Intelligent water meter control system and method based on Internet of Things

Abstract

The invention discloses an intelligent water meter control system and method based on the Internet of Things, and the method comprises the steps: mining the time sequence correlation characteristics of a real-time water consumption curve and historical synchronous data, building a dynamic comparison frame based on a time queue, and carrying out the deep query matching of the real-time data of an intelligent water meter and the multi-year synchronous historical water consumption time sequence characteristics, and self-adaptive identification of water consumption abnormity is realized based on a query result. By adopting a time queue matching analysis mode, the misjudgment risk caused by the change of the user behavior mode can be effectively reduced, and meanwhile, a complex prediction model does not need to be trained depending on a large amount of historical data, so that the operation of the whole system is more efficient and has sustainability.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to an intelligent water meter control system and method based on the Internet of Things. Background Art

[0002] With the rapid development of Internet of Things technology, smart water meters, as the core terminal equipment of smart water services, have gradually realized the real-time collection and remote transmission of water use data. In the existing technology, anomalies (such as water meter failure or pipeline leakage) are usually judged by comparing the statistical deviation of the total water consumption in the region with the data collected by the smart water meter. For example, by building a user water preference model, combining deep learning algorithms to analyze differences in water use behavior, or using Bayesian networks to predict the probability of water meter failure. However, such methods have significant limitations: on the one hand, the dynamic changes in user behavior patterns (such as seasonal water use fluctuations, new residents) may lead to model misjudgment, and rely on a large amount of historical data to train complex models, increasing the cost of system deployment and maintenance; on the other hand, threshold detection based on statistical deviations is difficult to accurately capture periodic or trend anomalies (such as a slight increase in daily average water consumption caused by slow pipeline leakage), and the response sensitivity to sudden short-term anomalies (such as instantaneous pipe bursts) is insufficient.

[0003] Therefore, an optimized IoT-based smart water meter control method is expected. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a smart water meter control system and method based on the Internet of Things, which mines the time series correlation characteristics of the real-time water usage curve and the historical data of the same period, and establishes a dynamic comparison framework based on the time queue, matches the real-time data of the smart water meter with the deep query of the time series characteristics of the historical water usage in the same period of multiple years, and realizes adaptive identification of water use anomalies based on the query results. By adopting the method of time queue matching analysis, the risk of misjudgment caused by changes in user behavior patterns can be effectively reduced. At the same time, there is no need to rely on a large amount of historical data to train complex prediction models, making the operation of the entire system more efficient and sustainable.

[0005] According to one aspect of the present application, a smart water meter control method based on the Internet of Things is provided, which includes: Obtaining a set of a time queue of water consumption data of a first smart water meter and a time queue of historical water consumption data; Based on the timestamp, a time series matching and a water consumption time series feature query matching analysis are performed between the time series of the water consumption data and the set of time series of the historical water consumption data to determine whether the first smart water meter has an abnormality; In response to an abnormality in the first smart water meter, the first smart water meter is remotely controlled.

[0006] According to another aspect of the present application, there is provided a smart water meter control system based on the Internet of Things, which includes: A data acquisition module, used to acquire a time queue of water consumption data of the first smart water meter and a collection of time queues of historical water consumption data; A water consumption time series feature query matching analysis module, used for performing time series matching and water consumption time series feature query matching analysis between a time queue of water consumption data and a set of time queues of historical water consumption data based on timestamps, to determine whether the first smart water meter has an abnormality; The smart water meter control module is used to remotely control the first smart water meter in response to an abnormality in the first smart water meter.

[0007] Compared with the prior art, the present application provides an IoT-based smart water meter control system and method, which mines the temporal correlation characteristics of real-time water usage curves and historical data of the same period, and establishes a dynamic comparison framework based on time queues, matches the real-time data of smart water meters with the deep query of the temporal characteristics of historical water usage in the same period of multiple years, and realizes adaptive identification of water usage anomalies based on the query results. By adopting the time queue matching analysis method, the risk of misjudgment caused by changes in user behavior patterns can be effectively reduced, and at the same time, there is no need to rely on a large amount of historical data to train complex prediction models, making the operation of the entire system more efficient and sustainable. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 It is a flow chart of a smart water meter control method based on the Internet of Things according to an embodiment of the present application; Figure 2 A schematic diagram of data flow of a smart water meter control method based on the Internet of Things according to an embodiment of the present application; Figure 3 Flow chart of sub-step S2 of the smart water meter control method based on the Internet of Things according to an embodiment of the present application; Figure 4 Flow chart of sub-step S22 of the smart water meter control method based on the Internet of Things according to an embodiment of the present application; Figure 5 It is a block diagram of an intelligent water meter control system based on the Internet of Things according to an embodiment of the present application. DETAILED DESCRIPTION

[0010] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0011] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0012] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0013] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0014] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0015] In the technical solution of the present application, a smart water meter control method based on the Internet of Things is proposed. Figure 1 The present invention is a flowchart of a smart water meter control method based on the Internet of Things according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the smart water meter control method based on the Internet of Things according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the smart water meter control method based on the Internet of Things includes the following steps: S1, obtaining a time queue of water consumption data of a first smart water meter and a set of time queues of historical water consumption data; S2, performing a timing matching and water consumption timing feature query matching analysis between the time queue of water consumption data and the set of time queues of historical water consumption data based on timestamps to determine whether there is an abnormality in the first smart water meter; S3, in response to the existence of an abnormality in the first smart water meter, remotely controlling the first smart water meter.

[0016] In particular, the S1 obtains a time queue of the water consumption data of the first smart water meter and a collection of time queues of historical water consumption data. That is, in an embodiment of the present application, first, the collected water consumption data information is transmitted to the IoT central control center through the communication module embedded in the first smart water meter. It should be understood that in traditional water management, manual meter reading or periodic data upload has delay and discrete defects, which makes it difficult to support the fine-grained analysis required for dynamic anomaly detection. The deployment of the embedded communication module essentially transforms the smart water meter from an isolated data collection node into an active sensing unit in the IoT ecosystem. Through low-power wide-area IoT protocols such as cellular networks (such as NB-IoT) and LoRa, the multi-sensor data such as water meter flow, pressure, valve status, etc. are encrypted and transmitted in a directional manner at preset time intervals or event trigger mechanisms. In addition, the real-time transmission mechanism enables the system to continuously capture micro-fluctuations in water use behavior, such as tiny leakage signals during low-flow periods at night, or abnormal patterns of sudden increases in water use during holidays, thereby breaking through the dependence of traditional threshold detection on fixed time windows. At the same time, the continuous injection of data streams creates conditions for the central control center to establish a dynamic baseline model. By comparing the real-time data stream with the multi-dimensional features of the historical period, the system can automatically identify pattern drift caused by normal variables such as seasonal factors and user increases and decreases, avoiding misjudgments caused by static models.

[0017] Then, in the IoT control center, the water consumption data information is sorted according to the timestamp to obtain the time queue of water consumption data, wherein the time queue of water consumption data has a start time and an end time. It should be understood that in the prior art, water consumption data is often simplified into discrete statistics or cumulative values ​​of fixed time windows (such as hourly water consumption). Although this processing method is convenient for calculation, it breaks the continuity of water consumption behavior on the time axis. For example, the initial stage of pipeline leakage may be manifested as a small flow fluctuation lasting for several hours in the early morning period. If the total amount is only counted in hours, its abnormal characteristics will be diluted by the zero flow data in the normal period; and the continuous time queue generated by timestamp sorting can completely retain the start and end boundaries and dynamic change trajectories of water consumption events, providing a data basis for capturing such progressive anomalies. It is worth mentioning that the start time represents the initial trigger point of a specific water use behavior or abnormal state on the time axis (such as valve opening, sudden increase in flow), and the end time corresponds to its end point (such as flow zero or recovery of baseline value). Unlike traditional fixed time windows (such as hourly division), the start and end times here are not preset manually, but are automatically identified by analyzing the inflection points of flow changes in the data stream. For example, when the flow is detected to rise continuously from zero, it is marked as the start time, and when the flow falls back to a stable baseline and maintains the set duration, it is marked as the end time. This dynamic segmentation strategy can accurately characterize the non-periodic characteristics of water use events in real scenarios. In a specific example of the present application, when the water use data information enters the central control center, the preprocessing module first normalizes the timestamp to eliminate the time drift error caused by network delay and device clock asynchrony. Then the sliding window algorithm is combined with flow gradient detection to dynamically identify the start and end points of the water use event: when the flow change rate exceeds the preset threshold, the start time mark is triggered, and the end time is confirmed after a certain protection period is delayed after the change rate returns to the stable interval. This mechanism can capture the instantaneous flow surge caused by sudden pipe bursts (the time interval between the start and end time is extremely short), and can also identify the long-term low flow caused by chronic leakage (the start and end time span is several days). The generated time queue of water consumption data not only contains the total water consumption, but also implicitly encodes the water consumption intensity change pattern (such as the difference between high-frequency pulse water consumption and smooth continuous flow) through timestamp density, providing a solid basis for subsequent time series anomaly analysis.

[0018] Then, the historical water consumption data information of the first smart water meter is dispatched, and based on the start time and end time, the time queues of historical water consumption data of different years are extracted from the historical water consumption data information to obtain a collection of time queues of historical water consumption data. Since user water consumption behavior often has periodicity (such as peak water consumption during holidays), seasonality (such as a surge in water consumption for irrigation in summer), and long-term evolution characteristics (such as the inter-annual water consumption baseline migration caused by changes in the community population structure), simply relying on recent data or a single historical period data to build an analysis benchmark is prone to misjudgment due to ignoring the dynamic association across time dimensions. By extracting a historical contemporaneous data set with the same start / end time characteristics as the current time queue (such as the irrigation period queue from 3:00 to 5:00 in the morning of June to August in the past three years), the system can establish a dynamically evolving benchmark model, while identifying total anomalies, capturing the morphological distortion of the flow curve (such as the rise in the flow baseline caused by leakage), thereby effectively decoupling normal seasonal fluctuations from persistent abnormal leakage. Through a collection of historical queues aligned in time and space, the system can strip away long-term trend interference (such as the year-on-year increase in water consumption of a family) and focus on detecting abnormal mutations that deviate from historical fluctuation patterns.

[0019] In particular, the S2 determines whether there is an abnormality in the first smart water meter based on the time series matching between the time series of the water consumption data and the set of time series of the historical water consumption data and the water consumption time series feature query matching analysis based on the timestamp. In a specific example of the present application, Figure 3 As shown, the S2 includes: S21, respectively extracting the water consumption time series characteristics of the time queue of water consumption data and the set of time queues of historical water consumption data to obtain the query water consumption time series correlation characteristics and the set of historical water consumption time series correlation characteristics for the same period; S22, performing a dynamic query and analysis of the water consumption time series characteristics on the query water consumption time series correlation characteristics and the set of historical water consumption time series correlation characteristics for the same period to obtain the water consumption time series query dynamic response coding characteristics; S23, determining whether the first smart water meter has an abnormality based on the water consumption time series query dynamic response coding characteristics.

[0020] Specifically, the S21 extracts the water consumption time series features of the time queue of water consumption data and the set of time queues of historical water consumption data respectively to obtain the query water consumption time series correlation features and the set of historical water consumption time series correlation features of the same period. That is, the time queue of water consumption data is input into the water consumption time series feature extractor based on the forward LSTM model to obtain the query water consumption time series correlation feature vector as the query water consumption time series correlation feature. Because traditional threshold detection or static statistical models are difficult to adapt to the non-stationarity of water use behavior, such as seasonal fluctuations, gradual changes in user habits or short-term anomalies caused by sudden events. The forward LSTM can adaptively and selectively memorize long-term historical information and filter noise interference through its unique gating mechanism (input gate, forget gate, output gate) and cell state structure. In the forward time series processing process, the model analyzes the correlation between the current water consumption and the historical state in turn according to the time step, and establishes a global time series representation from the start time to the end time through the continuous updating of the cell state. By mapping the multidimensional time queue into a low-dimensional dense query water consumption time series correlation feature vector, the system realizes the abstract expression of complex water use patterns. Compared with directly using the original time series data for statistical comparison, the feature vector encodes not only the water consumption value at each moment, but also the dynamic laws in the evolution of water consumption behavior (such as implicit features such as periodic fluctuation phase and trend change acceleration). This high-order representation enables the subsequent anomaly detection module to go beyond simple threshold comparison and make more accurate anomaly discrimination from the level of similarity measurement in feature space, especially for new anomaly patterns that have not appeared in historical data, with better generalization capabilities.

[0021] Similarly, the time queues of each historical water consumption data in the set of time queues of historical water consumption data are input into the water consumption time series feature extractor based on the forward LSTM model to obtain a set of historical water consumption time series correlation feature vectors as a set of historical water consumption time series correlation features. Because traditional methods are difficult to deal with the time-varying characteristics of water use behavior through static thresholds or fixed period pattern matching, for example, the peak water use in summer in the same community in different years may produce time series morphological shifts due to changes in population structure or climate anomalies. The forward LSTM model abstracts historical time series data layer by layer through its gated loop structure. Its input gate uses the sigmoid function to filter the importance weight of historical information, the forgetting gate dynamically adjusts the retention ratio of old memories in the cell state, and the output gate determines the contribution of the current state to the feature expression. This mechanism enables the model to automatically identify common patterns in water use curves in different years (such as periodic fluctuations in the morning and evening peaks on weekdays) when processing cross-year historical data, while suppressing non-critical noise (such as temporary water outage events on a specific day of a year), thereby constructing a benchmark feature space with temporal robustness. During this process, the forward LSTM accumulates temporal context information forward along the time axis when processing each historical time queue. The continuously updated memory units in its cell state not only record the absolute value changes in water consumption, but also capture the differential characteristics of the evolution of water consumption behavior (such as flow rate change rate, phase offset of periodic fluctuations, etc.). This deep feature expression allows the potential correlation between historical data to be explicitly encoded. For example, the difference in the impact of pipeline aging stages on leakage rates in different years will be mapped to a specific distribution pattern in the feature vector space. Compared with direct comparison of raw data, this high-order feature abstraction significantly improves the system's ability to identify long-term trend anomalies.

[0022] Specifically, in S22, the query water consumption time series correlation feature and the historical water consumption time series correlation feature set are subjected to dynamic query analysis of water consumption time series features to obtain dynamic response coding features of water consumption time series query. It should be understood that it is difficult to effectively capture the multi-scale spatiotemporal correlation patterns implicit in water use behavior through simple feature comparison or threshold judgment, such as the dynamic deviation relationship between the slight increase in daily average water consumption caused by slow leakage and the historical data of the same period, or the nonlinear difference between the local mutation pattern formed by the sudden pipe burst event in the feature space and the historical distribution boundary. Therefore, in order to overcome the above problems, the global topological association between water consumption features is captured to realize the essential characterization of abnormal patterns. In the technical solution of the present application, the query water consumption time series correlation feature and the historical water consumption time series correlation feature set are subjected to dynamic query analysis of water consumption time series features to obtain dynamic response coding features of water consumption time series query. That is, by constructing an implicit graph structure, the interaction between the query feature vector and the historical feature set is mapped into the topological space, and the spectral graph theory is used to reveal the inherent manifold structure characteristics of the data. This processing mechanism can overcome the limitations of traditional methods in feature similarity measurement. Specifically, first, the query water consumption time series correlation features are used as "dynamic decision anchor points", and the historical feature set constitutes a "space-time reference system". The two are mapped in hidden space through a nonlinear decision response unit. This unit encodes the implicit interaction between features (such as periodic resonance, trend deviation or mutation association between current features and historical features) into a decision point state implicit coding vector through nonlinear mapping of deep neural networks; then, the spectral decomposition of the Laplace matrix further extracts low-frequency components that reflect the essential structure of the data; then, by adaptively fusing these spectral domain feature components, the system can generate a dynamic response coding vector for water consumption time series queries that contains both local details and global structures. It not only encodes the degree of deviation of the current water use pattern from the historical benchmark, but also reveals the potential pattern of abnormal types (such as the differentiated representation of leakage and burst pipes in the spectral domain), thereby providing a multi-dimensional decision-making basis for subsequent abnormal judgment. In this way, the system's spatiotemporal perception of complex water use anomalies is significantly improved. In the feature expression dimension, by constructing the decision point state neighborhood matrix and Laplace matrix, the system converts discrete data points in the high-dimensional time series feature space into nodes on a continuous manifold structure, so that the "water use pattern similarity" that was originally difficult to quantify is converted into the connection strength in the graph structure and the characteristic vector distance in the spectral domain. For example, short-term fluctuations in historical data due to climate anomalies will be identified as local structural disturbances by the neighborhood matrix, while the spectral decomposition of the Laplace matrix filters such noise through low-dimensional embedding, retaining the true long-term trend information. In terms of computational efficiency, the spectral domain dimensionality reduction technology converts high-dimensional feature comparisons into vector operations in low-dimensional space, allowing the system to complete pattern matching of massive historical data within milliseconds.More importantly, the adaptive fusion mechanism dynamically adjusts the contribution of different spectral components through attention weights, enabling the system to automatically select the most relevant feature combination for a specific anomaly type (such as periodic leakage or random pipe burst), thereby reducing the false alarm rate while improving the generalization ability of new anomaly patterns. In particular, in a specific example of the present application, such as. Figure 4 As shown, the S22 includes: S221, inputting the query water consumption time series correlation feature vector and each historical water consumption time series correlation feature in the set of historical water consumption time series correlation feature vectors into the historical-query water consumption nonlinear decision response unit to obtain a set of historical-query water consumption decision point state implicit coding vectors; S222, performing spectrum-guided historical-query water consumption decision point adaptive aggregation analysis on the set of historical-query water consumption decision point state implicit coding vectors to obtain a water consumption time series query dynamic response coding vector as a water consumption time series query dynamic response coding feature.

[0023] More specifically, in S221, each historical water consumption time series correlation feature in the set of query water consumption time series correlation feature vectors and historical water consumption time series correlation feature vectors in the same period is input into the historical-query water consumption nonlinear decision response unit to obtain a set of implicit coding vectors of the state of historical-query water consumption decision points. Since the user's water consumption behavior is time-varying and uncertain (such as seasonal fluctuations, sudden events), simple statistical thresholds or linear models are difficult to distinguish the subtle differences between normal pattern evolution and real anomalies (such as slow leakage or instantaneous pipe burst). By introducing a nonlinear decision response unit, the system can reconstruct the potential decision relationship between the current water consumption characteristics and the historical pattern in a high-dimensional latent space, and implicitly decouple the water consumption behavior features that were originally intertwined in the original feature space, thereby providing a clearer decision basis for subsequent anomaly detection. Through nonlinear mapping and feature interaction, the system not only considers the temporal similarity between the current water consumption sequence and historical data, but also deeply explores the association rules between the two in implicit dimensions such as state transition and mutation response. For example, in a slow pipeline leakage scenario, the system can capture the nonlinear amplification effect of the small trend offset of the current water consumption relative to the historical data of the same period in the latent space, while the traditional linear weighted method may ignore such progressive anomalies due to noise interference. This implicit state encoding enables the system to break through the reliance of traditional statistical models on explicit thresholds and instead achieve more refined anomaly discrimination through decision boundary optimization in the latent space. In a specific example of the present application, each historical water consumption time series correlation feature in the set of query water consumption time series correlation feature vectors and historical water consumption time series correlation feature vectors for the same period is input into the historical-query water consumption nonlinear decision response unit according to the following formula to obtain a set of implicit coding vectors of the historical-query water consumption decision point states; wherein, the formula is:

[0024] in, is the set of time series associated feature vectors of water consumption in the same historical period, , , and are the first, second, and third time series associated feature vectors of the water consumption in the same period of history. and The time series correlation feature vector of water consumption in the same historical period, is the time series correlation feature vector of the query water consumption, is matrix multiplication, and They are The corresponding decision response weight matrix and decision response bias vector, is the activation function, , , , and are the first, second, and third implicit coding vectors of the history-query water consumption decision point state. , and The implicit coding vector of the state of the history-query water consumption decision point, is the set of implicit coding vectors of the historical-query water consumption decision point states.

[0025] More specifically, the S222 performs a spectrum-guided adaptive aggregation analysis of the history-query water consumption decision point state implicit coding vectors on the set of history-query water consumption decision point states to obtain a water consumption time series query dynamic response coding vector as a water consumption time series query dynamic response coding feature. In an embodiment of the present application, first, a graph structure analysis based on the history-query water consumption implicit state is performed on the set of history-query water consumption decision point state implicit coding features to obtain a history-query water consumption decision point state class neighborhood matrix and a history-query water consumption decision point state class degree matrix. It should be understood that when the system generates a set of implicit coding vectors through a nonlinear decision response unit, although these high-dimensional feature vectors contain the dynamic decision state of water use behavior, the implicit information such as the strength of association between nodes and the local manifold structure is still not explicitly modeled. By constructing a graph structure analysis framework, the system can convert abstract implicit coding vectors into a data topology network with clear geometric meaning, thereby providing a structured decision basis for subsequent spectral analysis. Among them, the construction of the neighborhood matrix is ​​essentially to convert the semantic correlation of decision points in the high-dimensional feature space into the edge weights of the graph structure. For example, when judging the abnormal water consumption in a certain period, the system can automatically identify the historical data nodes of the same period that are closest to the implicit state of the period through the neighborhood matrix (such as the water consumption pattern under similar temperature conditions in the same period last year), while excluding interference nodes caused by incidental factors such as holiday activities. The introduction of the degree matrix gives different nodes the ability to adjust weights differently. For example, when a decision point has a strong correlation with most historical nodes (such as a stable water consumption pattern caused by winter heating), the diagonal value of its degree matrix will be significantly higher than that of isolated abnormal points (such as sudden pipe burst events). This adaptive node importance quantification mechanism can effectively suppress the interference of noise nodes on the overall structural analysis, ensuring that the graph model more accurately reflects the distribution characteristics of real data. This graph structure analysis method significantly improves the adaptability and interpretability of the smart water meter system to complex water use scenarios. Through the synergy of the neighborhood matrix and the degree matrix, the system can not only capture the microscopic correlations between implicit states (such as identifying the hidden correlations between water consumption at multiple adjacent time nodes in the early stage of slow leakage), but also construct a dynamic topological map of the evolution of water use behavior at a macro level, so that the subsequent Laplace matrix spectral decomposition can more effectively extract discriminative core components, thereby accurately distinguishing normal pattern evolution from real abnormal events in real scenarios where complex variables such as seasonal fluctuations and infrastructure upgrades coexist, and providing reliable technical support for the refined management and predictive maintenance of smart water services. In a specific example of the present application, the set of implicit coding features of the historical-query water consumption decision point state is analyzed based on the graph structure of the historical-query water consumption implicit state using the following formula to obtain the historical-query water consumption decision point state class neighborhood matrix and the historical-query water consumption decision point state class degree matrix; wherein, the formula is:

[0026]

[0027] in, To calculate the second norm of a vector, is the inverse hyperbolic cosine function, , , , and are the eigenvalues ​​of each position in the state class neighborhood matrix of the history-query water consumption decision point, is the state class neighborhood matrix of the history-query water consumption decision point, To calculate the square of the vector-norm, yes The number of vectors in the equation is reduced by one. , and are the eigenvalues ​​of the diagonal positions in the state class matrix of the history-query water consumption decision point, It is the state class degree matrix of the history-query water consumption decision point.

[0028] Next, the state Laplace matrix of the historical-query water consumption decision point is calculated based on the state class neighborhood matrix of the historical-query water consumption decision point and the state class degree matrix of the historical-query water consumption decision point. It should be understood that although the neighborhood matrix can reflect the direct correlation strength between nodes (such as the similarity between the water consumption pattern of a certain period and the same period in history), and the degree matrix can characterize the connection density of nodes (such as the weak correlation between an abnormal decision point and most historical normal nodes), the two have not yet formed an overall description of the dynamic equilibrium state of the graph structure. By constructing the state Laplace matrix of the historical-query water consumption decision point, the system algebraically fuses the local connection relationship encoded in the neighborhood matrix with the global importance of the node quantified by the degree matrix. In this process, the Laplace matrix uniformly encodes the continuity (such as the annual accumulation of leakage trends) and the suddenness (such as the isolation of pipe burst events) in the evolution of water use behavior into the frequency domain components of the graph signal through the local connection strength of the neighborhood matrix and the node centrality parameters of the degree matrix. For example, slow leakage across years caused by pipeline aging forms a strongly connected node chain in the graph structure, and its corresponding Laplace matrix eigenvector shows smooth fluctuations in the low-frequency band, reflecting the continuity of the trend; while a sudden drop in water use caused by a temporary maintenance is manifested as a spike pulse in the high-frequency band. This frequency domain mapping mechanism enables the system to distinguish between natural fluctuations (such as the periodic spectrum corresponding to seasonal water use changes) and real anomalies (such as irregular high-frequency disturbances), providing a feature basis with clear physical meaning for subsequent spectral decomposition. In this way, the system can break through the limitations of traditional time domain analysis when facing multi-source heterogeneous water use scenarios, and provide a multi-dimensional decision-making basis for the precise anomaly diagnosis of smart water services. In a specific example of the present application, based on the historical-query water consumption decision point state class neighborhood matrix and the historical-query water consumption decision point state class degree matrix, the historical-query water consumption decision point state Laplace matrix is ​​calculated using the following formula; wherein, the formula is:

[0029] in, is the history-query water consumption decision point state Laplace matrix.

[0030] Then, the historical-query water consumption decision point state Laplace matrix is ​​spectrally decomposed to obtain a set of core component encoding vectors of the historical-query water consumption decision point. In an embodiment of the present application, first, the historical-query water consumption decision point state Laplace matrix is ​​optimized based on the cut-cycle space closure of the historical-query water consumption decision point state matrix using a topological closure mechanism to obtain an optimized historical-query water consumption decision point state Laplace matrix. It should be understood that when water use behavior distortion occurs due to pipe network leakage or water meter failure, its abnormal signal is often not isolated in the timing fluctuation of a single node, but forms an implicit association pattern across nodes through the physical connection of the water supply network and the data topology. For example, a pipeline rupture may cause the water consumption timing of three adjacent water meters to resonate in a specific frequency band. The traditional Laplace matrix can only characterize the direct connection strength between nodes, but cannot capture such non-explicit topological associations formed by fluid dynamics conduction. At this time, the spectral space of the original historical-query water consumption decision point state Laplace matrix is ​​susceptible to local noise interference, and its cut space component may mistakenly classify the real anomaly as seasonal fluctuations, while the cyclic space component may ignore the propagation path of the emergency. In the preferred example of the present application, a topological closure mechanism is used to optimize the historical-query water consumption decision point state Laplace matrix based on the cut-circular space closure to obtain the optimized historical-query water consumption decision point state Laplace matrix. That is, the cut space and cyclic space components of the historical-query water consumption decision point state Laplace matrix are separated by pseudo-inverse operation, which is essentially to decouple the static physical connection and dynamic data association of the water supply network at the algebraic topology level. The cut space closure strengthens the global correlation across time slices through exponential mapping (such as the cumulative effect of the slight increase in daily average water consumption caused by leakage in multi-year data), while the cyclic space closure constrains the propagation range of local disturbances through tensor product operations (such as the high-frequency oscillation of the burst pipe event only affects the adjacent nodes). For example, slow leakage is represented as a low-dimensional manifold in the cut space in the optimized Laplace matrix, and its spectral characteristics correspond to low-frequency energy concentration; while temporary water outage events form isolated high-frequency pulses in the cyclic space, which are significantly different from the real abnormal patterns. This optimization maps the spectral characteristics of the matrix to the physical evolution law of the water system, providing physical semantic support at the frequency domain level for abnormal diagnosis. This mechanism gives the system the ability to generalize and analyze unknown abnormalities: when a new type of fault occurs (such as progressive wear of the gears of smart water meters), its slowly developing characteristics will be mapped to the cut space, triggering a disposal strategy similar to that of known leakage; while sudden unknown faults (such as transient data anomalies caused by electromagnetic interference) fall into the high-frequency domain of the cyclic space, triggering immediate alarms. This frequency domain feature separation based on physical laws enables the system to achieve adaptive classification of diverse abnormal patterns without retraining the model, supporting the precise decision-making and rapid response of smart water systems in dynamic environments.

[0031] Preferably, the specific steps of using the topological closure mechanism to optimize the state Laplace matrix of the historical-query water consumption decision point based on the cut-cycle space closure are as follows: Specifically, firstly, a pseudo-inverse operation is performed to extract the connected component information of the state class neighborhood matrix of the history-query water consumption decision point and the state class degree matrix of the history-query water consumption decision point, that is, let:

[0032] Among them, the matrix and The historical-query water consumption decision point state Laplace matrices can be respectively represented as The cut space and cycle space of Represents matrix multiplication; Thus, the state Laplace matrix of the history-query water consumption decision point is obtained by cutting space closure and loop space closure. Optimize, expressed as:

[0033] in, The natural constant The exponential function value with base , It is added by location point. is the optimized history-query water consumption decision point state Laplace matrix.

[0034] In this way, the global structural invariant in the structural information of the graph is extracted by a closed operation based on pseudo-inverse, so that the state Laplace matrix of the history-query water consumption decision point is The graph topology information representation is more robust.

[0035] Furthermore, the optimized historical-query water consumption decision point state Laplace matrix is ​​spectrally decomposed to obtain a set of core component encoding vectors of the historical-query water consumption decision points. It should be understood that although the Laplace matrix optimized by the cut-loop space closure has eliminated the topological holes caused by short-term disturbances, its essence is still a complex operator that characterizes node associations in a high-dimensional space. For example, the synergistic water consumption decline caused by a cross-regional pipeline leak is manifested in the optimization matrix as an enhancement of weak connections between multiple non-adjacent nodes, but this association pattern may be masked by the strong connection characteristics of adjacent nodes in the original feature space. Therefore, in the technical solution of the present application, the optimized historical-query water consumption decision point state Laplace matrix is ​​spectrally decomposed to obtain a set of core component encoding vectors of the historical-query water consumption decision points. Here, through the spectral decomposition operation, the system decomposes the optimized Laplace matrix into a set of eigenvalues ​​and eigenvectors, where the eigenvector corresponding to the smallest eigenvalue carries the low-frequency global trend of water use behavior (such as slow leakage across years caused by pipeline aging), while the larger eigenvalue maps high-frequency local events (such as pulse-like disturbances caused by instantaneous pipe bursts). The essence of this mathematical operation is to project high-dimensional time series data into a frequency domain space with clear physical meaning. For example, long-term leakage is manifested as low-frequency energy concentration, while emergencies are manifested as a sudden increase in high-frequency energy, thereby breaking through the limitations of traditional methods in single-dimensional analysis in the time domain. In a specific example of the present application, the optimized historical-query water consumption decision point state Laplace matrix is ​​spectrally decomposed using the following formula to obtain a set of core component encoding vectors of the historical-query water consumption decision point; wherein, the formula is:

[0036] in, For Perform spectral decomposition operation, is the set of core component encoding vectors of the history-query water consumption decision points, , , and are the first, second, and third core component encoding vectors of the history-query water consumption decision point. and The core component encoding vector of the history-query water consumption decision point, The diagonal elements of the matrix are , , and History - query water consumption eigenvalue diagonal matrix, , , and They are , , and The corresponding eigenvalues ​​are is the diagonal matrix of historical-query water consumption eigenvalues.

[0037] Furthermore, the set of core component encoding vectors of historical-query water consumption decision points is adaptively fused to obtain the dynamic response encoding vector of water consumption time series query. It should be understood that the set of core component encoding vectors of historical-query water consumption decision points is a characteristic basis vector decoupled in the spectral domain. Although the abnormal signal components of different frequency bands have been separated, heterogeneous abnormalities such as pipe network leakage and equipment failure in actual scenarios often show cross-band coupling characteristics. For example, the low-frequency trend drift caused by slow leakage is often accompanied by medium-frequency micro-oscillations derived from pressure disturbances, and the failure of the control valve may simultaneously trigger the steady-state offset of high-frequency pulses and specific spatial encoding vectors. If a static fusion strategy is adopted, it is impossible to dynamically reconstruct the combination pattern of abnormal features according to the real-time monitoring data, and it is easy to weaken the perception ability of nonlinear associations between time-varying features due to fixed weight allocation. Through the adaptive fusion mechanism, the system can dynamically adjust the contribution of each core component according to the context of the current water use pattern, such as using the attention mechanism to learn the weight allocation strategy of different frequency band features. In this process, by continuously learning the implicit mapping relationship between historical core components and real-time query input, a context-driven feature reorganization mechanism is formed. When the abnormal pattern evolves over time (such as the increase in the amplitude of the low-frequency offset accompanied by the change in the frequency of pressure fluctuations during the leakage development stage), the fusion network can dynamically track the contribution migration trajectory of each component and synchronously adjust the feature synthesis direction. This dynamic synthesis mechanism enables the final generated response coding vector to have the ability of spatiotemporal generalization, which can not only characterize the stable topological characteristics of the pipe network structure, but also capture the transient distortion characteristics of water use behavior. In a specific example of the present application, the set of core component coding vectors of the historical-query water consumption decision point is adaptively fused using the following formula to obtain a dynamic response coding vector for water consumption time series query; wherein, the formula is:

[0038] in, For Perform adaptive fusion operations. and They are The corresponding fusion weight matrix and fusion bias vector, yes function, yes The corresponding history-query water consumption score weight vector, yes Corresponding history-query water consumption weight value, is a masking operation, is the preset threshold, yes The corresponding history-query water consumption mask weight value, is the dynamic response encoding vector of the water consumption time series query.

[0039] Specifically, the S23 determines whether the first smart water meter is abnormal based on the dynamic response coding characteristics of the water consumption time series query. That is, in the technical solution of the present application, the dynamic response coding vector of the water consumption time series query is input into the water consumption data diagnosis module based on the classifier to obtain the diagnosis result, and the diagnosis result is used to indicate whether the first smart water meter is abnormal. It should be understood that the dynamic response coding vector of the water consumption time series query is essentially an implicit state fusion of historical and current water use behaviors, but the abnormal discrimination information contained therein needs to be separated by a nonlinear decision boundary. However, due to the high dimensionality, nonlinearity and pattern confusion of the time series characteristics of water use behavior (such as the overlap of normal fluctuations and abnormal leakage characteristics), it is difficult to effectively distinguish abnormal states in complex scenarios by simply relying on artificial rules or static thresholds. By introducing a classifier, the system can map the dynamic response coding vector of the water consumption time series query after dynamic query analysis to an interpretable decision space, breaking through the shallow mining ability of the traditional method for feature correlation. In this way, the reliability and scene adaptability of abnormal diagnosis are significantly improved.

[0040] In particular, the S3, in response to an abnormality in the first smart water meter, remotely controls the first smart water meter. In an embodiment of the present application, first, in response to an abnormality in the first smart water meter, a smart water meter control instruction is generated. That is, the water consumption data diagnosis module based on the classifier will analyze the dynamic response encoding vector of the water consumption time series query. Once it is determined that the first smart water meter has an abnormality, such as it is found that the water consumption pattern deviates from the normal range, which may be due to potential leakage, equipment failure, etc., the system will automatically generate an early warning signal as a control instruction, wherein the early warning signal contains detailed information about the abnormal situation, such as the specific type of abnormality (such as leakage or illegal use), severity, and possible scope of impact.

[0041] Subsequently, the IoT control center sends the smart water meter control instruction to the remote control terminal of the first smart water meter to realize remote control of the first smart water meter through the remote control terminal. That is, once the abnormality of the smart water meter is confirmed, the corresponding control instruction will be quickly sent to the remote control terminal of the smart water meter to realize real-time control of the first smart water meter. Rapid intervention through remote control means ensures that the smart water meter and related water supply systems can resume normal working conditions as soon as possible. This not only helps to protect water resources from unnecessary waste, but also ensures service quality and avoids inconvenience to users. In addition, by handling abnormal situations in a timely manner, the service life of the equipment can be extended and maintenance costs can be reduced. After receiving the early warning signal, each recipient can take corresponding measures according to their own responsibilities. For example, the maintenance team may arrange technicians to go to the site for inspection and repair; the property management may contact the residents to remind them to pay attention to the recent water use; and the user can view the specific early warning details through the application to understand the abnormal conditions of their own water meters. In this way, it is ensured that any potential problems can be handled in a timely and effective manner, minimize losses and ensure the normal operation of the water supply system.

[0042] In summary, the IoT-based smart water meter control method according to the embodiment of the present application is explained, which mines the time series correlation characteristics of the real-time water usage curve and the historical data of the same period, and establishes a dynamic comparison framework based on the time queue, matches the real-time data of the smart water meter with the deep query of the time series characteristics of the historical water usage of the same period in multiple years, and realizes the adaptive identification of water usage anomalies based on the query results. By adopting the time queue matching analysis method, the risk of misjudgment caused by changes in user behavior patterns can be effectively reduced, and at the same time, there is no need to rely on a large amount of historical data to train complex prediction models, making the operation of the entire system more efficient and sustainable.

[0043] Furthermore, an intelligent water meter control system based on the Internet of Things is also provided.

[0044] Figure 5 FIG. 1 is a block diagram of an intelligent water meter control system based on the Internet of Things according to an embodiment of the present application. Figure 5 As shown, according to the IoT-based smart water meter control system 300 of the embodiment of the present application, it includes: a data acquisition module 310, which is used to acquire the time queue of water consumption data of the first smart water meter and the set of time queues of historical water consumption data; a water consumption time series feature query matching analysis module 320, which is used to determine whether there is an abnormality in the first smart water meter based on the water consumption time series feature query matching analysis between the time queue of water consumption data and the set of time queues of historical water consumption data; a smart water meter control module 330, which is used to remotely control the first smart water meter in response to the existence of an abnormality in the first smart water meter.

[0045] As described above, the smart water meter control system 300 based on the Internet of Things according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a smart water meter control algorithm based on the Internet of Things. In a possible implementation, the smart water meter control system 300 based on the Internet of Things according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the smart water meter control system 300 based on the Internet of Things can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the smart water meter control system 300 based on the Internet of Things can also be one of the many hardware modules of the wireless terminal.

[0046] Alternatively, in another example, the IoT-based smart water meter control system 300 and the wireless terminal may also be separate devices, and the IoT-based smart water meter control system 300 may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0047] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A smart water meter control method based on the Internet of Things, characterized in that: include: Obtaining a set of a time queue of water consumption data of a first smart water meter and a time queue of historical water consumption data; Based on the timestamp, a time series matching and a water consumption time series feature query matching analysis are performed between the time series of the water consumption data and the set of time series of the historical water consumption data to determine whether the first smart water meter has an abnormality; In response to an abnormality in the first smart water meter, the first smart water meter is remotely controlled.

2. The method for controlling a smart water meter based on the Internet of Things according to claim 1, characterized in that: Get the time queue of water consumption data and the collection of time queues of historical water consumption data, including: The collected water usage data information is transmitted to the IoT central control center through the communication module embedded in the first smart water meter; In the IoT control center, the water consumption data information is sorted according to the timestamp to obtain a time queue of water consumption data, wherein the time queue of water consumption data has a start time and an end time; The historical water consumption data information of the first smart water meter is scheduled, and based on the start time and the end time, the time queues of the historical water consumption data of different years are extracted from the historical water consumption data information to obtain a set of time queues of the historical water consumption data.

3. The method for controlling a smart water meter based on the Internet of Things according to claim 1, characterized in that: Based on the timestamp, a time series matching and a water consumption time series feature query matching analysis between a time series of water consumption data and a collection of time series of historical water consumption data are performed to determine whether the first smart water meter is abnormal, including: Extracting the water consumption time series features of the time queue of water consumption data and the set of time queues of historical water consumption data respectively to obtain a set of query water consumption time series correlation features and historical water consumption time series correlation features for the same period; Perform dynamic query analysis on the water consumption time series characteristics of the query water consumption time series and the historical water consumption time series correlation characteristics of the same period to obtain the dynamic response coding characteristics of the water consumption time series query; Based on the water consumption time series query dynamic response coding characteristics, determine whether the first smart water meter has an abnormality.

4. The method for controlling a smart water meter based on the Internet of Things according to claim 3 is characterized in that: The water consumption time series features of the time queue of water consumption data and the set of time queues of historical water consumption data are extracted respectively to obtain a set of query water consumption time series correlation features and historical water consumption time series correlation features for the same period, including: The time queue of water consumption data is input into the water consumption time series feature extractor based on the forward LSTM model to obtain the query water consumption time series correlation feature vector as the query water consumption time series correlation feature; Each time queue of historical water consumption data in the set of time queues of historical water consumption data is input into the water consumption time series feature extractor based on the forward LSTM model to obtain a set of historical water consumption time series correlation feature vectors for the same period as a set of historical water consumption time series correlation features.

5. The method for controlling a smart water meter based on the Internet of Things according to claim 4, characterized in that: The water consumption time series feature dynamic query analysis is performed on the set of query water consumption time series correlation features and historical water consumption time series correlation features in the same period to obtain the water consumption time series query dynamic response coding features, including: Input each historical water consumption time series correlation feature in the set of query water consumption time series correlation feature vector and historical water consumption time series correlation feature vector into the historical-query water consumption nonlinear decision response unit to obtain a set of historical-query water consumption decision point state implicit coding vectors; The set of historical-query water consumption decision point state implicit coding vectors is subjected to spectral-guided historical-query water consumption decision point adaptive aggregation analysis to obtain the water consumption time series query dynamic response coding vector as the water consumption time series query dynamic response coding feature.

6. The method for controlling a smart water meter based on the Internet of Things according to claim 5, characterized in that: The set of implicit coding vectors of the historical-query water consumption decision point states is subjected to a spectrum-guided adaptive aggregation analysis of the historical-query water consumption decision points to obtain a dynamic response coding vector for water consumption time series query, including: The set of implicit coding features of the historical-query water consumption decision point state is analyzed based on the graph structure of the historical-query water consumption implicit state to obtain the historical-query water consumption decision point state class neighborhood matrix and the historical-query water consumption decision point state class degree matrix; Based on the state class neighborhood matrix of the history-query water consumption decision point and the state class degree matrix of the history-query water consumption decision point, the state Laplace matrix of the history-query water consumption decision point is calculated; Performing spectral decomposition on the state Laplace matrix of the historical-query water consumption decision point to obtain a set of core component encoding vectors of the historical-query water consumption decision point; The set of core component encoding vectors of historical-query water consumption decision points is adaptively fused to obtain the dynamic response encoding vector of water consumption time series query.

7. The method for controlling a smart water meter based on the Internet of Things according to claim 6, characterized in that: The state Laplace matrix of the historical-query water consumption decision point is spectrally decomposed to obtain a set of core component encoding vectors of the historical-query water consumption decision point, including: The topological closure mechanism is used to optimize the state Laplace matrix of the historical-query water consumption decision point based on the cut-circular space closure to obtain the optimized state Laplace matrix of the historical-query water consumption decision point; The optimized historical-query water consumption decision point state Laplace matrix is ​​spectrally decomposed to obtain a set of core component encoding vectors of the historical-query water consumption decision point.

8. The method for controlling a smart water meter based on the Internet of Things according to claim 7, characterized in that: Determining whether the first smart water meter is abnormal based on the dynamic response coding characteristics of the water consumption time series query includes: The water consumption time series query dynamic response encoding vector is input into the classifier-based water consumption data diagnosis module to obtain a diagnosis result, and the diagnosis result is used to indicate whether there is an abnormality in the first smart water meter.

9. The method for controlling a smart water meter based on the Internet of Things according to claim 1, characterized in that: In response to an abnormality in the first smart water meter, remotely controlling the first smart water meter includes: In response to an abnormality in the first smart water meter, generating a smart water meter control instruction; The Internet of Things central control center sends the smart water meter control instruction to the remote control terminal of the first smart water meter to realize remote control of the first smart water meter through the remote control terminal.

10. An intelligent water meter control system based on the Internet of Things, characterized in that: include: A data acquisition module, used to acquire a time queue of water consumption data of the first smart water meter and a collection of time queues of historical water consumption data; A water consumption time series feature query matching analysis module is used to perform time series matching and water consumption time series feature query matching analysis between a time queue of water consumption data and a set of time queues of historical water consumption data based on timestamps to determine whether the first smart water meter has an abnormality; The smart water meter control module is used to remotely control the first smart water meter in response to an abnormality in the first smart water meter.

Images