Intelligent Water Meter Measurement Data Abnormality Monitoring System and Method

Abstract

This application relates to the field of intelligent monitoring technology. Specifically, it discloses an intelligent water meter measurement data abnormal monitoring system and method. It uses the built-in communication module of the intelligent water meter to send the intelligent water meter measurement data to the remote monitoring center. At the same time, a data processing algorithm is deployed in the remote monitoring center to divide the intelligent water meter measurement data into continuous time series segments at a predetermined time step, and uses the measurement data segment at the end as the verification benchmark. By performing two-level stacked time series encoding on the intelligent water meter measurement data segments in the historical period, the deep time series characteristics of the intelligent water meter measurement data are captured, and the water meter measurement data inference at the current time step is carried out accordingly. Furthermore, based on the comparison between the inference result and the verification benchmark, the abnormal judgment of the measurement data is realized. This method can effectively reduce the dependence on preset rules, dynamically adapt to diverse water usage scenarios, and reduce the risks of false alarms and missed detections.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring technologies, and more specifically, to an intelligent water meter measurement data abnormal monitoring system and method. Background Art

[0002] In modern intelligent water supply systems, intelligent water meters, as key terminal devices for water supply data collection, the accuracy of their measurement data is crucial for ensuring the efficient operation of the water supply system, reasonable billing, and timely detection of problems such as pipe network leakage. In this regard, the invention patent with the publication number CN118211160A proposes a method for monitoring abnormal water meter measurement data. It disassembles the original water meter measurement data into fine-grained units, and by mining its stage measurement state element vectors (short-term features) and periodic measurement state element vectors (long-term features), it comprehensively characterizes water usage behaviors. At the same time, it introduces knowledge embedding technology, transforms preset abnormal monitoring rules (such as time period thresholds, seasonal water usage patterns) into computable embedding vectors, and establishes a deep association between the data pattern and the rules. Finally, it generates a comprehensive decision vector by fusing the above short-term stage features, long-term periodic features, and rule features to quantitatively evaluate the deviation degree of the water usage state and identify data anomalies.

[0003] In the prior art, abnormal monitoring rule features refer to a series of predefined rules or criteria based on which data abnormal monitoring is carried out, describing the patterns or behaviors that data should follow under normal circumstances. When the actual data does not conform to the rule features, it may be identified as abnormal. However, the abnormal monitoring rules are constructed based on historical data and expert knowledge and are used to describe and identify common abnormal behavior patterns in water meter measurement data. The fixed nature of such preset rules limits their ability to identify new abnormal behaviors and is difficult to cope with complex and changeable water usage scenarios. For example, sudden water usage events, introduction of new water usage equipment, or regional water usage behavior changes may cause the preset rules to fail, resulting in false alarms or missed detections.

[0004] Therefore, an optimized intelligent water meter measurement data abnormal monitoring system and method are expected. Summary of the Invention

[0005] To solve the above technical problems, the present application is proposed. An embodiment of the present application provides an intelligent water meter measurement data anomaly monitoring system and method, which uses the built-in communication module of the intelligent water meter to send the intelligent water meter measurement data to the remote monitoring center. At the same time, a data processing algorithm is deployed in the remote monitoring center to divide the intelligent water meter measurement data into continuous time series segments at a predetermined time step, and use the measurement data segment at the end as the verification benchmark. By performing secondary stacked time series coding on the intelligent water meter measurement data segments in the historical period, that is, extracting the short-term time series fluctuation characteristics inside the segments and mining the long-term time series correlation characteristics between the segments, the deep time series characteristics of the intelligent water meter measurement data can be captured, and based on this, the water meter measurement data inference at the current time step is performed. Furthermore, based on the comparison between the inference result and the verification benchmark, the anomaly judgment of the measurement data is realized. This method can effectively reduce the dependence on preset rules, dynamically adapt to diverse water usage scenarios, and reduce the risks of false alarms and missed detections.

[0006] According to one aspect of the present application, there is provided an intelligent water meter measurement data anomaly monitoring method, which includes:

[0007] Using the built-in communication module of the intelligent water meter to send the intelligent water meter measurement data to the remote monitoring center;

[0008] At the remote monitoring center, perform data segmentation on the intelligent water meter measurement data based on a fixed time step to obtain the sequence distribution of the intelligent water meter measurement time step data;

[0009] Extract the intelligent water meter measurement time step data of the last time step from the sequence distribution of the intelligent water meter measurement time step data as the intelligent water meter measurement self-supervised verification data;

[0010] Perform measurement data time series secondary stacked coding on the intelligent water meter measurement time step data in the sequence distribution except the intelligent water meter measurement self-supervised verification data to obtain the intelligent water meter measurement time series context coding vector;

[0011] Perform intelligent water meter measurement inference based on the intelligent water meter measurement time series context coding vector to obtain the intelligent water meter measurement inference time step data;

[0012] Based on the comparison between the intelligent water meter measurement inference time step data and the intelligent water meter measurement self-supervised verification data, determine whether there is data anomaly.

[0013] According to another aspect of the present application, there is provided an intelligent water meter measurement data anomaly monitoring system, which includes:

[0014] A water meter measurement data transmission module for using the built-in communication module of the intelligent water meter to send the intelligent water meter measurement data to the remote monitoring center;

[0015] The water meter measurement data segmentation module is used to segment the intelligent water meter measurement data at a fixed time step in the remote monitoring center to obtain the sequence distribution of the intelligent water meter measurement time step data;

[0016] The verification data extraction module is used to extract the intelligent water meter measurement time step data of the last time step from the sequence distribution of the intelligent water meter measurement time step data as the intelligent water meter measurement self-supervised verification data;

[0017] The time series stacking encoding module is used to perform two-level stacking encoding of the measurement data time series on the intelligent water meter measurement time step data in the sequence distribution except the intelligent water meter measurement self-supervised verification data to obtain the intelligent water meter measurement time series context encoding vector;

[0018] The measurement inference module is used to perform intelligent water meter measurement inference based on the intelligent water meter measurement time series context encoding vector to obtain the intelligent water meter measurement inference time step data;

[0019] The anomaly detection module is used to determine whether there is data anomaly based on the comparison between the intelligent water meter measurement inference time step data and the intelligent water meter measurement self-supervised verification data.

[0020] Compared with the prior art, the intelligent water meter measurement data anomaly monitoring system and method provided by this application use the built-in communication module of the intelligent water meter to send the intelligent water meter measurement data to the remote monitoring center. At the same time, a data processing algorithm is deployed in the remote monitoring center to segment the intelligent water meter measurement data into continuous time series segments at a predetermined time step, and the measurement data segment at the end is used as the verification benchmark. By performing two-level stacked time series encoding on the intelligent water meter measurement data segments in the historical period, that is, extracting the short-term time series fluctuation characteristics inside the segment and mining the long-term time series correlation characteristics between segments, the deep time series characteristics of the intelligent water meter measurement data can be captured, and the water meter measurement data inference at the current time step is performed accordingly. Furthermore, based on the comparison between the inference result and the verification benchmark, the anomaly judgment of the measurement data is realized. This method can effectively reduce the dependence on preset rules, dynamically adapt to diverse water usage scenarios, and reduce the risks of false alarms and missed detections. Description of the Drawings

[0021] By describing the embodiments of the present application in more detail in combination with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The 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 to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 It is a flowchart of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application.

[0023] Figure 2 It is a schematic diagram of data flow of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application.

[0024] Figure 3 It is a flowchart of sub-step S4 of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application.

[0025] Figure 4 It is a flowchart of sub-step S43 of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application.

[0026] Figure 5 It is a flowchart of sub-step S432 of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application.

[0027] Figure 6 It is a flowchart of sub-step S6 of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application.

[0028] Figure 7 It is a block diagram of the intelligent water meter measurement data abnormal monitoring system according to an embodiment of the present application. Detailed implementation manners

[0029] As shown in the present application, unless the context clearly indicates an exceptional situation, words such as "a", "an", "one" and / or "the" are not specifically singular and may also include 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.

[0030] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0031] 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 operations before or below do not necessarily need to be executed precisely in order. On the contrary, according to the need, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0032] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0033] It should be noted that in the present application, all actions of obtaining data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where the data is located and obtaining authorization from the owner of the corresponding device.

[0034] To address the above technical problems, the present application proposes an intelligent water meter measurement data anomaly monitoring method. It uses the built-in communication module of the intelligent water meter to send the measurement data of the intelligent water meter to a remote monitoring center. At the same time, a data processing algorithm is deployed in the remote monitoring center. The measurement data of the intelligent water meter is segmented into continuous time series segments at a predetermined time step, and the end measurement data segment is used as a verification benchmark. By performing a two-level stacked time series encoding on the historical measurement data segments of the intelligent water meter, that is, extracting the short-term time series fluctuation characteristics within the segments and mining the long-term time series correlation characteristics between the segments, the deep time series characteristics of the intelligent water meter measurement data can be captured. Based on this, the measurement data of the water meter at the current time step is inferred, and then based on the comparison between the inference result and the verification benchmark, the anomaly judgment of the measurement data is realized. This method can effectively reduce the dependence on preset rules, dynamically adapt to diverse water usage scenarios, and reduce the risks of false alarms and missed detections.

[0035] Figure 1 It is a flowchart of the intelligent water meter measurement data anomaly monitoring method according to an embodiment of the present application. Figure 2 It is a schematic diagram of data flow of the intelligent water meter measurement data anomaly monitoring method according to an embodiment of the present application. As Figure 1 and Figure 2As shown, the abnormal monitoring method for the metering data of the intelligent water meter includes the steps of: S1, sending the metering data of the intelligent water meter to the remote monitoring center by using the built-in communication module of the intelligent water meter; S2, in the remote monitoring center, performing data segmentation on the metering data of the intelligent water meter based on a fixed time step to obtain the sequence distribution of the metering time step data of the intelligent water meter; S3, extracting the metering time step data of the last time step from the sequence distribution of the metering time step data of the intelligent water meter as the self-supervised calibration data for the metering of the intelligent water meter; S4, performing two-level stacking coding on the time series of the metering time step data of the intelligent water meter except the self-supervised calibration data for the metering of the intelligent water meter in the sequence distribution to obtain the context coding vector of the metering time series of the intelligent water meter; S5, performing metering inference on the intelligent water meter based on the context coding vector of the metering time series of the intelligent water meter to obtain the metering inference time step data; S6, determining whether there is data abnormality based on the comparison between the metering inference time step data and the self-supervised calibration data for the metering of the intelligent water meter.

[0036] In the above abnormal monitoring method for the metering data of the intelligent water meter, in step S1, the metering data of the intelligent water meter is sent to the remote monitoring center by using the built-in communication module of the intelligent water meter. It should be understood that since intelligent water meters are widely distributed and numerous in quantity, in order to achieve centralized management of data, in this application, data is remotely transmitted through the built-in communication module of the intelligent water meter, so as to facilitate centralized data management and abnormal monitoring work by using the high-efficiency data processing capabilities of the remote monitoring center. In a specific example of this application, the built-in communication module is an NB-IoT module or a LoRaWAN module.

[0037] Specifically, for the NB-IoT module, it can provide deep coverage based on the existing cellular network, so that even in basements or indoor environments with weak signals, a good connection state can be maintained. In addition, NB-IoT supports the simultaneous access of a large number of devices, which is particularly important for cities with a large number of deployed intelligent water meters. By using narrowband technology, NB-IoT can effectively reduce energy consumption, extend battery life, and reduce maintenance costs. Therefore, in an urban environment, especially for intelligent water meters that need to operate stably for a long time and are not easy to replace batteries, NB-IoT is an ideal choice.

[0038] On the other hand, the LoRaWAN module utilizes unlicensed frequency bands for long-distance data transmission, which is particularly suitable for locations in remote areas or where traditional cellular network coverage cannot be relied upon. LoRaWAN can not only support long-distance data transmission but also has low power consumption, making it a reliable choice for smart water meters in complex geographical environments. At the same time, since LoRaWAN adopts a star topology structure, that is, all nodes communicate directly with the gateway, it simplifies the network architecture and reduces the deployment cost. This feature enables LoRaWAN to demonstrate unparalleled advantages in some special application scenarios, such as water meter monitoring in rural areas.

[0039] After selecting the communication modules, the next step is to configure these modules. The configuration process involves multiple aspects, including but not limited to network parameter settings, security policy formulation, etc. First of all, it is necessary to ensure that each smart water meter can correctly identify and connect to the designated remote monitoring center. This requires the communication module to have good compatibility and stability. Secondly, in order to ensure data security, the built-in communication module needs to adopt a series of encryption measures, such as using the SSL / TLS protocol to protect the security of data during transmission and prevent information leakage. In addition, strict access control mechanisms can also be implemented to restrict access to the data of smart water meters to only authenticated users.

[0040] After completing the initial configuration, the smart water meter starts to collect metering data at a predetermined time interval and prepares to send it out. In this process, the built-in communication module is responsible for managing the entire data transmission process, from establishing the initial connection to maintaining stable communication, and then to handling any possible error situations. For example, in the case of data loss or damage, the communication module should be able to automatically detect these problems and initiate a retransmission mechanism to ensure data integrity. An effective error handling mechanism is crucial for ensuring the reliability of data transmission. It is worth noting that in addition to the basic data transmission function, the built-in communication module also needs to support advanced functions, such as firmware updates. With the development of technology and the continuous change of security threats, it has become increasingly important to regularly update the software version of the communication module. Through Over-the-Air (OTA) technology, remote firmware upgrades can be achieved without affecting the normal operation of the smart water meter, thereby enhancing the security and performance of the system.

[0041] In the above intelligent water meter measurement data anomaly monitoring method, in step S2, at the remote monitoring center, the intelligent water meter measurement data is segmented based on a fixed time step to obtain the sequence distribution of the intelligent water meter measurement time step data. It should be understood that the intelligent water meter measurement data is essentially a continuous time series. Considering that directly processing the complete data stream may lead to low computing efficiency due to too long time span and it is difficult to capture local abnormal fluctuations in the measurement data. Therefore, this application adopts a data segmentation strategy based on a predetermined time step. By segmenting the data according to a fixed time step (such as 7 days), the intelligent water meter measurement data can be transformed from continuous data into discrete serialized segments, and each segment represents an independent time unit, which helps to hierarchically model the short-term fluctuations (such as instantaneous water consumption mutation) of the measurement data within each independent time unit and the long-term correlation (such as daily water use pattern change) between each independent time unit.

[0042] In a specific embodiment, 7 days is selected as a time step, which can not only cover the changes in common water use patterns in daily life but also ensure the computing efficiency of data processing. Of course, the fixed time step can also be adjusted according to the actual situation, and this embodiment is not restrictive. In this way, the originally huge data stream spanning a long time period can be transformed into discrete serialized segments that are easy to manage and analyze. Each segment represents the water use situation within a specific time period, which not only helps to capture the short-term fluctuation characteristics in the measurement data, such as the mutation of instantaneous water consumption, but also can explore the long-term correlation between each independent time unit, such as the change trend of water use patterns between different days.

[0043] In the specific implementation process, it is necessary to preprocess all the original measurement data collected from the intelligent water meter to ensure the data quality and consistency. Then, according to the selected 7-day time step, start to segment these data. At this stage, it is important to ensure that the selection of the segmentation point does not split any meaningful data patterns or events. Therefore, in practical applications, the segmentation strategy may be adjusted according to specific water use habits and known water use peak periods, so that each time step data segment can as completely as possible reflect the water use behavior characteristics of users within that time period.

[0044] In addition, since the intelligent water meter measurement data is essentially a continuous time series, directly processing the entire data stream may lead to an excessive computing burden due to a large time span and it is difficult to accurately capture local abnormal fluctuations. After adopting the method of segmenting the data according to a fixed time step (such as 7 days), not only can the above problems be alleviated, but also the short-term time series fluctuation characteristics can be deeply explored within each time step, and at the same time, the long-term time series correlation characteristics can be explored between different time steps, which helps to build a more detailed and comprehensive data model, so as to more accurately reflect the water use behavior pattern of users and its change law over time.

[0045] In the above intelligent water meter measurement data anomaly monitoring method, in step S3, the intelligent water meter measurement time step data of the last time step is extracted from the sequence distribution of the intelligent water meter measurement time step data as the intelligent water meter measurement self-supervised verification data. It should be understood that traditional anomaly monitoring methods rely on preset rules, but the rules are prone to obsolescence in dynamic water usage scenarios, which limits the generalization ability of the model to new anomaly patterns. To overcome this defect, this application adopts a self-supervised learning strategy, that is, uses unlabeled intelligent water meter measurement data for self-verification. Here, the end intelligent water meter measurement time step data (the latest data segment) represents the current water usage state. By using it as the verification benchmark, historical data can be used to predict the current state, and a self-supervised learning framework can be constructed, thereby reducing the dependence on expert rules and dynamically adapting to changes in water usage behavior.

[0046] In the above intelligent water meter measurement data anomaly monitoring method, in step S4, the intelligent water meter measurement time step data in the sequence distribution of the intelligent water meter measurement time step data except the intelligent water meter measurement self-supervised verification data is subjected to two-level stacking encoding of the measurement data time series to obtain the intelligent water meter measurement time series context encoding vector. It should be understood that since the intelligent water meter measurement data has significant time series correlation, the time series pattern and dynamic change trend of water usage behavior can be mined through time series modeling and learning of the intelligent water meter measurement data in historical periods, so as to provide a basis for predicting the current water usage state and judging anomalies. Here, to avoid the verification data interfering with the model's independent learning of historical time series patterns, the verification data needs to be excluded from the sequence distribution of the intelligent water meter measurement time step data, and only the remaining historical data is used for time series feature learning to ensure that the time series reasoning process is not affected by the current state. Among them, Figure 3 FIG. is a flowchart of sub-step S4 of the intelligent water meter measurement data anomaly monitoring method according to an embodiment of the present application. As Figure 3 shown, step S4 includes the steps of: S41, defining the intelligent water meter measurement time step data in the sequence distribution of the intelligent water meter measurement time step data except the intelligent water meter measurement self-supervised verification data as the intelligent water meter measurement historical time step data to obtain a time series of the intelligent water meter measurement historical time step data; S42, respectively performing first-level time series encoding based on the forward LSTM model on each intelligent water meter measurement historical time step data in the time series of the intelligent water meter measurement historical time step data to obtain a time series of intelligent water meter measurement data first-level time series correlation feature vectors; S43, performing second-level time series context encoding based on a multi-dimensional constraint optimization mechanism on the time series of the intelligent water meter measurement data first-level time series correlation feature vectors to obtain the intelligent water meter measurement time series context encoding vector.

[0047] Specifically, in step S41, the water meter measurement time step data in the sequence distribution of the smart water meter measurement time step data, excluding the self-supervised verification data of the smart water meter measurement, is defined as the historical time step data of the smart water meter measurement to obtain a time series of the historical time step data of the smart water meter measurement. That is, by clearly defining all time step data except the self-supervised verification data as historical time step data, the boundary between training and verification data is structurally divided, providing a clear input range for subsequent time series modeling.

[0048] Specifically, in step S42, each of the historical time step data in the time series of the smart water meter measurement historical time step data is respectively encoded by a first-level time series based on a forward LSTM model to obtain a time series of the first-level time series correlation feature vectors of the smart water meter measurement data. It should be understood that since the water meter measurement data has strong time series dependence (such as daily and nightly water usage patterns), in order to capture the local time series fluctuations of the smart water meter measurement data in historical periods, the forward LSTM model, which has excellent performance in time series processing tasks, is used in this application to extract short-term time series fluctuation features from each of the historical time step data of the smart water meter measurement. Specifically, the LSTM model (Long Short-Term Memory) has memory units and gating mechanisms inside. When processing the historical time step data of the smart water meter measurement, it can selectively remember key historical states and filter out noise, effectively learning the time series dependence relationship in the data, thereby capturing the time series fluctuation features of the water consumption in each independent time unit, such as the daily or weekly water usage peaks and valleys, and then generating a time series of the first-level time series correlation feature vectors of the smart water meter measurement data. In addition, the "forward LSTM" is selected here instead of the bidirectional model because the water meter anomaly detection pays more attention to the one-way causal chain from history to the present and does not require reverse time series information.

[0049] Specifically, in step S43, a secondary time series context encoding based on a multi-dimensional constraint optimization mechanism is performed on the time series of the first-level time series correlation feature vectors of the smart water meter measurement data to obtain the smart water meter measurement time series context encoding vector. Specifically, since the above-mentioned first-level time series encoding only captures the short-term time series fluctuation characteristics within each independent time unit and cannot capture the long-term correlations across time steps (such as monthly or seasonal changes in water usage patterns), it may lead to a lack of a global perspective in model prediction and anomaly judgment. Therefore, in order to comprehensively reflect the time series characteristics of the smart water meter measurement data, the present application further performs secondary time series context encoding on the time series of the first-level time series correlation feature vectors of the smart water meter measurement data to achieve the mining of long-term time series correlation features of water consumption. In particular, considering that traditional sequence models (such as the standard Transformer) may introduce redundant correlations during sequence feature association and fusion, resulting in key patterns being drowned out by noise. In response to this, the present application introduces an independence constraint mechanism. During the message dynamic propagation process of the time series of the first-level time series correlation feature vectors of the smart water meter measurement data, by dynamically adjusting the information interaction intensity between time nodes, while retaining the cross-time step dependence relationship, the coupling of irrelevant features is suppressed, thereby effectively avoiding the interference of redundant information on the model prediction ability and generating a smart water meter measurement time series context encoding vector that integrates global time series context information. Among them, Figure 4 is a flowchart of sub-step S43 of the smart water meter measurement data anomaly monitoring method according to an embodiment of the present application. As Figure 4 shown, step S43 includes the steps of: S431, performing message passing space end constraint anchoring and axis constraint anchoring on the time series of the first-level time series correlation feature vectors of the smart water meter measurement data to obtain a first-level time series correlation message passing space end constraint anchoring encoding vector and a first-level time series correlation message passing space axis constraint anchoring encoding vector of the smart water meter measurement data; S432, using the first-level time series correlation message passing space end constraint anchoring encoding vector and the first-level time series correlation message passing space axis constraint anchoring encoding vector as constraint conditions, calculating the message passing constraint factors of each first-level time series correlation feature vector in the time series of the first-level time series correlation feature vectors of the smart water meter measurement data to obtain a time series of first-level time series correlation feature message passing constraint factors; S433, based on the time series of the first-level time series correlation feature message passing constraint factors, performing dynamic constraint transfer encoding on the time series of the first-level time series correlation feature vectors of the smart water meter measurement data to obtain the smart water meter measurement time series context encoding vector.

[0050] More specifically, step S431 includes: First, extract the last first-level time-series correlation feature vector of the smart water meter measurement data from the time series of the first-level time-series correlation feature vectors of the smart water meter measurement data as the end-constrained anchoring coding vector of the first-level time-series correlation message passing space of the smart water meter measurement data, which is expressed by the formula:

[0051]

[0052] Wherein, represents the time series of the first-level time-series correlation feature vectors of the smart water meter measurement data, 、 、 and respectively represent the 1st, 2nd, th, and th first-level time-series correlation feature vectors of the smart water meter measurement data in the time series of the first-level time-series correlation feature vectors of the smart water meter measurement data, is the number of the first-level time-series correlation feature vectors of the smart water meter measurement data, is the end-constrained anchoring coding vector of the first-level time-series correlation message passing space of the smart water meter measurement data.

[0053] Here, a dynamic anchoring mechanism for time-series features is constructed to strengthen the self-supervised verification ability of the model. It should be understood that by converting the last first-level time-series correlation feature vector of the smart water meter measurement data into the end-constrained anchoring coding vector of the first-level time-series correlation message passing space of the smart water meter measurement data, both the end-state information of the time series is retained, and the effective guidance of the dynamic propagation process is achieved through the feature independence constraint. The end-constrained anchoring coding vector of the first-level time-series correlation message passing space of the smart water meter measurement data generated by this anchoring coding method can set boundary conditions in the feature distribution sequence, thereby ensuring that the subsequent message passing process focuses on the nearest local fluctuation features.

[0054] Then, perform clustering analysis on the time series of the first-level time-series correlation feature vectors of the smart water meter measurement data to obtain the axis-constrained anchoring coding vector of the first-level time-series correlation message passing space of the smart water meter measurement data, which is expressed by the formula:

[0055]

[0056] Wherein, represents the clustering analysis function, and respectively represent the maximum value taking function and the minimum value taking function, is a preset hyperparameter used to adjust the sensitivity of the clustering analysis, represents The feature importance factor, represents the normalized exponential function, represents the clustering weight coefficient of represents the axis constraint anchoring coding vector of the first-level time series correlation message passing space of the intelligent water meter measurement data.

[0057] That is, through clustering analysis, the time series of the first-level time series correlation feature vectors of the intelligent water meter measurement data is skeletonized, and the extreme value boundary features represented by it can be automatically extracted. Combining the vector mapping mechanism, the original discrete time series fluctuations are transformed into a compact representation in the continuous probability distribution space, so as to reveal the potential structure of the data flow manifold through the normalization operation of the softmax function without explicitly defining rules. Through this clustering analysis method, the axis constraint anchoring coding vector of the first-level time series correlation message passing space of the generated intelligent water meter measurement data can reveal the time series dominant change mode of the water meter measurement data, which helps to limit the distortion amplitude of the feature space in the subsequent message passing process, reduce local noise interference, and improve the accuracy of abnormal detection of measurement data.

[0058] Figure 5 is a flowchart of sub-step S432 of the intelligent water meter measurement data abnormal monitoring method according to an embodiment of the present application. As Figure 5 shown, the step S432 includes steps: S4321, calculating the tail message passing constraint factor of each first-level time series correlation feature vector of the intelligent water meter measurement data in the time series of the first-level time series correlation feature vectors of the intelligent water meter measurement data relative to the tail constraint anchoring coding vector of the first-level time series correlation message passing space of the intelligent water meter measurement data; S4322, calculating the axis message passing constraint factor of each first-level time series correlation feature vector of the intelligent water meter measurement data in the time series of the first-level time series correlation feature vectors of the intelligent water meter measurement data relative to the axis constraint anchoring coding vector of the first-level time series correlation message passing space of the intelligent water meter measurement data; S4323, based on the tail message passing constraint factor and the axis message passing constraint factor of each first-level time series correlation feature vector of the intelligent water meter measurement data in the time series of the first-level time series correlation feature vectors of the intelligent water meter measurement data, calculating the message passing constraint factor of each first-level time series correlation feature vector of the intelligent water meter measurement data to obtain the time series of the first-level time series correlation feature message passing constraint factors.

[0059] In a specific example of the present application, the step S4321 is expressed by the formula:

[0060]

[0061] Wherein, represents the The position eigenvalue of the first-level time-series correlation feature vector of the intelligent water meter measurement data, represents the position eigenvalue of the end constraint anchoring coding vector of the first-level time-series correlation message passing space of the intelligent water meter measurement data, represents the logarithmic function with base 2, represents taking the absolute value, represents end message passing constraint factor of

[0062] That is, by measuring the interaction response strength between the first-level time-series correlation feature vector of the intelligent water meter measurement data and the end constraint anchoring coding vector of the first-level time-series correlation message passing space of the intelligent water meter measurement data, a local constraint weight system with conditional probability distribution is constructed, so as to dynamically adjust the importance of different features during the message passing process. When a certain feature vector highly matches the end reference coding, its constraint weight increases, indicating that this feature carries key time-series information compatible with the verification reference. Otherwise, the weight decreases, suggesting potential abnormal disturbances. Through this adaptive attention mechanism, the generated end message passing constraint factor enables the model to effectively focus on the core features maintaining time-series coherence and suppress noise interference.

[0063] In a specific example of this application, the step S4322 is expressed by the formula:

[0064]

[0065] where represents the square of the norm of the vector, represents the inverse hyperbolic cosine function, represents axis message passing constraint factor of

[0066] That is, similar to the calculation of the tail-end message passing constraint factor, in this application, by introducing the axis message passing constraint factor, a regularization constraint is imposed in the feature space to compress the time series of the first-order time series correlation feature vectors of the originally unbounded intelligent water meter measurement data onto a multi-dimensional sphere with clear semantic boundaries. Specifically, the axis message passing constraint factor measures the Poincaré distance between the first-order time series correlation feature vectors of the intelligent water meter measurement data and the axis reference encoding (i.e., the axis constraint anchoring encoding vector of the first-order time series correlation message passing space of the intelligent water meter measurement data), realizing the refined regulation of the feature importance. When the first-order time series correlation feature vectors of the intelligent water meter measurement data are close to the axis reference encoding, it indicates that this feature dominates in the time series evolution process, and its axis message passing constraint factor increases, otherwise it decreases, thereby enhancing the model's ability to identify the time series dominant features and further improving the accuracy and robustness of the measurement data anomaly detection.

[0067] Particularly, in a preferred example of this application, the step S4323 includes: First, perform message passing collaborative constraint optimization on the tail-end message passing constraint factor and the axis message passing constraint factor of the first-order time series correlation feature vectors of the intelligent water meter measurement data to obtain the optimized tail-end message passing constraint factor and the optimized axis message passing constraint factor, which is expressed by the formula:

[0068]

[0069] Wherein, and respectively represent the covariant unified tail-end message passing constraint factor and the covariant unified axis message passing constraint factor, represents the transpose of the matrix, and respectively represent the optimized tail-end message passing constraint factor and the optimized axis message passing constraint factor.

[0070] That is, considering that there may be contradictions between the local mutations (tail-end constraints) and the global trends (axis constraints) of the water meter measurement data in message passing, the present application further balances the two constraints by constructing a message passing collaborative constraint optimization mechanism to ensure the consistency and stability of the temporal features during the transmission process. Specifically, first, through the horizontal tight covariance mechanism of the message passing field, the standard space constraint matrix is applied to the tail-end message passing constraint factor and the axis message passing constraint factor, forcing them to establish a unified transmission paradigm in the global topological structure. At the same time, the component rotation matrix of the reversible space metric is used to impose a spinor tight constraint on the covariant representation, so that the optimized tail-end message passing constraint factor and the optimized axis message passing constraint factor not only maintain the elasticity of the horizontal feature propagation but also satisfy the continuity of the longitudinal temporal evolution. Through this collaborative optimization method, the metric parameters of the feature space are dynamically adjusted to balance short-term sensitivity and long-term memory. It should be understood that the horizontal tight covariance eliminates the problem of gradient disappearance on the feature propagation path, ensuring the complete transmission of abnormal signals at different time scales, avoiding premature dilution or amplification of local perturbations, and the spinor tight constraint maps the originally separated long-term and short-term constraints to the same geometric space, and the generated optimized tail-end message passing constraint factor and optimized axis message passing constraint factor significantly reduce the overfitting risk.

[0071] Then, the optimized tail-end message passing constraint factor and the optimized axis message passing constraint factor of the first-level temporal correlation feature vector of the intelligent water meter measurement data are weighted and fused and then input into the Sigmoid function for normalization processing to obtain the first-level temporal correlation feature message passing constraint factor of the intelligent water meter measurement data, which is expressed by the formula:

[0072]

[0073] Wherein, and respectively represent different weight coefficients, is the sigmoid function, is the first-level temporal correlation feature message passing constraint factor of the intelligent water meter measurement data.

[0074] It should be understood that at the feature expression level, the first-level temporal correlation feature message passing constraint factor of the weighted and fused intelligent water meter measurement data not only combines the advantages of the tail-end message passing constraint factor and the axis message passing constraint factor, but also through the normalization processing of the sigmoid function, the first-level temporal correlation feature message passing constraint factor of the intelligent water meter measurement data is mapped between 0 and 1, realizing the smooth regulation of the feature importance. This processing method not only enhances the sensitivity and recognition ability of the model to the feature importance, but also improves the stability and reliability of the abnormal detection results.

[0075] More specifically, step S433 is expressed by the formula:

[0076]

[0077] where represents the encoding vector of the metering time series context of the intelligent water meter.

[0078] That is, through the dynamic weighted pooling operation, the time series of the first-level time series correlation feature vectors of the intelligent water meter metering data is compressed into the encoding vector of the metering time series context of the intelligent water meter. By assigning different weight coefficients to the features of different time steps, the model's ability to capture composite abnormal patterns is significantly enhanced, and its adaptability to new abnormal scenarios is significantly improved.

[0079] In the above intelligent water meter metering data anomaly monitoring method, in step S5, intelligent water meter metering inference is performed based on the encoding vector of the metering time series context of the intelligent water meter to obtain the metering inference time step data of the intelligent water meter. In a specific example of the present application, step S5 includes: performing time series feature decoding based on the RNN model on the encoding vector of the metering time series context of the intelligent water meter to obtain the metering inference time step data of the intelligent water meter. It should be understood that after obtaining the encoding vector of the metering time series context of the intelligent water meter that integrates the global time series context information, it is necessary to further map it back to the original data space to predict the metering data of the intelligent water meter at the next time step. Here, the present application uses an RNN model (Recurrent Neural Network) to perform time series feature decoding on the encoding vector of the metering time series context of the intelligent water meter. Specifically, the encoding vector of the metering time series context of the intelligent water meter is used as the input of the RNN model, and through the hidden layer state transfer mechanism of the RNN model, the global time series context information of the metering data contained in the encoding vector is learned, and based on the learned evolution information of the metering data time series context, the predicted metering data of the intelligent water meter at the next time step is gradually deduced, thereby generating the metering inference time step data of the intelligent water meter.

[0080] In the above intelligent water meter metering data anomaly monitoring method, in step S6, based on the comparison between the metering inference time step data of the intelligent water meter and the self-supervised calibration data of the intelligent water meter metering, it is determined whether there is data anomaly. It should be understood that the metering inference time step data of the intelligent water meter is the current state predicted through the self-supervised learning framework based on historical intelligent water meter metering data, while the self-supervised calibration data of the intelligent water meter metering is the actually observed current state. By comparing the deviation of the actual data from the expected pattern, it can be intuitively judged whether there is an anomaly in the current intelligent water meter metering data. Where Figure 6It is a flowchart of sub-step S6 of the intelligent water meter measurement data anomaly monitoring method according to an embodiment of the present application. As Figure 6 shown, the step S6 includes steps: S61, calculating the cosine function value between the intelligent water meter measurement inference time step data and the intelligent water meter measurement self-supervised verification data as the measurement data time series consistency operator; S62, based on the comparison between the measurement data time series consistency operator and a preset threshold, determining whether there is data anomaly.

[0081] Specifically, in the step S61, the cosine function value between the intelligent water meter measurement inference time step data and the intelligent water meter measurement self-supervised verification data is calculated as the measurement data time series consistency operator. It should be understood that the cosine function value, as an index for measuring the similarity of the directions of two vectors, has a value range between [-1, 1]. When the directions of the two vectors are exactly the same, the cosine function value is 1; when the directions are exactly opposite, the cosine function value is -1; when the directions are perpendicular, the cosine function value is 0. Therefore, by calculating the cosine function value between the intelligent water meter measurement inference time step data and the intelligent water meter measurement self-supervised verification data, the similarity of the predicted data and the actual observed data in terms of direction can be quantitatively evaluated, thereby reflecting the consistency of the measurement data time series.

[0082] Specifically, in the step S62, based on the comparison between the measurement data time series consistency operator and a preset threshold, it is determined whether there is data anomaly. That is, if the measurement data time series consistency operator is greater than or equal to the preset threshold, it is determined that the intelligent water meter measurement data is normal, indicating that the current measurement state of the intelligent water meter conforms to the expected mode and no further intervention is required; otherwise, if the measurement data time series consistency operator is less than the preset threshold, it is determined that there is an anomaly in the intelligent water meter measurement data, prompting that the intelligent water meter needs to be inspected or maintained to avoid potential water resource waste or measurement error problems. This intelligent water meter anomaly detection method based on self-supervised learning not only improves the accuracy and efficiency of anomaly detection, but also provides strong support for the remote monitoring and maintenance of intelligent water meters.

[0083] In summary, the intelligent water meter measurement data anomaly monitoring method based on the embodiments of the present application is elucidated. It uses the built-in communication module of the intelligent water meter to send the intelligent water meter measurement data to the remote monitoring center. At the same time, a data processing algorithm is deployed in the remote monitoring center to divide the intelligent water meter measurement data into continuous time series segments at a predetermined time step, and uses the measurement data segment at the end as the verification benchmark. By performing two-level stacked time series encoding on the intelligent water meter measurement data segments in the historical period, that is, extracting the short-term time series fluctuation characteristics inside the segment and mining the long-term time series correlation characteristics between segments, the deep time series characteristics of the intelligent water meter measurement data are captured, and the water meter measurement data inference at the current time step is carried out based on this. Furthermore, based on the comparison between the inference result and the verification benchmark, the anomaly judgment of the measurement data is realized. This method can effectively reduce the dependence on preset rules, dynamically adapt to diverse water usage scenarios, and reduce the risks of false alarms and missed detections.

[0084] Furthermore, an intelligent water meter measurement data anomaly monitoring system is also provided.

[0085] Figure 7 FIG. is a block diagram of the intelligent water meter measurement data anomaly monitoring system according to the embodiments of the present application. As Figure 7 shown, the intelligent water meter measurement data anomaly monitoring system 100 according to the embodiments of the present application includes: a water meter measurement data transmission module 110, configured to send the intelligent water meter measurement data to the remote monitoring center by using the built-in communication module of the intelligent water meter; a water meter measurement data segmentation module 120, configured to segment the intelligent water meter measurement data at a fixed time step in the remote monitoring center to obtain a sequence distribution of the intelligent water meter measurement time step data; a verification data extraction module 130, configured to extract the intelligent water meter measurement time step data of the last time step from the sequence distribution of the intelligent water meter measurement time step data as the intelligent water meter measurement self-supervised verification data; a time series stacked encoding module 140, configured to perform two-level stacked encoding of the measurement data time series on the intelligent water meter measurement time step data in the sequence distribution except the intelligent water meter measurement self-supervised verification data to obtain an intelligent water meter measurement time series context encoding vector; a measurement inference module 150, configured to perform intelligent water meter measurement inference based on the intelligent water meter measurement time series context encoding vector to obtain intelligent water meter measurement inference time step data; and an anomaly detection module 160, configured to determine whether there is data anomaly based on the comparison between the intelligent water meter measurement inference time step data and the intelligent water meter measurement self-supervised verification data.

[0086] Here, those skilled in the art can understand that the specific operations of each module in the above intelligent water meter measurement data anomaly monitoring system have been described above with reference to Figures 1 to 6The description of the intelligent water meter measurement data anomaly monitoring method has been introduced in detail, and therefore, its repeated description will be omitted.

[0087] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, advantages, effects, etc. mentioned in the present invention are only examples and not limitations. It cannot be considered that these advantages, advantages, effects, etc. are essential for each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purposes of illustration and easy understanding, rather than limitations. The above details do not limit the present invention to necessarily adopt the above specific details to be implemented.

[0088] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0089] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting.

[0090] Finally, it should be noted that the above description has been given for the purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for monitoring abnormal measurement data of a smart water meter, characterized in that: include: Use the built-in communication module of the smart water meter to send the smart water meter measurement data to the remote monitoring center; At the remote monitoring center, data segmentation is performed on the smart water meter measurement data based on a fixed time step to obtain a sequence distribution of the smart water meter measurement time step data; Extracting the last time step of the smart water meter metering time step data from the sequence distribution of the smart water meter metering time step data as the smart water meter metering self-supervision verification data; The metering time step data of the smart water meter except the metering self-supervisory verification data of the smart water meter in the sequence distribution of the metering time step data of the smart water meter are subjected to metering data time series secondary stacking coding to obtain a smart water meter metering time series context coding vector, including: defining the metering time step data of the smart water meter except the metering self-supervisory verification data of the smart water meter in the sequence distribution of the metering time step data of the smart water meter as the metering historical time step data of the smart water meter to obtain a time series of the metering historical time step data of the smart water meter; performing a first-level time series coding based on a forward LSTM model on each smart water meter metering historical time step data in the time series of the metering historical time step data of the smart water meter to obtain a time series of a first-level time series correlation feature vector of the metering data of the smart water meter; performing a message passing space tail constraint anchoring and an axis constraint anchoring on the time series of the first-level time series correlation feature vector of the metering data of the smart water meter to obtain a smart water meter metering time series context coding vector. The first-level time series associated message transmission space tail constraint anchor coding vector of the smart water meter metering data and the first-level time series associated message transmission space axis constraint anchor coding vector of the smart water meter metering data; using the first-level time series associated message transmission space tail constraint anchor coding vector of the smart water meter metering data and the first-level time series associated message transmission space axis constraint anchor coding vector of the smart water meter metering data as constraint conditions, calculating the message transmission constraint factor of each first-level time series associated feature vector of the smart water meter metering data in the time series of the first-level time series associated feature vector of the smart water meter metering data to obtain the time series of the first-level time series associated feature message transmission constraint factor of the smart water meter metering data; based on the time series of the first-level time series associated feature message transmission constraint factor of the smart water meter metering data, dynamically constrain the transmission coding of the time series of the first-level time series associated feature vector of the smart water meter metering data to obtain the smart water meter metering time series context coding vector; Performing smart water meter measurement reasoning based on the smart water meter measurement time series context coding vector to obtain smart water meter measurement reasoning time step data; Based on the comparison between the smart water meter measurement reasoning time step data and the smart water meter measurement self-supervision verification data, it is determined whether there is data anomaly.

2. The method for monitoring abnormal measurement data of smart water meters according to claim 1, characterized in that: The built-in communication module is a NB-IoT module or a LoRaWAN module.

3. The method for monitoring abnormal measurement data of smart water meters according to claim 2 is characterized in that: The time series of the first-level time series correlation feature vector of the smart water meter measurement data is anchored by the tail constraint of the message transmission space and the axis constraint to obtain the first-level time series correlation message transmission space tail constraint anchor coding vector of the smart water meter measurement data and the first-level time series correlation message transmission space axis constraint anchor coding vector of the smart water meter measurement data, including: Extract the last first-level time series correlation feature vector of the smart water meter metering data from the time series of the first-level time series correlation feature vector of the smart water meter metering data as the tail-end constraint anchor coding vector of the first-level time series correlation message transmission space of the smart water meter metering data; A cluster analysis is performed on the time series of the first-level time series correlation feature vector of the smart water meter measurement data to obtain the first-level time series correlation message transmission space axis constraint anchor coding vector of the smart water meter measurement data.

4. The method for monitoring abnormal measurement data of smart water meters according to claim 3 is characterized in that: Taking the tail constraint anchor coding vector of the first-level time series associated message transmission space of the smart water meter metering data and the axis constraint anchor coding vector of the first-level time series associated message transmission space of the smart water meter metering data as constraint conditions, calculating the message transmission constraint factor of each first-level time series associated feature vector of the smart water meter metering data in the time series of the first-level time series associated feature vector of the smart water meter metering data to obtain the time series of the first-level time series associated feature message transmission constraint factor of the smart water meter metering data, including: Calculate the tail end message transmission constraint factor of each first-level time series correlation feature vector of the smart water meter metering data in the time series of the first-level time series correlation feature vector of the smart water meter metering data relative to the tail end constraint anchor coding vector of the first-level time series correlation message transmission space of the smart water meter metering data; Calculate the axis message transmission constraint factor of each first-level time series correlation feature vector of the smart water meter metering data in the time series of the first-level time series correlation feature vector of the smart water meter metering data relative to the axis constraint anchor coding vector of the first-level time series correlation message transmission space of the smart water meter metering data; Based on the tail message transmission constraint factor and the axis message transmission constraint factor of each first-level time series correlation feature vector of the smart water meter metering data in the time series of the first-level time series correlation feature vector of the smart water meter metering data, the message transmission constraint factor of each first-level time series correlation feature vector of the smart water meter metering data is calculated to obtain the time series of the first-level time series correlation feature message transmission constraint factor of the smart water meter metering data.

5. The method for monitoring abnormal measurement data of smart water meters according to claim 4 is characterized in that: Based on the tail message transmission constraint factor and the axis message transmission constraint factor of each first-level time series correlation feature vector of the smart water meter metering data in the time series of the first-level time series correlation feature vector of the smart water meter metering data, the message transmission constraint factor of each first-level time series correlation feature vector of the smart water meter metering data is calculated to obtain the time series of the first-level time series correlation feature message transmission constraint factor of the smart water meter metering data, including: Perform message transmission collaborative constraint optimization on the tail end message transmission constraint factor and the axis line message transmission constraint factor of the first-level time series correlation feature vector of the smart water meter measurement data to obtain an optimized tail end message transmission constraint factor and an optimized axis line message transmission constraint factor; The optimized tail end message transmission constraint factor and the optimized axis message transmission constraint factor of the first-level time series correlation feature vector of the smart water meter measurement data are weighted and fused, and then input into the Sigmoid function for normalization processing to obtain the first-level time series correlation feature message transmission constraint factor of the smart water meter measurement data.

6. The method for monitoring abnormal measurement data of smart water meters according to claim 5, characterized in that: Performing smart water meter measurement reasoning based on the smart water meter measurement time series context coding vector to obtain smart water meter measurement reasoning time step data includes: The smart water meter measurement time series context encoding vector is decoded based on the time series feature of the RNN model to obtain the smart water meter measurement reasoning time step data.

7. The method for monitoring abnormal measurement data of smart water meters according to claim 6, characterized in that: Based on the comparison between the smart water meter measurement reasoning time step data and the smart water meter measurement self-supervision verification data, determining whether there is data anomaly includes: Calculate the cosine function value between the smart water meter measurement reasoning time step data and the smart water meter measurement self-supervision verification data as a measurement data time series consistency operator; Based on the comparison between the measurement data time series consistency operator and a preset threshold, it is determined whether there is data anomaly.

8. A smart water meter measurement data abnormality monitoring system, using the smart water meter measurement data abnormality monitoring method according to claim 1, characterized in that: include: A water meter measurement data transmission module, used to send the smart water meter measurement data to a remote monitoring center using the built-in communication module of the smart water meter; A water meter measurement data segmentation module is used to perform data segmentation on the smart water meter measurement data based on a fixed time step in the remote monitoring center to obtain a sequence distribution of the smart water meter measurement time step data; A verification data extraction module is used to extract the last time step of the smart water meter measurement time step data from the sequence distribution of the smart water meter measurement time step data as the smart water meter measurement self-supervision verification data; A time series stacking coding module is used to perform metering data time series secondary stacking coding on the smart water meter metering time step data in the sequence distribution of the smart water meter metering time step data except the smart water meter metering self-supervision verification data to obtain a smart water meter metering time series context coding vector; A metering reasoning module, used for performing metering reasoning of the smart water meter based on the smart water meter metering time series context coding vector to obtain metering reasoning time step data of the smart water meter; The anomaly detection module is used to determine whether there is a data anomaly based on the comparison between the intelligent water meter measurement reasoning time step data and the intelligent water meter measurement self-supervision verification data.

Images