Water management method, system and device

Abstract

The application relates to the technical field of water affair electronic data processing, and discloses a water affair management method, system and device. First, users are divided into different water use mode sets through cluster analysis; second, individual fluctuation and mode scarcity are comprehensively considered to calculate a monitoring priority, hierarchical collection is realized, the core lies in defining a resilience index which integrates short-term stability and group universality, and the resilience index is used to accurately identify abnormal users; once an abnormal user is found, the similarity of the water use curves of other users in the set to which the abnormal user belongs is analyzed, early warning and positioning of risk transmission are realized, and the monitoring frequency of abnormal users and users affected by the transmission is automatically increased. The scheme realizes the change from uniform monitoring to dynamic and accurate monitoring based on behavior portraits and risk transmission, and significantly improves the monitoring efficiency and risk warning capability.

Description

Technical Field

[0001] This invention relates to the field of water affairs electronic data processing technology, specifically to a water affairs management method, system, and device. Background Technology

[0002] Traditional water data monitoring systems typically collect water consumption data from all users at a fixed frequency, a model with significant technical drawbacks. On the one hand, with a massive number of users, evenly distributing monitoring resources leads to heavy system load and high storage costs. On the other hand, it lacks the ability to identify the differentiated water consumption characteristics of users of different types, such as residential, commercial, and industrial users, making it impossible to achieve focused monitoring of high-risk users.

[0003] More advanced water management solutions have attempted to classify users through cluster analysis or introduce mechanisms for dynamically adjusting data collection frequency. However, these solutions often suffer from logical inconsistencies or application limitations. For example, user profiling fails to establish deep integration with subsequent monitoring resource allocation and anomaly detection mechanisms; frequency adjustment strategies often rely on single indicators without comprehensively considering the generality of the behavior of the user's group, leading to unnecessary over-monitoring of specific but stable users or insufficient response to minor disturbances in vulnerable groups. Furthermore, existing methods generally lack consideration for the propagation of abnormal events; when a user exhibits an anomaly, it cannot automatically determine whether similar users face collateral risks, thus missing the opportunity for intervention. Summary of the Invention

[0004] This invention provides a water management method, system, and apparatus.

[0005] The technical solution of this invention is as follows:

[0006] A water management method includes the following steps:

[0007] S1. Collect water consumption data from all users at a fixed frequency to form a baseline time series dataset. Extract the amplitude and phase features of each user's baseline time series dataset. Based on the amplitude and phase features of all users, perform clustering to divide all users into K sets. Then, perform topological analysis according to the user's geographical location to divide each set into multiple subsets of users in different regions.

[0008] S2. Based on the total number of users, the number of users in the subset, and the user's water consumption variance, calculate the monitoring priority score for each user, and classify users into different frequency collection levels according to the monitoring priority score.

[0009] S3. Based on the standard deviation of the user's water consumption and the number of users in the subset, calculate the resilience index of each user. When the resilience index is lower than the first threshold, the user is identified as having abnormal resilience, and the collection frequency is doubled and an alarm is triggered.

[0010] S4. For each other user in the subset of users with abnormal resilience, calculate the similarity between their water usage curve and the water usage curve of users with abnormal resilience during the selected assessment period. If the similarity is greater than the similarity threshold, it is determined that the risk has spread to that user. For the user whose risk has been spread, double the collection frequency and issue an alarm.

[0011] In S2, a monitoring priority score is calculated for each user based on the total number of users, the number of users in the same set, and the user's water consumption variance. Specifically:

[0012] ,

[0013] in, The variance of water consumption for users, The number of users in the subset. This represents the total number of users.

[0014] In S2, users are divided into different collection frequency levels based on their monitoring priority scores:

[0015] Find the first and second target percentiles of the monitoring priority score, and classify users whose monitoring priority score is greater than or equal to the first target percentile as high-frequency collection level;

[0016] Users whose monitoring priority score is greater than or equal to the second target percentile but less than the first target percentile are classified as mid-frequency acquisition level.

[0017] Users whose monitoring priority score is less than the second target percentile are classified as low-frequency collection level.

[0018] In S3, the resilience index for each user is calculated based on the standard deviation of their water consumption and the number of users in their set, specifically as follows:

[0019] ,

[0020] in, It is the standard deviation of user i's water consumption data in the most recent evaluation period. The number of users in the subset. , These are the weighting coefficients.

[0021] In S3, the first threshold is the average value of each user over several past evaluation periods.

[0022] The method for selecting the K sets in S1 is as follows: Based on historical water usage, determine the range of K values, perform preliminary clustering for each K value, calculate the sum of squared errors within all preliminary clusters, and based on the sum of squared errors within all preliminary clusters, select the K value of the inflection point according to the elbow rule for clustering, thus dividing all users into K sets.

[0023] In S1, the amplitude and phase features of all users are clustered to divide all users into K sets. Specifically, a sample is randomly selected from the sample dataset composed of the amplitude and phase features of all users as the first initial cluster center. For each sample in the sample dataset, the shortest distance from it to the first initial cluster center is calculated. The shortest distance of each sample is used as the weight for it to be selected as the next center. The selection probability of each sample being selected as the next center is calculated. The sample with the highest probability is selected as the next cluster center. This process continues until K initial cluster centers are selected.

[0024] For each sample in the sample dataset, calculate its Euclidean distance to the K initial cluster centers, assign each sample to the cluster represented by the cluster center closest to it, resulting in multiple clusters. Calculate the mean of all samples in each cluster and use it as the new center of that cluster. When the new center of the iterative cluster reaches the maximum number of iterations, the iteration terminates, and the current cluster partition is output as K sets.

[0025] Users in each subset of S1 are users within the same area and using the same water usage pattern.

[0026] A water management system for implementing the above-mentioned water management method includes:

[0027] The set partitioning module collects water consumption data from all users at a fixed frequency to form a baseline time series dataset. It extracts the amplitude and phase features of each user's baseline time series dataset, performs clustering based on the amplitude and phase features of all users, divides all users into K sets, and then performs topological analysis according to the user's geographical location to divide each set into multiple subsets of users in different regions.

[0028] The frequency collection setting module calculates a monitoring priority score for each user based on the total number of users, the number of users in the subset, and the user's water consumption variance. Based on the monitoring priority score, the user is divided into different frequency collection levels.

[0029] The anomaly detection module calculates the resilience index for each user based on the standard deviation of the user's water consumption and the number of users in the subset. When the resilience index is lower than the first threshold, the user is identified as having abnormal resilience, and the sampling frequency is doubled and an alarm is triggered.

[0030] The propagation detection module calculates the similarity between the water usage curve of each other user in the subset of users with abnormal resilience during the selected assessment period and the water usage curve of users with abnormal resilience. If the similarity is greater than the similarity threshold, it is determined that the risk has been propagated to that user. For the user whose risk has been propagated, the sampling frequency is doubled and an alarm is triggered.

[0031] A water management device includes a processor and a memory, wherein the processor implements a water management method when executing a computer program stored in the memory.

[0032] The beneficial effects of this application are as follows:

[0033] 1. By clustering users, if other users in the same area also experience abnormal water usage during the same period, it can be more accurately determined that this is not an anomaly of a single user, but is likely a regional event. This provides a scientific and interpretable data foundation for subsequent differentiated and precise management.

[0034] 2. By integrating priority scores based on individual volatility and pattern scarcity, the limited monitoring resources are optimized from average allocation to on-demand allocation, enabling focused coverage of high-risk and high-uncertainty users while controlling overall costs.

[0035] 3. The resilience index can not only sensitively capture abnormal fluctuations, but also assess anomalies by combining user behavior patterns, thus achieving more accurate alarms. It also breaks through the traditional single-point anomaly detection approach. By analyzing the behavioral similarity within the same user group, it can proactively discover potential risk spread, achieving an upgrade from point monitoring to area monitoring. Detailed Implementation

[0036] Example

[0037] A water management method includes the following steps:

[0038] S1. Collect water consumption data from all users at a fixed frequency to form a baseline time series dataset. Extract the amplitude and phase features of each user's baseline time series dataset. Based on the amplitude and phase features of all users, perform clustering to divide all users into K sets. Then, perform topological analysis according to the user's geographical location to divide each set into multiple subsets of users in different regions.

[0039] Specifically, after forming the baseline time series dataset, a water consumption line graph is drawn with time as the horizontal axis and water consumption as the vertical axis. Amplitude features are extracted. During the evaluation period, the standard deviation of water consumption and the difference between peak and trough values ​​are calculated. The two are then weighted and summed to obtain the amplitude features, which reflect the degree of fluctuation in the user's water consumption during this period.

[0040] Phase feature extraction is performed, and by analyzing the benchmark time series dataset, the user's daily water usage time and duration are identified.

[0041] Clustering is performed based on the amplitude and phase characteristics of all users.

[0042] Furthermore, the clustering steps are as follows: First, based on historical water usage, determine the range of K values, perform preliminary clustering for each K value, calculate the sum of squared intra-cluster errors of all preliminary clusters, and based on the sum of squared intra-cluster errors of all preliminary clusters, select the K value of the inflection point according to the elbow rule for clustering, and divide all users into K sets.

[0043] Then, randomly select a sample from the sample dataset composed of the amplitude and phase features of all users as the first initial cluster center. For each sample in the sample dataset, calculate its shortest distance to the first initial cluster center. Use the shortest distance of each sample as the weight for it to be selected as the next center. Calculate the selection probability of each sample being selected as the next center. Select the sample with the highest probability as the next cluster center, until K initial cluster centers are selected.

[0044] Finally, for each sample in the sample dataset, calculate its Euclidean distance to the K initial cluster centers, assign each sample to the cluster represented by the cluster center closest to it, resulting in multiple clusters. Calculate the mean of all samples in each cluster as the new center of that cluster. When the new center of the iterative cluster reaches the maximum number of iterations, the iteration terminates, and the current cluster partition is output as K sets.

[0045] Obtain the geographic tags of users within the set, and based on the geographic tags, divide the set into several subsets based on its topology.

[0046] Ultimately, users in each subset simultaneously meet two conditions: all users share the same water usage behavior pattern, and all users have the same geographic tag.

[0047] One of the purposes of this application is to monitor whether there are any anomalies among users, including but not limited to leaks in the main water supply pipeline and branch water supply pipelines. Therefore, it is necessary to clarify that the users to be clustered and the subsequent steps are those in the same region and on the same main water supply pipeline.

[0048] Amplitude and phase characteristics can be understood as similar water usage behaviors among users within the same set. This division facilitates subsequent anomaly investigation, checking whether the anomaly is caused by a user's water usage behavior, and whether users with similar water usage behaviors have also experienced anomalies. When a user uses a large amount of water, it may cause fluctuations in the pipeline pressure, thus slightly affecting the water consumption of neighboring users. It may also cause a large number of users within the same set to experience anomalies due to leaks in branch water supply pipes. Therefore, regional characteristics are also essential. This set division method can lay the foundation for subsequent identification of users with resilience anomalies, making subsequent judgment steps faster and more reliable.

[0049] S2. Based on the total number of users, the number of users in the subset, and the user's water consumption variance, calculate the monitoring priority score for each user, and classify users into different frequency collection levels according to the monitoring priority score.

[0050] The method for calculating the monitoring priority score for each user is as follows:

[0051] ,

[0052] in, The variance of water consumption for users, The number of users in the subset. This represents the total number of users.

[0053] This measure is used to assess the instability of a user's water consumption. The higher the value, the more drastic the fluctuation in the user's water consumption, the more unpredictable their behavior, and the greater the possibility or impact of sudden increases or decreases. Therefore, a higher frequency of data collection is needed to capture these changes.

[0054] This ratio measures the rarity of a user's subset and the uncertainty of how people perceive them; the smaller the subset, the larger the ratio. Directly using the ratio can overemphasize the influence of extremely small subsets, potentially leading to a few users dominating the monitoring priority score. The logarithmic transformation converts the linearly growing ratio into a logarithmically growing one, thus smoothing and compressing the data.

[0055] To illustrate this, consider an example: a subset containing only one user has an uncertainty of log( / 2); a subset containing 10 users has an uncertainty of log( / 11). The difference between the two is logarithmic, rather than absolute, making the monitoring priority scores more balanced.

[0056] Furthermore, the increase in scarcity from a set of 1000 users to a subset of 500 users is far greater than the increase from a set of 20 users to a subset of 10 users. Therefore, using a logarithmic function here better reflects this idea: when the subset already has a small number of users, reducing the number of people further will result in a gradual and slower increase in uncertainty, rather than a linear or exponential increase.

[0057] Users are categorized into different collection frequency levels based on their monitoring priority scores:

[0058] Find the first and second target percentiles of the monitoring priority score, and classify users whose monitoring priority score is greater than or equal to the first target percentile as high-frequency collection level;

[0059] Users whose monitoring priority score is greater than or equal to the second target percentile but less than the first target percentile are classified as mid-frequency acquisition level.

[0060] Users whose monitoring priority score is less than the second target percentile are classified as low-frequency collection level.

[0061] It's important to note that percentiles are a statistical concept. They represent the percentage of monitoring priority scores that are lower than or equal to this value. Let's illustrate this with a specific example of dividing the monitoring priority scores of 10 users into different collection frequency levels. Suppose the monitoring priority scores of the 10 users, sorted from lowest to highest, are (1.2, 1.5, 2.0, 2.8, 3.3, 4.1, 5.0, 6.2, 8.5, 12.0). The first target percentile and the second target percentile are 80 and 50, respectively. First, we need to know the 80th percentile, which represents 80% of the users' monitoring priority scores that are lower than or equal to it. Therefore, the ninth and tenth users are assigned to the high-frequency collection level. Similarly, the sixth to eighth users are assigned to the medium-frequency collection level, and the first to fifth users are assigned to the low-frequency collection level.

[0062] This tiered approach takes into account both the water usage risk of each user and the uncertainty of the subset they belong to. When a user is a high-risk factory with large and volatile water usage, even if they belong to a subset, their monitoring priority score will be high, and they will receive high-frequency monitoring. When a user is a single-person household with small and stable water usage and a unique water usage pattern, and they are the only one in the subset, their monitoring priority score will also increase due to the scarcity factor, and they will receive medium / high-frequency monitoring to prevent unknown types of anomalies from occurring.

[0063] In summary, this step can allocate appropriate monitoring frequencies to users of various types or water usage behaviors, achieving good monitoring results and enabling timely detection and response to various anomalies.

[0064] S3. Based on the standard deviation of the user's water consumption and the number of users in the subset, calculate the resilience index for each user. When the resilience index is lower than the first threshold, the user is identified as having abnormal resilience, and the collection frequency is doubled and an alarm is triggered.

[0065] The resilience index for each user is calculated as follows:

[0066] ,

[0067] in, It is the standard deviation of user i's water consumption data in the most recent evaluation period. The number of users in the subset. , These are the weighting coefficients.

[0068] It directly measures short-term water usage fluctuations. The smaller the standard deviation, the more stable the water usage. After the reciprocal transformation, the small fluctuation is converted into a large value, so that it is consistent with the positive meaning of high resilience. That is, the more stable the water usage, the higher the short-term stability component. This is directly related to the user's current real-time water usage behavior and is a dynamic indicator that changes at any time.

[0069] The numerical value reflects the prevalence or commonness of user behavior patterns. For example, the peak water usage patterns of millions of households during morning and evening hours are more resilient than the unique water usage patterns of a special factory that relies on a single production line. The patterns of special factories are more fragile and more prone to drastic fluctuations due to changes in a single internal factor. In other words, a small fluctuation in the water usage patterns of a general user may just be random noise, while a fluctuation of the same magnitude in a niche user may mean that their water usage patterns are abnormal, thus improving the ability to detect anomalies.

[0070] The weighting coefficients can be set based on the actual situation of the region. If real-time fluctuations are considered more important, α can be set to be large and β to be small; if the robustness of the group is considered more important, β can be adjusted to be larger.

[0071] Preferably, α=β=1 can be initially set, and then fine-tuned based on the effect during operation.

[0072] Furthermore, when the resilience index is lower than the first threshold, which is the average value of each user over several past evaluation periods, the user is identified as having abnormal resilience. The collection frequency is doubled and an alarm is triggered. For example, if the current collection frequency is once per hour, then the collection frequency will need to be once every half hour.

[0073] Alarms may not be processed immediately; more data on resilient and abnormal users needs to be collected to inform the handling of alarms.

[0074] S4. For each other user in the subset of users with abnormal resilience, calculate the similarity between their water usage curve and the water usage curve of users with abnormal resilience during the selected assessment period. If the similarity is greater than the similarity threshold, it is determined that the risk has spread to that user. For the user whose risk has been spread, double the collection frequency and issue an alarm.

[0075] When S3 determines that a user is a user with abnormal resilience, S4 is started to obtain the baseline time series dataset of the user with abnormal resilience in S1, obtain the water use behavior subset to which the user with abnormal resilience belongs, and the baseline time series dataset of all other users in the subset.

[0076] For each other user, obtain their water consumption time-series data within a selected assessment period during the abnormal period of the user with resilience anomalies. The selected assessment period includes the period containing the resilience anomaly data. Determine whether the two time series are similar in shape, specifically using the following method:

[0077] Calculate the first-order difference sequences for other users and users with abnormal resilience. The first-order difference sequence is the difference between adjacent time points within the selected evaluation period, which represents the water usage change trend at each time point.

[0078] Calculate the cosine similarity between two first-order difference sequences:

[0079] ,

[0080] Where S ranges from [-1, 1], where 1 indicates that the trends are completely consistent, and -1 indicates that they are completely opposite. For the first-order difference sequence of users with abnormal resilience, For other users' first-order difference sequences.

[0081] To map it to the [0,1] range for comparison, the cosine similarity is further processed to obtain the similarity score:

[0082] .

[0083] The design of this application allows for rapid similarity calculation, is insensitive to the absolute amount of water used, and can effectively detect synchronous fluctuations among different users.

[0084] If the similarity to other users is greater than the similarity threshold, it is determined that the risk has spread to that user, and the user is marked as the one whose risk has been spread.

[0085] If the similarity is less than or equal to the similarity threshold, the user is considered not to have been spread.

[0086] The data collection frequency of the affected users will be immediately doubled, for example, from once every 2 hours to once every 1 hour. At the same time, a risk observation period will be initiated for the affected users, during which they will be subject to closer monitoring.

[0087] This step directly measures the synchronicity of the water usage change rhythms of two users. It is not affected by the user scale or the size of the basic water consumption. It can fairly compare residential households and industrial households. The closer the similarity is to 1, the closer the water usage behavior of the two is to each other in time, and the higher the possibility of risk propagation along this path.

[0088] A water management system for implementing the above-mentioned water management method includes:

[0089] The set partitioning module collects water consumption data from all users at a fixed frequency to form a baseline time series dataset. It extracts the amplitude and phase features of each user's baseline time series dataset, performs clustering based on the amplitude and phase features of all users, divides all users into K sets, and then performs topological analysis according to the user's geographical location to divide each set into multiple subsets of users in different regions.

[0090] The frequency collection setting module calculates a monitoring priority score for each user based on the total number of users, the number of users in the subset, and the user's water consumption variance. Based on the monitoring priority score, the user is divided into different frequency collection levels.

[0091] The anomaly detection module calculates the resilience index for each user based on the standard deviation of the user's water consumption and the number of users in the subset. When the resilience index is lower than the first threshold, the user is identified as having abnormal resilience, and the sampling frequency is doubled and an alarm is triggered.

[0092] The propagation detection module calculates the similarity between the water usage curve of each other user in the subset of users with abnormal resilience during the selected assessment period and the water usage curve of users with abnormal resilience. If the similarity is greater than the similarity threshold, it is determined that the risk has been propagated to that user. For the user whose risk has been propagated, the sampling frequency is doubled and an alarm is triggered.

[0093] A water management device includes a processor and a memory, wherein the processor implements a water management method when executing a computer program stored in the memory.

Claims

1. A water management method, characterized in that, Includes the following steps: S1. Collect water consumption data from all users at a fixed frequency to form a baseline time series dataset. Extract the amplitude and phase features of each user's baseline time series dataset. Based on the amplitude and phase features of all users, perform clustering to divide all users into K sets. Then, perform topological analysis according to the user's geographical location to divide each set into multiple subsets of users in different regions. Clustering is performed on the amplitude and phase features of all users to divide all users into K sets. Specifically, a sample is randomly selected from the sample dataset composed of the amplitude and phase features of all users as the first initial cluster center. For each sample in the sample dataset, the shortest distance from it to the first initial cluster center is calculated. The shortest distance of each sample is used as the weight for it to be selected as the next center. The selection probability of each sample being selected as the next center is calculated. The sample with the highest probability is selected as the next cluster center. This process continues until K initial cluster centers are selected. For each sample in the sample dataset, calculate its Euclidean distance to the K initial cluster centers, assign each sample to the cluster represented by the cluster center closest to it, resulting in multiple clusters. Calculate the mean of all samples in each cluster and use it as the new center of that cluster. When the new center of the iterative cluster reaches the maximum number of iterations, the iteration terminates, and the current cluster partition is output as K sets. The method for selecting the K sets is as follows: Based on historical water usage, determine the range of K values, perform preliminary clustering for each K value, calculate the sum of squared intra-cluster errors of all preliminary clusters, and based on the sum of squared intra-cluster errors of all preliminary clusters, select the K value of the inflection point according to the elbow rule for clustering, and divide all users into K sets. S2. Based on the total number of users, the number of users in the subset, and the user's water consumption variance, calculate the monitoring priority score for each user, and classify users into different frequency collection levels according to the monitoring priority score. Based on the total number of users, the number of users in the same set, and the user's water consumption variance, a monitoring priority score is calculated for each user, specifically as follows: ， in, The variance of water consumption for users, The number of users in the subset. Total number of users; S3. Based on the standard deviation of the user's water consumption and the number of users in the subset, calculate the resilience index of each user. When the resilience index is lower than the first threshold, the user is identified as having abnormal resilience, and the collection frequency is doubled and an alarm is triggered. Based on the standard deviation of a user's water consumption and the number of users in their set, the resilience index for each user is calculated as follows: ， in, It is the standard deviation of user i's water consumption data in the most recent evaluation period. The number of users in the subset. , These are the weighting coefficients; S4. For each other user in the subset of users with abnormal resilience, calculate the similarity between their water usage curve and the water usage curve of users with abnormal resilience during the selected assessment period. If the similarity is greater than the similarity threshold, it is determined that the risk has spread to that user. For the user whose risk has been spread, double the collection frequency and issue an alarm.

2. The water management method according to claim 1, characterized in that, In S2, users are divided into different collection frequency levels based on their monitoring priority scores: Find the first and second target percentiles of the monitoring priority score, and classify users whose monitoring priority score is greater than or equal to the first target percentile as high-frequency collection level; Users whose monitoring priority score is greater than or equal to the second target percentile but less than the first target percentile are classified as mid-frequency acquisition level. Users whose monitoring priority score is less than the second target percentile are classified as low-frequency collection level.

3. The water management method according to claim 1, characterized in that, The first threshold in S3 is the average value of each user over several past evaluation periods.

4. The water management method according to claim 1, characterized in that, Users in each subset of S1 are users within the same area and using the same water usage pattern.

5. A water management system for implementing the water management method according to any one of claims 1-4, characterized in that, include: The set partitioning module collects water consumption data from all users at a fixed frequency to form a baseline time series dataset. It extracts the amplitude and phase features of each user's baseline time series dataset, performs clustering based on the amplitude and phase features of all users, divides all users into K sets, and then performs topological analysis according to the user's geographical location to divide each set into multiple subsets of users in different regions. Clustering is performed on the amplitude and phase features of all users to divide all users into K sets. Specifically, a sample is randomly selected from the sample dataset composed of the amplitude and phase features of all users as the first initial cluster center. For each sample in the sample dataset, the shortest distance from it to the first initial cluster center is calculated. The shortest distance of each sample is used as the weight for it to be selected as the next center. The selection probability of each sample being selected as the next center is calculated. The sample with the highest probability is selected as the next cluster center. This process continues until K initial cluster centers are selected. For each sample in the sample dataset, calculate its Euclidean distance to the K initial cluster centers, assign each sample to the cluster represented by the cluster center closest to it, resulting in multiple clusters. Calculate the mean of all samples in each cluster and use it as the new center of that cluster. When the new center of the iterative cluster reaches the maximum number of iterations, the iteration terminates, and the current cluster partition is output as K sets. The method for selecting the K sets is as follows: Based on historical water usage, determine the range of K values, perform preliminary clustering for each K value, calculate the sum of squared intra-cluster errors of all preliminary clusters, and based on the sum of squared intra-cluster errors of all preliminary clusters, select the K value of the inflection point according to the elbow rule for clustering, and divide all users into K sets. The frequency collection setting module calculates a monitoring priority score for each user based on the total number of users, the number of users in the subset, and the user's water consumption variance. Based on the monitoring priority score, the user is divided into different frequency collection levels. Based on the total number of users, the number of users in the same set, and the user's water consumption variance, a monitoring priority score is calculated for each user, specifically as follows: ， in, The variance of water consumption for users, The number of users in the subset. Total number of users; The anomaly detection module calculates the resilience index for each user based on the standard deviation of the user's water consumption and the number of users in the subset. When the resilience index is lower than the first threshold, the user is identified as having abnormal resilience, and the sampling frequency is doubled and an alarm is triggered. Based on the standard deviation of a user's water consumption and the number of users in their set, the resilience index for each user is calculated as follows: ， in, It is the standard deviation of user i's water consumption data in the most recent evaluation period. The number of users in the subset. , These are the weighting coefficients; The propagation detection module calculates the similarity between the water usage curve of each other user in the subset of users with abnormal resilience during the selected assessment period and the water usage curve of users with abnormal resilience. If the similarity is greater than the similarity threshold, it is determined that the risk has been propagated to that user. For the user whose risk has been propagated, the sampling frequency is doubled and an alarm is triggered.

6. A water management device, characterized in that, It includes a processor and a memory, wherein the processor implements a water management method as described in any one of claims 1-4 when executing a computer program stored in the memory.

Images