collected, a temperature change model is constructed based on a deterministic function and a to-be-estimated parameter vector in combination with an autoregressive model;

[0016] The environmental temperature is abnormally identified and corrected by the temperature change model to generate a temperature feature vector;

[0017] The temperature feature vector, the battery voltage, the working current, the reference signal received power and the transmission power are processed by Min-Max linear normalization to generate a multi-dimensional operation feature vector.

[0018] In the above scheme, the temperature feature is ensured to be accurate and reliable by abnormally identifying and correcting the environmental temperature by the temperature change model to effectively eliminate the temperature abnormal value caused by extreme working conditions or sensor error. The dimension difference of different parameters such as battery voltage, working current and communication power is eliminated by Min-Max linear normalization, so that each feature has balanced weight in model training, which significantly improves the overall quality of the multi-dimensional operation feature vector. The generated feature vector is combined with the time series capability of the LSTM model to provide high-quality data support for capacity prediction, which greatly improves the stability and accuracy of battery capacity prediction. The technical problems of temperature data abnormal interference in traditional feature processing, model training deviation caused by inconsistent dimensions of multiple parameters, and low-quality features that cannot effectively support accurate capacity evaluation can be solved.

[0019] Further, the constructing a battery capacity prediction model comprises:

[0020] Based on a preset fixed interval time, the remaining battery capacity of a smart water meter in an open-circuit state at different temperatures is measured using a high-precision coulomb meter.

[0021] Based on the remaining battery capacity at different temperatures, corresponding simulated self-discharge coefficients are generated.

[0022] Based on the simulated self-discharge coefficients and the self-discharge coefficients at the reference temperature, the corresponding logarithmic fitting equations for the self-discharge coefficients are constructed using the Arrhenius equation.

[0023] The temperature sensitivity coefficient is calculated using a linear regression method; based on the temperature sensitivity coefficient and the current ambient temperature, the self-discharge coefficient at the current moment is generated by fitting an equation with a logarithmic self-discharge coefficient.

[0024] An LSTM neural network was constructed and trained using the backpropagation time algorithm to generate a battery capacity prediction model.

[0025] Furthermore, the battery capacity prediction model includes:

[0026] Collect training multi-dimensional operating parameters and their labels under different working conditions and temperatures;

[0027] Based on the temperature features in the labels and the multi-dimensional running feature vectors of training, the corresponding remaining capacitance-temperature curve is plotted.

[0028] The remaining capacitance-temperature curve is converted to generate a capacitance-temperature feature that is identical to other feature dimensions in the multi-dimensional running feature vector used for training and can be used for model prediction.

[0029] Update the multi-dimensional running feature vector based on capacity-temperature features;

[0030] The updated training multi-dimensional running feature vector is input into the LSTM neural network and trained using the backpropagation time algorithm to generate a battery capacity prediction model.

[0031] In the above scheme, to transform the nonlinear effect of temperature on self-discharge into a calculable quantitative index, breaking the limitations of traditional static compensation and allowing the model to adapt to self-discharge differences under different temperatures in real time, this invention introduces a temperature sensitivity coefficient and a self-discharge coefficient. A simulated self-discharge coefficient is generated by measuring the remaining capacity in an open circuit. This is combined with the Arrhenius equation and linear regression to accurately quantify the temperature sensitivity coefficient, achieving dynamic generation of the self-discharge coefficient with the current ambient temperature. Simultaneously, capacity-temperature features adapted to the model are generated through capacity-temperature curve transformation and the training vector is updated. Then, an LSTM is trained using the BPTT algorithm. This strengthens the quantitative correlation between self-discharge effects and temperature and capacity, and improves the adaptability of features to the model's temporal modeling capabilities, significantly enhancing the dynamic adaptability and accuracy of capacity prediction. Furthermore, introducing the temperature sensitivity coefficient and the self-discharge coefficient and combining them with the output of the LSTM neural network solves the problems of insufficient generalization caused by static self-discharge compensation and discrete temperature features in traditional prediction, as well as the difficulty in accurately capturing capacity change patterns when multiple parameters are coupled with the self-discharge effect.

[0032] Furthermore, the generated battery capacity prediction result includes:

[0033] Based on the temperature data in the multi-dimensional running feature vector, the corresponding temperature curve is plotted, and the multi-dimensional running feature vector is updated.

[0034] The updated multi-dimensional running feature vector is input into the battery capacity prediction model to generate the first battery capacity prediction result.

[0035] Calculate the time interval between the last forecast and the current forecast;

[0036] Based on the time interval, self-discharge coefficient, and the first battery capacity prediction result, the battery capacity prediction result is calculated.

[0037] In the above scheme, based on the first battery capacity prediction result, the capacity loss caused by self-discharge within the prediction period is accurately deducted by combining the self-discharge coefficient and time interval. This separates the inherent characteristic of self-discharge from the load consumption, quantifies the impact of time and temperature on natural capacity loss, realizes dynamic compensation for natural battery loss, and greatly improves the fit between the capacity prediction result and the actual remaining capacity.

[0038] A multi-parameter-based smart water meter battery capacity dynamic monitoring system is divided into a top-down application layer, network layer, and sensing layer; wherein:

[0039] The sensing layer is used to periodically collect multi-dimensional operating parameters of the smart water meter battery through sensors; these multi-dimensional operating parameters include battery voltage, operating current, ambient temperature, reference signal receiving power, and transmitting power.

[0040] The network layer is used to transmit multi-dimensional operating parameters to the application layer;

[0041] The application layer is used to manage smart water meters, analyze and monitor multi-dimensional operating parameter data, generate monitoring results, and implement them.

[0042] Furthermore, the application layer includes:

[0043] The feature generation unit is used to construct a temperature change model and combine the mean filtering algorithm to extract features from multi-dimensional operating parameters and generate multi-dimensional operating feature vectors.

[0044] The self-discharge coefficient calculation unit is used to calculate the self-discharge coefficient based on the Arrhenius equation and linear regression method.

[0045] The model building unit is used to build a battery capacity prediction model based on an LSTM neural network.

[0046] The prediction unit is used to predict the multi-dimensional operating feature vector based on the self-discharge coefficient through the battery capacity prediction model, and generate the battery capacity prediction result.

[0047] The monitoring unit is used to monitor the smart water meter battery based on the battery capacity prediction results and battery voltage, combined with a three-level early warning mechanism.

[0048] Furthermore, the perception layer includes:

[0049] The timer is used to set the data collection interval.

[0050] The MCU microcontroller is used to generate battery acquisition commands and temperature acquisition commands, and to control the network layer for data interaction through a serial interface, acquiring the reference signal received power and transmitted power of the smart water meter.

[0051] A resistor voltage divider circuit is used to implement the battery acquisition command issued by the MCU microcontroller and acquire the battery voltage of the smart water meter battery.

[0052] The temperature sensor is used to implement temperature acquisition commands issued by the MCU microcontroller and collect the ambient temperature of the smart water meter battery.

[0053] Furthermore, the resistor divider circuit includes a MOSFET T1 of model SP3407S; the first port of T1 is connected to the battery voltage VBAT; the second port of T1 is connected to one end of resistor R1, and the other end of resistor R1 serves as the control signal pin V_Ctr; the third port of T1 is connected to one end of resistor R2, and the other end of resistor R2 is connected to one end of resistor R4 and grounding resistor R3 respectively; the other end of resistor R4 is connected to grounding capacitor C1 and serves as the voltage detection output pin V_Det.

[0054] In the above scheme, this system uses multiple components such as a resistive voltage divider circuit and a temperature sensor in the sensing layer to periodically collect multi-dimensional operating parameters such as battery voltage, operating current, ambient temperature, and communication power. After reliable transmission through the network layer, the feature generation unit in the application layer optimizes feature quality using a temperature change model and mean filtering. The self-discharge coefficient calculation unit combines the Arrhenius equation and linear regression to accurately quantify the impact of temperature on self-discharge. The LSTM neural network in the model building unit captures the temporal variation law of capacity. Finally, the prediction and monitoring unit generates accurate battery capacity prediction results and implements dynamic monitoring in conjunction with a three-level early warning mechanism. This achieves comprehensive collection of multi-dimensional parameters, dynamic compensation for self-discharge effects, accurate prediction of battery capacity, and timely triggering of graded early warnings. It effectively solves the problems of single monitoring dimensions, inaccurate capacity prediction, and delayed early warning in traditional systems, greatly improving the accuracy and timeliness of smart water meter battery status monitoring and ensuring the stability of equipment power supply. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of the method in an embodiment of the present invention;

[0057] Figure 2 This is a resistor voltage divider circuit diagram in an embodiment of the present invention;

[0058] Figure 3 This is a graph showing the remaining capacitance of the capacitor versus temperature in an embodiment of the present invention.

[0059] Figure 4 This is a system structure diagram in an embodiment of the present invention;

[0060] Figure 5 This is an application layer structure diagram in an embodiment of the present invention;

[0061] Figure 6 This is a structural diagram of the sensing layer in an embodiment of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0063] Please see Figure 1 This embodiment provides a method for dynamic monitoring of the battery capacity of a smart water meter based on multiple parameters. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this.

[0064] like Figure 1 As shown, a method for dynamic monitoring of the battery capacity of a smart water meter based on multiple parameters includes:

[0065] S1. Collect multi-dimensional operating parameters of the smart water meter battery at regular intervals using sensors; the multi-dimensional operating parameters include battery voltage, operating current, ambient temperature, reference signal receiving power, and transmitting power;

[0066] Specifically, battery voltage, operating power consumption, and ambient temperature are measured every hour. Operating power consumption includes the power consumption (average current consumption) and continuous operating time of each functional module in the smart water meter. These multi-dimensional operating parameters are saved to memory and uploaded to the platform / cloud via low-power networks (NB-IoT, LoRa). The functional modules in the smart water meter are a communication module, a metering module, a threshold orifice module, a display module, and a storage module.

[0067] S1 includes:

[0068] S1-1. Construct a resistor voltage divider circuit, adjust the battery voltage of the resistor voltage divider circuit using an MCU microcontroller, and calculate the battery voltage.

[0069] like Figure 2 As shown, the resistor divider circuit includes a MOSFET T1 of model SP3407S; the first port of T1 is connected to the battery voltage VBAT; the second port of T1 is connected to one end of resistor R1, and the other end of resistor R1 serves as the control signal pin V_Ctr; the third port of T1 is connected to one end of resistor R2, and the other end of resistor R2 is connected to one end of resistor R4 and grounding resistor R3 respectively; the other end of resistor R4 is connected to grounding capacitor C1 and serves as the voltage detection output pin V_Det.

[0070] During operation, the control signal pin V_Ctr outputs high and low levels to control the conduction and cutoff of MOSFET T1 (conduction starts data acquisition, cutoff disconnects the circuit to reduce power consumption). In the conduction state, the battery voltage VBAT is divided by resistors R2 and R3 in series. The voltage at the intermediate node after the voltage division is output to the MCU's ADC acquisition module via the voltage detection output pin V_Det. Simultaneously, it is filtered by an RC circuit composed of resistor R4 and capacitor C1 to eliminate interference. Based on Ohm's law and the voltage division principle, the measured voltage is acquired by the ADC. Then, according to the formula:

[0071] ;

[0072] Calculate the battery voltage It can accurately reproduce the actual battery voltage, ensuring that the ADC acquisition module is protected from overvoltage damage due to the input voltage matching range. It also ensures the accuracy of voltage data by using the 1% high-precision resistors R2 and R3 and filtering processing, providing reliable raw data for the subsequent model to predict the remaining battery capacity. At the same time, it reduces invalid power consumption and extends battery life by working with the timing control of V_Ctr.

[0073] S1-2, The ambient temperature is collected by the temperature sensor controlled by the MCU microcontroller;

[0074] S1-3. Collect the operating power consumption, reference signal receiving power, and transmission power of each functional module in the smart water meter.

[0075] S2. Construct a temperature change model, and combine the mean filtering algorithm to extract features from multi-dimensional operating parameters to generate multi-dimensional operating feature vectors.

[0076] S2 includes:

[0077] S2-1. Collect the historical ambient temperature of the smart water meter battery under various working conditions, and construct a temperature change model based on the deterministic function and the vector of parameters to be estimated, combined with an autoregressive model.

[0078] Specifically, the parameter vector to be estimated is solved using the least squares method or the maximum likelihood estimation method. Based on historical environmental temperatures and the parameter vector to be estimated, a deterministic function is generated, resulting in an autoregressive model (temperature change model). The formula corresponding to the temperature change model is:

[0079] ;

[0080] ;

[0081] in, This indicates that the mean is 0 and the variance is 0. White noise distribution, Indicates the current time Lower residual (noise term) Indicates the current time The ambient temperature below, , Representing time respectively ,time Corresponding historical ambient temperature, Represents a constant. , Representing the first autoregressive coefficient and the second, respectively. One autoregressive coefficient.

[0082] S2-2. Anomalies in ambient temperature are identified and corrected using a temperature change model, generating a temperature feature vector.

[0083] Specifically, an ambient temperature is predicted using a temperature change model to obtain a predicted temperature value. The difference between the ambient temperature and the corresponding predicted temperature value is calculated, and it is determined whether the absolute value of the difference is less than a preset residual threshold. If so, the ambient temperature is retained and used as the temperature feature vector. Otherwise, the predicted temperature replaces the ambient temperature and is used as the temperature feature vector.

[0084] S2-3. The temperature feature vector, battery voltage, operating current, reference signal received power, and transmitted power are processed by Min-Max linear normalization to generate a multi-dimensional operating feature vector.

[0085] S3. Calculate the self-discharge coefficient based on the Arrhenius equation and linear regression method; construct a battery capacity prediction model based on LSTM neural network.

[0086] S3 includes:

[0087] S3-1. Based on a preset fixed interval time, the remaining battery capacity of a smart water meter battery in an open circuit state is measured at different temperatures using a high-precision coulomb meter.

[0088] Specifically, when the smart water meter battery is in an open-circuit state, with no external load and the circuit completely disconnected, the battery capacity change is solely caused by self-discharge (natural loss due to internal chemical reactions), accurately isolating interference from load power consumption. Therefore, this invention selects to measure the remaining capacity of the smart water meter battery in an open-circuit state. By measuring the remaining capacity at different temperatures using a high-precision coulomb meter at fixed intervals, the self-discharge coefficient (SDR, the proportion of capacity consumed by self-discharge per unit time) at each temperature can be accurately calculated. Combined with measurement data from multiple temperature points, a quantitative relationship between "temperature and self-discharge coefficient" (such as the Arrhenius equation) can be established, ultimately providing clean and reliable basic data for the self-discharge compensation of the LSTM model. This ensures that the battery remaining capacity prediction accurately adapts to real-time operating temperatures, improving the overall accuracy and reliability of smart water meter battery capacity monitoring.

[0089] S3-2. Based on the remaining battery capacity at different temperatures, generate the corresponding simulated self-discharge coefficient;

[0090] In this embodiment, the proportion of capacity consumed by self-discharge per unit time is calculated based on the remaining battery capacity at different temperatures, and this proportion is used as the simulated self-discharge coefficient. Finally, the temperature, each measurement time, and their corresponding simulated self-discharge coefficients are integrated to create a table, namely Table 1. In subsequent processing, the corresponding simulated self-discharge coefficient is searched from Table 1 using the temperature and remaining battery capacity.

[0091] Table 1

[0092]

[0093] Table 1 shows the self-discharge coefficients obtained under different temperature conditions based on preset fixed intervals. Here, n represents the total number of samples collected within each fixed interval, and t... ij Q represents the time of the j-th measurement at the i-th temperature point. ij SDR represents the remaining battery capacity measured at the j-th temperature point. ij This represents the simulated self-discharge coefficient for the j-th measurement at the i-th temperature point.

[0094] S3-3. Based on the simulated self-discharge coefficients and the self-discharge coefficients at the reference temperature, the corresponding logarithmic fitting equations for the self-discharge coefficients are constructed using the Arrhenius equation.

[0095] Specifically, the formula corresponding to S3-3 is:

[0096] ;

[0097] ;

[0098] in, Indicates reference temperature. This represents the self-discharge coefficient at the reference temperature. This represents the self-discharge coefficient at the current moment. Indicates the temperature sensitivity coefficient. This indicates the ambient temperature at the current moment. Indicated by the natural constant as the base The logarithmic function.

[0099] S3-4. Calculate the temperature sensitivity coefficient using the linear regression method; based on the temperature sensitivity coefficient and the current ambient temperature, generate the self-discharge coefficient at the current moment by fitting the equation with the logarithm of the self-discharge coefficient.

[0100] While LSTM neural networks, as purely data-driven models, excel at capturing the temporal dependencies of battery capacity changes, they are susceptible to incomplete data coverage and noise interference, making it difficult to accurately reflect the actual physical characteristic that "battery self-discharge is an inherent physical loss." In contrast, the self-discharge coefficient, derived from the Arrhenius equation and linear regression, anchors the deterministic physical law between temperature and battery self-discharge. Combining the self-discharge coefficient with LSTM neural networks integrates the temporal learning capabilities of data-driven models with the deterministic common sense of physical models, compensating for the shortcomings of pure data models that deviate from the actual physical loss laws of batteries. This results in predictions that both align with data trends and conform to the true physical logic of battery self-discharge, making them more accurate and reliable.

[0101] S3-5. Construct an LSTM neural network and train it using the backpropagation time algorithm to generate a battery capacity prediction model.

[0102] S3-5 includes:

[0103] S3-5-1. Collect training multi-dimensional operating parameters and their labels (real battery remaining capacity) under different operating conditions and temperatures, and generate corresponding training multi-dimensional operating feature vectors through the same processing as S2.

[0104] S3-5-2. Based on the temperature features in the label and training multi-dimensional running feature vector, plot the corresponding remaining capacitance-temperature curve, such as... Figure 3 As shown;

[0105] Specifically, directly using temperature features for model prediction and training at each temperature point results in a large amount of data, a heavy workload, and enormous computational resources required. However, plotting a capacitor's remaining capacity versus temperature curve provides prior knowledge by intuitively identifying the correlation (e.g., linear, exponential, or piecewise nonlinear), avoiding blind learning by the model in high-dimensional raw data. Furthermore, key parameters (such as temperature sensitivity coefficients) can be extracted by fitting mathematical formulas, transforming multiple temperature points into a few fitting parameters. This simplifies feature dimensions, reduces computational complexity, and allows for the addition of reasonable values ​​for sparse temperature points through smooth fitting, alleviating data sparsity. Based on this, targeted features (such as power transformations of temperature) can be constructed, enabling the model to more efficiently capture core dependencies. Ultimately, this solves the problems of large data volume, heavy workload, and enormous computational resource consumption caused by directly training at each temperature point, improving model training efficiency and generalization ability.

[0106] S3-5-3. Convert the format of the remaining capacitance-temperature curve to generate a capacitance-temperature feature that is the same as other feature dimensions in the multi-dimensional running feature vector of the training and can be used for model prediction.

[0107] S3-5-4. Based on the capacity-temperature feature, update the training multi-dimensional running feature vector, that is, replace the temperature feature in the training multi-dimensional running feature vector with the capacity-temperature feature.

[0108] S3-5-5. Input the updated training multi-dimensional running feature vector into the LSTM neural network, train it using the backpropagation time algorithm, and generate a battery capacity prediction model.

[0109] Specifically, after the updated training multi-dimensional running feature vector is input into the LSTM neural network, the LSTM neural network selectively memorizes and forgets the features at each time step through the synergistic effect of the forget gate, input gate, and output gate. This captures the long-term temporal dependencies between features (such as the dynamic correlation between capacity-temperature features and battery remaining capacity, and the change law of voltage over time), and transforms the temporal information into a hidden state vector to extract core features. Subsequently, the backpropagation time (BPTT) algorithm first calculates the error between the battery capacity prediction value output by the model and the actual remaining capacity (label), and then propagates this error back along the time dimension to each time step. The gradient of all trainable parameters, such as the LSTM gating parameters and hidden layer weights, is calculated through the chain rule. Based on the gradient information, the parameters are adjusted to minimize the prediction error. After multiple rounds of iterative training until the model converges, a prediction model that can accurately map temporal features and battery remaining capacity is finally generated.

[0110] S4. Based on the self-discharge coefficient, the multi-dimensional operating feature vector is predicted using the battery capacity prediction model to generate a battery capacity prediction result; the battery capacity prediction result is the remaining capacity of the battery.

[0111] S4 includes:

[0112] S4-1. Based on the temperature data in the multi-dimensional running feature vector, draw the corresponding temperature curve and update the multi-dimensional running feature vector; the processing procedure of S4-1 is the same as that of S3-5-2 to S3-5-4.

[0113] S4-2. Input the updated multi-dimensional running feature vector into the battery capacity prediction model to generate the first battery capacity prediction result;

[0114] S4-3, Calculate the time interval between the previous forecast and the current forecast. ;

[0115] S4-4. Prediction results based on time interval, self-discharge coefficient, and first battery capacity. Calculate the predicted battery capacity. .

[0116] The formula corresponding to S4-4 is:

[0117] .

[0118] S5. Based on the battery capacity prediction results and battery voltage, the smart water meter battery is monitored in conjunction with a three-level early warning mechanism.

[0119] The three-tiered early warning mechanism consists of three levels: Level 1, Level 2, and Level 3. Level 1 early warning indicates the smart water meter can operate for at least 180 days. The platform sends a notification to the water utility's maintenance department and generates a battery replacement recommendation list, reminding the water utility to replace the battery within these 180 days. Level 2 early warning occurs when the predicted battery capacity is less than 5% and the battery voltage is less than 3.4V. This indicates the smart water meter is operating normally with only metering function enabled. The meter records a Level 1 low-voltage event, requiring an alarm and immediate battery replacement. Level 3 early warning occurs when the predicted battery capacity is less than 2% and the battery voltage is less than 3.2V. This indicates the smart water meter is in a low-power state. The meter records a Level 2 low-voltage event and reminds the user to replace the battery.

[0120] In summary, this invention collects parameters such as battery voltage, operating energy consumption, ambient temperature, and communication power from multiple dimensions. It optimizes feature quality using a temperature change model and mean filtering, and combines the Arrhenius equation and LSTM neural network to accurately quantify the self-discharge effect and capture capacity time-series changes. Furthermore, it dynamically corrects the prediction results through time intervals and self-discharge coefficients. Finally, with a three-level early warning mechanism, it achieves comprehensive monitoring of the smart water meter's battery status, accurate capacity prediction, and tiered early warning, significantly improving the accuracy and timeliness of battery monitoring, effectively ensuring the stability of equipment power supply, and reducing operation and maintenance costs.

[0121] like Figure 4 As shown, a multi-parameter-based smart water meter battery capacity dynamic monitoring system is divided into a top-down application layer, network layer, and sensing layer; wherein:

[0122] The sensing layer is used to periodically collect multi-dimensional operating parameters of the smart water meter battery through sensors; these multi-dimensional operating parameters include battery voltage, operating current, ambient temperature, reference signal receiving power, and transmitting power.

[0123] The network layer is used to transmit multi-dimensional operating parameters to the application layer; the network layer can use NB-IoT network, CAT.1 communication and LoRa communication to transmit multi-dimensional operating parameters.

[0124] The application layer is used to manage smart water meters, analyze and monitor multi-dimensional operating parameter data, generate monitoring results, and implement them.

[0125] like Figure 5 As shown, the application layer includes:

[0126] The feature generation unit is used to construct a temperature change model and combine the mean filtering algorithm to extract features from multi-dimensional operating parameters and generate multi-dimensional operating feature vectors.

[0127] The self-discharge coefficient calculation unit is used to calculate the self-discharge coefficient based on the Arrhenius equation and linear regression method.

[0128] The model building unit is used to build a battery capacity prediction model based on an LSTM neural network.

[0129] The prediction unit is used to predict the multi-dimensional operating feature vector based on the self-discharge coefficient through the battery capacity prediction model, and generate the battery capacity prediction result.

[0130] The monitoring unit is used to monitor the smart water meter battery based on the battery capacity prediction results and battery voltage, combined with a three-level early warning mechanism.

[0131] like Figure 6 As shown, the perception layer includes:

[0132] The timer is used to set the data collection interval.

[0133] The MCU microcontroller is used to generate battery acquisition commands and temperature acquisition commands, and to control the network layer for data interaction through a serial interface, acquiring the reference signal received power and transmitted power of the smart water meter.

[0134] A resistor voltage divider circuit is used to implement the battery acquisition command issued by the MCU microcontroller and acquire the battery voltage of the smart water meter battery.

[0135] The temperature sensor is used to implement temperature acquisition commands issued by the MCU microcontroller and collect the ambient temperature of the smart water meter battery.

[0136] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0137] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0138] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-parameter based intelligent water meter battery capacity dynamic monitoring method, characterized in that, The application relates to a battery capacity prediction method and device for an intelligent water meter. The method comprises the following steps: Collecting multi-dimensional operation parameters of the battery of the intelligent water meter through sensor timing; The multi-dimensional operation parameters include battery voltage, working current, ambient temperature, reference signal receiving power and transmitting power; A temperature change model is constructed, and a mean filtering algorithm is used to extract features of the multi-dimensional operation parameters to generate a multi-dimensional operation feature vector; Based on the Arrhenius equation and the linear regression method, the self-discharge coefficient is calculated; Based on the LSTM neural network, a battery capacity prediction model is constructed; Based on the self-discharge coefficient, the multi-dimensional operation feature vector is predicted through the battery capacity prediction model to generate a battery capacity prediction result; 2. The multi-parameter based intelligent water meter battery capacity dynamic monitoring method according to claim 1, characterized in that, Based on the battery capacity prediction result and the battery voltage, a three-level early warning mechanism is used to monitor the battery of the intelligent water meter. The method for generating the multi-dimensional operation feature vector comprises the following steps: Collecting historical ambient temperatures of the battery of the intelligent water meter under various working conditions, constructing a temperature change model based on a deterministic function and a to-be-estimated parameter vector, and combining an autoregressive model; Abnormalities in the ambient temperature are identified and corrected through the temperature change model to generate a temperature feature vector; 3. The multi-parameter based intelligent water meter battery capacity dynamic monitoring method according to claim 1, characterized in that, The temperature feature vector, the battery voltage, the working current, the reference signal receiving power and the transmitting power are processed through Min-Max linear normalization to generate the multi-dimensional operation feature vector. The method for constructing the battery capacity prediction model comprises the following steps: Based on a preset fixed interval time, the residual capacity of the battery of the intelligent water meter under different temperatures in an open circuit state is measured through a high-precision coulomb counter; Based on the residual capacity of the battery under different temperatures, corresponding simulated self-discharge coefficients are generated; Based on the simulated self-discharge coefficients and the self-discharge coefficient under a reference temperature, a logarithmic fitting equation of the self-discharge coefficient is constructed through the Arrhenius equation; A temperature-sensitive coefficient is calculated through the linear regression method; based on the temperature-sensitive coefficient and the ambient temperature at the current moment, the self-discharge coefficient at the current moment is generated through the logarithmic fitting equation of the self-discharge coefficient; 4. The multi-parameter based intelligent water meter battery capacity dynamic monitoring method according to claim 3, characterized in that, An LSTM neural network is constructed and trained through a back propagation through time algorithm to generate the battery capacity prediction model. The method for generating the battery capacity prediction model comprises the following steps: Collecting training multi-dimensional operation parameters and labels under different working conditions and different temperatures; Based on the labels and the temperature features in the training multi-dimensional operation feature vector, a corresponding residual capacity-temperature curve graph is drawn; The residual capacity-temperature curve graph is format-converted to generate a capacity-temperature feature which has the same dimension as other feature dimensions in the training multi-dimensional operation feature vector and can be used for model prediction; Based on the capacity-temperature feature, the training multi-dimensional operation feature vector is updated; 5. The multi-parameter based intelligent water meter battery capacity dynamic monitoring method according to claim 1, characterized in that, The updated training multi-dimensional operation feature vector is input into the LSTM neural network, and the battery capacity prediction model is generated through the back propagation through time algorithm. The method for generating the battery capacity prediction result comprises the following steps: Based on the temperature data in the multi-dimensional operation feature vector, a corresponding temperature curve is drawn, and the multi-dimensional operation feature vector is updated; The updated multi-dimensional operation feature vector is input into the battery capacity prediction model to generate a first battery capacity prediction result; The time interval between the last prediction and the current prediction is calculated; The battery capacity prediction result is calculated based on the time interval, the self-discharge coefficient and the first battery capacity prediction result.

6. A multi-parameter based intelligent water meter battery capacity dynamic monitoring system characterized in that, The monitoring system is divided into an application layer, a network layer and a perception layer from top to bottom; wherein: The perception layer is configured to collect multi-dimensional operation parameters of the battery of the intelligent water meter by a sensor at a timing; the multi-dimensional operation parameters include battery voltage, working current, ambient temperature, reference signal receiving power and transmitting power; The network layer is configured to transmit the multi-dimensional operation parameters to the application layer; The application layer is configured to manage the intelligent water meter, analyze and monitor the multi-dimensional operation parameter data, and generate and implement monitoring results.

7. A multi-parameter based intelligent water meter battery capacity dynamic monitoring system as claimed in claim 6, wherein, The application layer comprises: a feature generation unit configured to construct a temperature change model, extract features of the multi-dimensional operation parameters by combining a mean filtering algorithm, and generate a multi-dimensional operation feature vector; a self-discharge coefficient calculation unit configured to calculate the self-discharge coefficient based on the Arrhenius equation and a linear regression method; a model construction unit configured to construct a battery capacity prediction model based on an LSTM neural network; a prediction unit configured to predict the multi-dimensional operation feature vector by the battery capacity prediction model based on the self-discharge coefficient, and generate a battery capacity prediction result; a monitoring unit configured to monitor the battery of the intelligent water meter based on the battery capacity prediction result and the battery voltage, and in combination with a three-level early warning mechanism.

8. A multi-parameter based intelligent water meter battery capacity dynamic monitoring system as claimed in claim 6, wherein, The perception layer comprises: a timer configured to set an acquisition interval time; an MCU single-chip microcomputer configured to generate battery acquisition instructions and temperature acquisition instructions, and control data interaction of the network layer by a serial interface to collect reference signal receiving power and transmitting power of the intelligent water meter; a resistance voltage division circuit configured to implement the battery acquisition instructions issued by the MCU single-chip microcomputer to collect the battery voltage of the battery of the intelligent water meter; a temperature sensor configured to implement the temperature acquisition instructions issued by the MCU single-chip microcomputer to collect the ambient temperature of the battery of the intelligent water meter.

9. A multi-parameter based intelligent water meter battery capacity dynamic monitoring system as claimed in claim 8, wherein, The resistance voltage division circuit comprises a MOS tube T1 of model SP3407S; a first port of the T1 is connected to the battery voltage VBAT; a second port of the T1 is connected to one end of a resistor R1, and the other end of the resistor R1 is used as a control signal pin V_Ctr; a third port of the T1 is connected to one end of a resistor R2, and the other end of the resistor R2 is respectively connected to one end of a resistor R4 and a grounding resistor R3; the other end of the resistor R4 is connected to a grounding capacitor C1, and is used as a voltage detection output pin V_Det.

Images