Urban water prediction and scheduling method based on user behavior modeling and storage device

Abstract

The application discloses a city water use prediction scheduling method based on user behavior modeling and a storage device, relates to the technical field of water use prediction scheduling, and comprises the following steps: for the user cluster topology of a target city, a vibration cascade relationship is established by dividing multiple water use units; a behavior phase transition point is determined, a water use behavior portrait sequence is constructed, and a scheduling decision maker is developed; a water use task is received by a city water system, the scheduling decision maker is triggered, and a prediction scheduling strategy is determined; and according to the city water system, the prediction scheduling strategy is executed and managed. The application solves the technical problems that the existing city water use unit division is extensive, the scheduling strategy lacks precision and real-time performance, and the water use prediction error is large and the scheduling management efficiency is low, achieves the technical effects of realizing water use unit division and formulating a prediction scheduling strategy, and improves the city water use prediction precision and the scheduling management efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water prediction scheduling, in particular to a city water prediction scheduling method based on user behavior modeling and a storage device. BACKGROUND

[0002] With the acceleration of urbanization, the scale of urban water use is expanding, and the demand for water use is diversifying and dynamic. The existing city water prediction scheduling technology relies on traditional statistical models or simple time series analysis, which has the problems of rough division of water use units, inability to accurately depict user behavior differences and dynamic influence of environmental factors, and difficulty in effectively capturing the phase transition point of water use behavior. At the same time, the scheduling decision lacks deep fusion analysis of real-time sensor data and behavior patterns, resulting in low water use prediction accuracy and lagging scheduling strategy, which cannot meet the needs of fine urban water management, causing waste of water resources and imbalance between supply and demand.

[0003] The existing technology has the technical problems of rough division of city water use units and lack of accuracy and real-time of scheduling strategy, resulting in large water use prediction error and low scheduling management efficiency. SUMMARY

[0004] The present application provides a city water prediction scheduling method based on user behavior modeling and a storage device, which is used to solve the technical problems of rough division of city water use units and lack of accuracy and real-time of scheduling strategy in the prior art, resulting in large water use prediction error and low scheduling management efficiency.

[0005] In view of the above problems, the present application provides a city water prediction scheduling method based on user behavior modeling and a storage device.

[0006] In a first aspect of the present application, a city water prediction scheduling method based on user behavior modeling is provided, the method comprising:

[0007] For the user cluster topology of the target city, a vibration cascade relationship is established by dividing multiple levels of water use units, wherein the vibration cascade relationship is a voiceprint sensing topology corresponding to the water use unit network at different cluster granularities; the behavior phase transition points are determined and the water use behavior portrait sequence is constructed according to the time-sharing behavior phase transition and the external environmental behavior phase transition, and the voiceprint fingerprint library is combined to develop a scheduling decision maker for the city water system, wherein the voiceprint fingerprint library is constructed by pipeline vibration characteristics-water flow behavior; the city water system receives a water use task, interprets the task and makes a decision based on the vibration cascade relationship, performs directional sensing at the cluster granularity, returns and triggers the scheduling decision maker, performs signal constant and abnormality binary classification and parallel judgment of behavior phase state, constructs a behavior state space and makes long and short time decisions to determine the prediction scheduling strategy; and according to the city water system, the prediction scheduling strategy is executed and managed.

[0008] In a second aspect of the present application, a storage device is provided, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the city water prediction and scheduling method based on user behavior modeling.

[0009] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0010] For the user cluster topology of the target city, a vibration cascade relationship is established by dividing multiple levels of water use units; the behavior phase change points are determined and the water use behavior portrait sequence is constructed based on the time-sharing behavior phase change and the external environment behavior phase change, and the city water system development scheduling decision maker is constructed in combination with the voiceprint fingerprint library; the city water system receives the water use task, interprets the task and makes decisions based on the vibration cascade relationship, performs directional sensing at the task cluster granularity, returns and triggers the scheduling decision maker, performs parallel judgment of signal constant and variable classification and behavior phase, constructs the behavior state space and makes long and short time decisions to determine the prediction and scheduling strategy; and the prediction and scheduling strategy is executed according to the city water system. The technical effects of realizing water use unit division and formulating prediction and scheduling strategy are achieved, and the city water prediction accuracy and scheduling management efficiency are improved. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0012] Figure 1 A flowchart of the city water prediction and scheduling method based on user behavior modeling provided by the embodiments of the present application is shown.

[0013] Figure 2 A structural diagram of the storage device provided by the embodiments of the present application is shown.

[0014] Explanation of reference signs: input device 201, processor 202, memory 203, output device 204. DETAILED DESCRIPTION

[0015] The present application provides a city water prediction and scheduling method based on user behavior modeling and a storage device, which are used to solve the technical problems of extensive division of city water use units, lack of accuracy and real-time of scheduling strategy, large water prediction error and low scheduling management efficiency in the prior art.

[0016] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0017] Embodiment one, as shown in the present application provides a city water prediction scheduling method based on user behavior modeling, the method comprises: Figure 1

[0018] Step S100: For the user cluster topology of the target city, a vibration cascade relationship is established by dividing multiple levels of water use units, wherein the vibration cascade relationship is a voiceprint sensing topology corresponding to the water use unit network under different cluster granularities.

[0019] Specifically, for the user cluster topology structure of the target city, multiple levels of water use units are divided according to the cluster granularity from small to large, for example, a first level of water use units with a household as the smallest cluster granularity, a second level of water use units with a building as the granularity, and so on, until an N level of water use units with a region as the granularity. For each level of water use units, a voiceprint sensing array distributed in the city water pipe network is matched to establish a cascade relationship between the water use unit and the sensing array. For example, a first level of water use units correspond to a first level of sensing arrays deployed in the user's home, a second level of water use units correspond to a second level of sensing arrays deployed in the building's total water pipe, forming a multi-level vibration cascade relationship from the household to the region. The vibration cascade relationship is essentially a mapping of the water use unit network and the voiceprint sensing topology under different cluster granularities, which can realize the vibration signal perception and transmission from the micro user water behavior to the macro regional water trend.

[0020] Step S200: Determine the behavior phase change point and construct the water use behavior image sequence by the time-sharing behavior phase change and the external environment behavior phase change, and develop a scheduling decision maker in the city water system in combination with a voiceprint fingerprint library, wherein the voiceprint fingerprint library is constructed by pipeline vibration characteristics-water flow behavior.

[0021] ​Specifically, first, the phase change points of water use behavior are mined from the time dimension and the environmental dimension. By analyzing the historical water use records of the target city, the periodic phase change points of user water use behavior (such as early morning washing water use, evening shower water use, etc.) are mined according to the day / week / month cycle, and the water use behavior phase change points caused by external environmental factors such as seasonal temperature changes and sudden rainfall (such as a sharp increase in water use during the summer high temperature period, changes in water supply pipe network pressure during heavy rain) are considered. The two types of phase change points are fitted to form a complete set of behavior phase change points. Based on the phase change points, a water use behavior portrait sequence is constructed, such as distinguishing between “morning water use portrait” and “weekday water use portrait”, and combining the voiceprint fingerprint library constructed by the correlation between pipe vibration characteristics and water flow behavior (such as vibration frequency characteristics corresponding to different water flow rates). Through the generation of adversarial training, signal discriminators (to distinguish between normal / abnormal water use signals) and behavior phase state discriminators (to identify water use behavior phase changes) are determined, and finally a scheduling decision maker is integrated in the city water system for subsequent signal analysis and scheduling decisions.

[0022] Step S300: Receive water use tasks through the city water system, perform task interpretation and decision-making based on vibration cascade relationship, perform directional sensing at task cluster granularity, return and trigger the scheduling decision maker, perform signal normal / abnormal classification and behavior phase state parallel judgment, construct behavior state space and make long / short time decision, and determine the predicted scheduling strategy.

[0023] Specifically, when the city water system receives a water use task (such as global water supply scheduling, local pipeline maintenance, etc.), the task type is first interpreted and the best cluster granularity of the target water use unit is located (such as local task corresponding to building-level water use unit, global task corresponding to regional-level water use unit). Based on the vibration cascade relationship, the sensing array corresponding to the target cascade is matched, the pipe water ripple vibration signal is directionally collected and returned. The signal discriminator in the scheduling decision maker classifies the vibration signal as normal / abnormal, and the phase discriminator is triggered based on the task scenario, and the current water use behavior portrait is determined (such as whether it is in the early morning water use phase). The signal classification result and the behavior portrait are fused to construct a behavior state space, the water use trend in the state space is analyzed through algorithms such as long short-term memory network (LSTM), and a predicted scheduling strategy covering short-term scheduling (such as real-time water pressure adjustment) and long-term planning (such as regional water allocation) is developed.

[0024] Step S400: According to the city water system, execute the issuing and execution management of the predicted scheduling strategy.

[0025] Specifically, based on the control architecture of the urban water system, the determined predictive scheduling strategy is issued to each level of execution unit, such as sending water pressure adjustment instructions to regional water supply stations and sending valve opening and closing control signals to local pipe networks. In this process, the multi-cluster division mechanism enables different tasks to dynamically adapt to the corresponding unit cluster architecture and sensing topology (such as automatically switching emergency repair tasks to the sensing topology of the building level in the fault area), and to switch the directional behavior image (such as switching from the regular water use image to the anti-freezing emergency image) in real time based on phase elements (such as sudden changes in water consumption and sudden drops in environmental temperature). By dynamically reconstructing the sensing topology and updating the phase image, the behavior state space is continuously updated to ensure that the water use prediction and scheduling decision is always based on the most accurate water use state, thereby achieving intelligent scheduling and dynamic management of urban water use.

[0026] In one possible implementation manner, the step S100 further includes:

[0027] Step S110: deploying a voiceprint sensing array, wherein the voiceprint sensing array is distributedly deployed in the urban water pipe network.

[0028] Step S120: obtaining a user cluster topology of the target city, and establishing a vibration cascade relationship between the user cluster topology and the voiceprint sensing array.

[0029] Specifically, the voiceprint sensing array is distributedly deployed at key nodes and user ends of the urban water pipe network, specifically including installing vibration sensors at user household water terminals (such as faucets, water meters), building main water pipes, regional water supply trunk pipes and the like, to form a multi-level sensing network covering from user terminals to the main line of the urban water supply pipe network. The voiceprint sensing array is connected to the data center of the urban water system through wired or wireless communication, and real-time vibration signals generated by water flow in the pipeline are collected, thereby providing original data support for subsequent water use behavior analysis and scheduling decision.

[0030] The user cluster topology structure of the target city is obtained through the geographic information system (GIS) and the water use management database of the urban water system, and the topology structure reflects the distribution relationship of users at levels such as regions, buildings, units, etc. Based on different cluster granularities (such as households, buildings, communities, regions) of the user cluster topology, each level of the voiceprint sensing array is matched with the corresponding water use unit to establish a vibration cascade relationship. For example, a first-level sensing array deployed in a user's home is cascaded with a household-level water use unit, and a second-level sensing array of a building main water pipe is cascaded with a building-level water use unit, so that the water use vibration signals of each water use unit can be collected in real time by the corresponding sensing array, forming a vibration signal transmission link from micro user water use behavior to macro regional water use condition, thereby laying a foundation for subsequent task interpretation and directional sensing based on the vibration cascade relationship.

[0031] In one possible implementation manner, the step S100 further includes:

[0032] Step S130: dividing the multi-level water use units at the cluster granularity of the user cluster topology.

[0033] Step S140: determining a primary water use unit for the multi-level water use units, wherein the primary water use unit is the smallest cluster granularity, taking a household as a water use unit.

[0034] Step S150: determining a primary sensing array in the voiceprint sensing array for the primary water use unit, wherein the primary sensing array is deployed in the primary water use unit.

[0035] Step S160: establishing a cascade of the primary water use unit and the primary sensing array as a first vibration cascade relationship.

[0036] Specifically, the multi-level water use units are divided at the cluster granularity of the user cluster topology, specifically referring to hierarchical division of water use units according to the spatial distribution and hierarchical characteristics of the target city user cluster topology, in the order of cluster granularity from small to large. First, taking a household as the most basic bottom unit, it is aggregated into buildings, communities, regions and other different granularity water use units in turn, forming a multi-level architecture from micro users to macro regions. During the division process, combined with the physical topology structure of the city water supply network and the aggregation characteristics of user water behavior, it is ensured that each level of water use unit can independently reflect the water use characteristics of a specific cluster, and can form a water use trend transmission chain from the bottom to the top level on the whole, laying a foundation for subsequent establishment of vibration cascade relationship and realization of water use monitoring and dispatching decision at different granularity.

[0037] Determining a primary water use unit for the multi-level water use units specifically refers to defining a household as a primary water use unit in the multi-level water use unit system based on user cluster topology division. This process takes a family or a single user as the basic unit, combines the user profile and geographic distribution information in the city water management system, defines the boundary of each primary water use unit (such as specific residential number, physical location corresponding to water use account), and associates its water use attributes (such as water meter number, water use equipment type, etc.). The primary water use unit, as the basis of the entire water use unit hierarchical structure, can be aggregated into larger granularity water use units such as buildings and communities, and can be refined to the behavior monitoring of different water use terminals in the household, ensuring that the whole chain data collection and analysis from micro user water behavior to macro regional dispatching has logical coherence and traceability.

[0038] The first-level sensing array in the acoustic fingerprint sensing array is determined for the first-level water unit, specifically, in the acoustic fingerprint sensing array distributed in the urban water pipe network, a combination of sensing devices directly serving the first-level water unit is located and selected. The first-level water unit takes a household as the minimum cluster granularity, so the first-level sensing array is usually deployed near the water terminal nodes in the user's home, such as the household water pipe, water meter, faucet, water heater and other devices, and the vibration signals generated during single household water use are collected in real time by installing vibration sensors, pressure sensors and other devices. These dispersed sensors form an array based on wireless or wired communication protocols, forming full-coverage sensing of the water use behavior of the first-level water unit, ensuring that the vibration characteristics such as frequency and amplitude of water flow during indoor water use can be accurately captured, providing a data collection basis for subsequent establishment of the vibration cascade relationship between the first-level water unit and the sensing array.

[0039] The cascade relationship between the first-level water unit and the first-level sensing array is established as the first vibration cascade relationship, and the first-level water unit in units of households is mapped and associated with the first-level sensing array deployed in the user's home in terms of physical location and signal transmission. Through the geographic information system (GIS) and sensing network topology data of the urban water system, the water unit of a certain household (such as a specific residential house number) is corresponded to the vibration sensor combination (i.e. the first-level sensing array) deployed at the positions such as the water meter and faucet in the house. When the household has a water use behavior, the first-level sensing array collects the pipeline vibration signals in real time and transmits them back to the urban water system through the communication link, forming a direct cascade link of household water use behavior-vibration signal sensing. The first vibration cascade relationship constitutes the bottom layer of the multi-level vibration cascade system, ensuring that subsequent aggregation can be based on this to form the cascade relationship between the water unit and the sensing array of larger granularity such as the building and the region, and to realize signal transmission and collaborative monitoring from micro user water use behavior to macro regional water use state.

[0040] In one possible implementation manner, step S130 further includes:

[0041] Step S131: traversing the multi-level water unit, performing matching and cascade based on the acoustic fingerprint sensing array to determine the Nth vibration cascade relationship, wherein the Nth vibration cascade relationship takes the water unit with the maximum cluster granularity.

[0042] Step S132: adding the first vibration cascade relationship to the Nth vibration cascade relationship into the vibration cascade relationship.

[0043] Specifically, after completing the cascade of the first water unit (with a house as the minimum cluster granularity) and the first sensor array, all divided multi-level water units are traversed in order of cluster granularity from small to large (such as a house, a building, a community, a region, etc.), and for each level of water unit, a corresponding level of sensor combination is matched in the distributed deployment of the acoustic fingerprint sensor array. For example, the second water unit (building level) corresponds to the second sensor array deployed on the building main water pipe, the third water unit (community level) corresponds to the third sensor array deployed on the community water supply main pipe, and so on, until the N-level water unit (such as a specific regional city) matches the N-level sensor array of the regional water supply pipeline. In this way, the vibration cascade relationship between each level of water unit and the corresponding sensor array is established, and the Nth vibration cascade relationship is the mapping of the maximum cluster granularity water unit and the regional level sensor array, forming a full-level vibration signal perception link from a micro single house to a macro region, ensuring that the city water system can realize the cascade collection and transmission of vibration signals based on water units of different granularities.

[0044] The first vibration cascade relationship to the Nth vibration cascade relationship is added to the vibration cascade relationship, and the established first vibration cascade relationship corresponding to the minimum cluster granularity house (the mapping of the first water unit and the first sensor array) to the Nth vibration cascade relationship corresponding to the maximum cluster granularity (such as a region) (the mapping of the N-level water unit and the N-level sensor array) is integrated into the vibration cascade relationship database of the city water system in order of level. This operation forms a complete multi-level vibration cascade system, for example, including house-level sensor array, building-level sensor array, community-level sensor array, and other mapping relationships at each level, so that the cascade relationship between water units of different cluster granularities and acoustic fingerprint sensor arrays is recorded and managed by the system. Through the integrated vibration cascade relationship, the city water system can call the corresponding level of cascade relationship to realize directional sensing according to the water task demand (such as local pipeline monitoring or global water supply scheduling), and provide multi-level infrastructure support for subsequent task interpretation, signal collection and scheduling decisions based on the vibration cascade relationship.

[0045] In one possible implementation, step S200 further includes:

[0046] Step S210: generating an adversarial training with normal water use and abnormal water use to determine a signal discriminator.

[0047] Step S220: generating an adversarial training with a behavior phase point and the behavior image sequence to determine a behavior phase discriminator.

[0048] Step S230: constructing the scheduling decision maker according to the signal discriminator and the phase discriminator, wherein the acoustic fingerprint library is built-in in the scheduling decision maker.

[0049] Specifically, a generative adversarial training method is used to determine the signal discriminator, employing both normal and abnormal water usage scenarios (such as water ripples and flow vibrations caused by user water usage, or vibrations caused by pipeline malfunctions). This involves using pipeline vibration signals generated by daily user water usage (such as washing and showering) and vibration signals caused by non-user water usage behaviors such as pipeline leaks and equipment failures as training samples, trained using a generative adversarial network (GAN) architecture. The generator simulates various water usage vibration signals, while the discriminator performs a dual judgment on the input signal: distinguishing between real-world data collection and generator simulation, and identifying whether the signal belongs to normal user water usage vibration or abnormal pipeline vibration. Through iterative adversarial game between the generator and discriminator, the discriminator gradually learns and masters the differences between normal and abnormal vibration signals in terms of frequency, amplitude, and waveform, ultimately determining a signal discriminator with accurate discrimination capabilities, capable of performing binary classification of normal / abnormal water usage on real-time transmitted target vibration signals.

[0050] Using behavioral phase transition points and the behavioral profile sequences, a generative adversarial training (GAN) framework is employed to determine the behavioral phase discriminator. Based on the identified behavioral phase transition points (including first-class transition points obtained through periodic time-sharing mining, such as morning peak water usage periods and evening peak water usage periods, and second-class transition points mined by combining seasonal environmental evolution and sudden environmental changes, such as surges in water consumption during summer high temperatures and sudden changes in pipeline pressure during rainstorms) and constructed water usage behavioral profile sequences (such as morning water usage profiles, weekday water usage profiles, and emergency water usage profiles), the GAN framework is used to determine the behavioral phase discriminator. The generator simulates and generates water usage behavioral feature sequences under different behavioral phases. These feature sequences contain various parameters related to behavioral phase transition points, such as water consumption changes, water usage time distribution, and the influence of environmental factors. The discriminator analyzes the input behavioral feature sequences to determine their corresponding specific behavioral phase (e.g., whether they are in the phase corresponding to a certain type of transition point). During training, the generator continuously optimizes to generate behavioral feature sequences that are closer to the real situation, while the discriminator continuously improves its ability to identify different phase features. Through the adversarial game between the two, the discriminator is eventually able to accurately identify the current phase of water use behavior based on the specific identification elements of the behavioral phase transition point (i.e., phase transition elements), thereby achieving dynamic switching and accurate judgment of the water use behavior profile and determining the behavioral phase discriminator.

[0051] A scheduling decision-maker is constructed based on a signal discriminator and a phase discriminator, with an acoustic fingerprint database built into this decision-maker. The signal discriminator (used to distinguish between normal and abnormal water use signals) and the behavioral phase discriminator (used to identify the current water use behavior phase), determined through adversarial training, are integrated into the urban water system's scheduling decision-maker as core algorithm modules. Simultaneously, an acoustic fingerprint database (storing feature templates of vibration signals under different water use scenarios, such as frequency and amplitude features corresponding to user washing, pipe leaks, etc.), constructed by associating pipeline vibration characteristics with water flow behavior, is built into the scheduling decision-maker as a benchmark database for signal discrimination. The scheduling decision-maker performs feature matching on the real-time transmitted vibration signals based on the acoustic fingerprint database, combining the normal / abnormal classification results of the signal discriminator with the phase determination of the behavioral phase discriminator (e.g., whether it is the morning peak water use phase), forming a collaborative decision-making mechanism of signal state-behavioral phase.

[0052] In one possible implementation, step S220 further includes:

[0053] Step S221: Call the water usage records of the target city, set the user water usage behavior cycle and perform periodic time-sharing mining to determine a type of behavior phase transition point.

[0054] Step S222: Based on seasonal environmental evolution and sudden environmental changes, identify the phase transition points of the two types of behaviors.

[0055] Step S223: Fit the phase transition points of the first type of behavior and the phase transition points of the second type of behavior to determine the phase transition points of the behavior.

[0056] Specifically, the process involves accessing water usage records from the target city, establishing user water usage behavior cycles, and performing periodic time-based data mining to identify a type of behavioral phase transition point. Historical water usage records for the target city are obtained from the city's water system database, covering data such as user water consumption and appliance usage at different time periods. This data is used to establish daily, weekly, and monthly user water usage behavior cycles. Water usage data within each cycle is then further subdivided and mined along the time dimension, for example, by statistically analyzing water consumption fluctuations hourly. This identifies recurring and regular peak and off-peak water usage periods, such as the peak water usage for washing and grooming between 6 and 8 AM and the peak water usage for showering between 7 and 9 PM. These key time points of recurring water usage behavior changes within fixed cycles constitute a type of behavioral phase transition point, reflecting the periodic patterns in the temporal distribution of users' daily water usage behavior. These provide key time-based nodes for subsequently constructing a sequence of water usage behavior profiles.

[0057] The two types of behavior phase change points are mined from seasonal environmental evolution and sudden environmental changes. From the influence of external environment on water use behavior, two dimensions are identified to determine the mutation nodes of water use behavior. On the one hand, the influence of seasonal environmental evolution on water use is analyzed, such as the sharp increase of water use in summer (June-August) due to residents' showering and greening irrigation, and the adjustment of water use mode in winter (December-February) due to water pipe anti-freezing, so as to determine the seasonal phase change point. On the other hand, attention is paid to sudden environmental changes, such as rainstorm weather causing water supply pipe network pressure fluctuation, and users' concentrated water storage after sudden water stop event. By comparing the water use data before and after the sudden environmental event, the water use behavior mutation point caused by sudden environmental change is located. This type of phase change point reflects the interference of non-periodic environmental factors on water use behavior, and together with the first type of behavior phase change point (periodic time phase change point) forms a complete set of behavior phase change points, which provides key basis for subsequent water use behavior portrait switching and phase state identification in the environmental dimension.

[0058] The first type of behavior phase change point and the second type of behavior phase change point are fitted to determine the behavior phase change point. Specifically, the following implementation means are used: using the regression fitting algorithm (such as support vector regression) in machine learning, inputting the time series data of the first type of behavior phase change point (such as the water use amount and time stamp corresponding to the daily / weekly / monthly periodic water use peak period) and the environmental correlation data of the second type of behavior phase change point (such as the water use mutation value corresponding to the seasonal temperature, sudden rainfall and other environmental factors) into a multi-dimensional feature space, and constructing a time-water amount-environmental factor coupling model. The feature parameters (such as periodic weight and environmental influence coefficient) of the two types of phase change points are standardized by feature engineering, and the model hyperparameters are optimized by grid search, so that the model can identify the phase change threshold that meets the periodicity and environmental disturbance at the same time (such as when the summer temperature exceeds 30°C and is at 19-21 o'clock every day, the water use amount increases by 20% to determine the phase change point). Finally, the fused behavior phase change point set is output by the model to realize the accurate positioning of the user water use behavior turning point.

[0059] In one possible implementation manner, step S220 further includes:

[0060] Step S224: determining a phase change element for the behavior phase change point, wherein the phase change element is a specific recognition element of behavior switching.

[0061] Step S225: performing behavior phase state recognition and identification and water use behavior portrait switching training according to the phase change element, and constructing the phase state discriminator.

[0062] Specifically, for the aforementioned behavioral phase transition points, phase transition elements are determined. These phase transition elements are specific identification elements for behavioral switching. From the determined behavioral phase transition points, key characteristic parameters that can characterize the switching of water use behavior are extracted. These phase transition elements include features in the time dimension, such as the specific time period and periodicity of the phase transition point; features in the environmental dimension, such as the threshold values ​​for changes in environmental factors such as quarterly temperature and rainfall; and features in the water use behavior dimension, such as the fluctuation range of water consumption and changes in the combination of water-using equipment. For example, for the phase transition point of peak evening water use caused by high summer temperatures, its phase transition elements may include the specific time period (7-9 pm), the environmental temperature threshold (≥30℃), and the increase in water consumption compared to the daily average (≥20%). These elements together constitute specific indicators for identifying the switching of water use behavior from the daily state to the high-load state in summer, providing a key basis for subsequent identification of behavioral phases and the switching of water use behavior profiles.

[0063] Based on the phase transition factors, a generative adversarial algorithm is used to train the switching between behavioral phase identification and water use behavior profiles, constructing the phase discriminator. The model is trained using the generative adversarial network (GAN) framework, with the phase transition factors as input. The generator is responsible for generating simulated behavioral phase feature sequences, striving to approximate real water use behavior phases; the discriminator is used to determine whether the input feature sequence comes from real data or is generated by the generator, and simultaneously identifies its corresponding specific behavioral phase. During training, the generator continuously optimizes to generate more realistic phase features, while the discriminator enhances its ability to distinguish between real and generated phases. Through the adversarial game between the two, the discriminator can accurately identify different behavioral phases. For example, when inputting phase transition factors such as high summer temperatures, evening hours, and surges in water consumption, the discriminator can accurately identify the high water use phase in summer evenings and trigger the corresponding water use behavior profile switching, ultimately constructing a highly robust phase discriminator that achieves dynamic identification and profile switching of water use behavior phases.

[0064] In one possible implementation, step S300 further includes:

[0065] Step S310: Determine the water usage task, wherein the water usage task is any combination of global task-local task and actively initiated task-passively generated task.

[0066] Step S320: By interpreting the water use task, locate the target water use unit, wherein the target water use unit is the optimal cluster granularity for task matching.

[0067] Step S330: Based on the target water-using unit, match and determine the target cascade relationship in the vibration cascade relationship.

[0068] Specifically, determining the water use task needs to clarify its combination attributes in the spatial dimension and the trigger mechanism dimension. Among them, the spatial dimension covers global tasks and local tasks. Global tasks are water management tasks facing the entire city, such as city-wide water resource allocation, city-wide water supply network pressure regulation, etc. Local tasks focus on local areas of the city, such as pipeline maintenance in a community, water monitoring in a specific industrial park, etc. The trigger mechanism dimension includes tasks initiated by the initiative and tasks generated by the initiative. The task initiated by the initiative is a task initiated by the initiative according to the pre-designed plan or target, such as the scheduling plan formulated before the arrival of the summer water peak, the annual water supply facility maintenance plan, etc. The task generated by the initiative is a task triggered by a sudden situation or external event, such as emergency repair after a sudden leakage of the water supply network, troubleshooting after a user reports water anomalies, etc. The water use task can be any combination of the two dimensions, such as global-initiative task, local-generated task, etc. By clarifying the combination attributes of the task, a clear logical basis is provided for subsequent task interpretation and positioning of the target water unit.

[0069] By interpreting the water use task, the target water unit is located. The target water unit is the best cluster granularity that fits the task. After clarifying the combination attributes of the water use task (such as global-initiative, local-generated, etc.), the specific requirements of the task are analyzed, such as the spatial range, time requirements, precision indicators, etc. Then, in the multi-level water units (such as households, buildings, communities, regions, etc.) that have been divided, the cluster granularity that best meets the task requirements is matched. For example, for the local-generated task of detecting a leakage in a community water supply network, it is necessary to accurately locate to a specific building or household, so the best cluster granularity is the building level or household level, and the target water unit is the specific building or household in the community. For the global-initiative task of summer city water peak scheduling, it is necessary to allocate water resources from the city-wide level, so the best cluster granularity is the city level or regional level, and the target water unit is the entire city or a specific water supply area. This process ensures that the subsequent sensing and decision-making can be executed at the most appropriate scale through precise matching of task requirements and water unit granularity, thereby improving the efficiency and accuracy of task execution.

[0070] According to the target water unit, a target cascade relationship is determined in the vibration cascade relationship, and after locating the target water unit that fits the water task (such as a cluster granularity of a house, a building, a community, or a region), the cascade mapping relationship corresponding to the target water unit is retrieved from a pre-built vibration cascade relationship database. The vibration cascade relationship records the correspondence between the water units at each level and the distributed deployment of the acoustic fingerprint sensing array. For example, when the target water unit is a community, the cascade relationship between the community-level water unit and the sensing array deployed on the community water supply main pipe needs to be matched. If the target water unit is a single house, the cascade relationship between the house-level water unit and the primary sensing array installed on the pipe in the house needs to be matched. Through this matching mechanism, the sensing path and signal acquisition range required for the task can be accurately determined, providing infrastructure support for subsequent calls to the target sensing array for directional vibration signal acquisition and triggering the scheduling decision maker to perform signal discrimination and phase state analysis, ensuring that the entire water prediction and scheduling process runs efficiently at the corresponding cluster granularity.

[0071] In one possible implementation, step S300 further includes:

[0072] Step S340: According to the target cascade relationship, a target sensing array is called to perform pipe water ripple sensing to determine a target vibration signal.

[0073] Step S350: The target vibration signal is returned to trigger the signal discriminator to distinguish between normal water signals and abnormal water signals.

[0074] Step S360: Based on the task scenario of the water task, a cascade determination based on the phase state discriminator is triggered to determine a target behavior portrait.

[0075] Step S370: According to the normal water signal and the target behavior portrait, the behavior state space is constructed.

[0076] Step S380: The behavior state space is used as a reference for water prediction and scheduling analysis and management.

[0077] Specifically, after the mapping relationship between the target water unit and the acoustic fingerprint sensing array (i.e., the target cascade relationship) is determined, the target sensing array deployed at the corresponding location is called, such as a house-level sensing array, a building-level sensing array, or a regional-level sensing array. These distributed acoustic fingerprint sensing arrays deployed in the urban water pipe network can sense the water flow vibration in the pipe in real time, convert the water ripple vibration into corresponding electrical signals by capturing the vibration frequency, amplitude, waveform, and other characteristics of the pipe wall. Then, the electrical signals are preprocessed, such as filtering and amplifying, to remove noise interference, so as to determine the target vibration signal that can accurately represent the current water state, providing reliable raw data support for subsequent signal discrimination and water behavior analysis.

[0078] The target vibration signal collected and pre-processed by the target sensor array is returned to the dispatch decision maker through the communication network of the urban water system, automatically activating the built-in signal discriminator. The signal discriminator is constructed based on the generative adversarial training (GAN) mechanism and has completed training using a large amount of historical data, and can accurately identify the vibration feature differences corresponding to normal water use scenarios (such as washing, showering, etc.) and abnormal water use scenarios (such as pipeline leakage, valve failure, equipment anomaly, etc.). When the target vibration signal is input, the signal discriminator will compare and analyze it with the standard vibration templates (covering frequency, amplitude, waveform, and other characteristic parameters of different water use behaviors) pre-stored in the voiceprint fingerprint library, and perform normal / abnormal binary classification determination through pattern recognition algorithms. For example, if the signal characteristics match the high-frequency vibration template corresponding to pipeline leakage with a matching degree exceeding the preset threshold, it is determined as an abnormal water use signal; if it matches the low-frequency stable vibration characteristics of daily washing, it is determined as a normal water use signal, thereby providing key signal classification basis for subsequent water use scheduling and abnormal warning.

[0079] According to the specific scene of the water use task (such as global water use peak scheduling, local pipeline leakage detection, etc.), the cascade determination mechanism of the phase state discriminator is activated. The phase state discriminator will combine the determined phase change elements (such as time, environmental temperature, water consumption fluctuation amplitude, and other specific recognition elements of behavior switching) to determine the phase state of the current water use behavior at multiple levels. For example, in the global water use peak scheduling scene in summer high temperature, if it is detected that the time is at 19-21 o'clock, the environmental temperature exceeds 30°C, and the water consumption is increased by 20% compared with the daily average, the phase state discriminator will match the corresponding phase change elements in sequence through the cascade determination mechanism, and finally determine that the target behavior portrait is the summer evening high-load water use portrait. This cascade determination process relies on the pre-stored behavior phase state switching rules and water use behavior portrait sequence in the phase state discriminator, realizes the accurate mapping from the task scene to the specific behavior portrait, and provides the basis for the behavior pattern level for subsequent construction of behavior state space and development of prediction scheduling strategy.

[0080] According to the normal water use signal and the target behavior image, the behavior state space is constructed, the normal water use signal features (such as time domain and frequency domain feature parameters such as vibration frequency, amplitude, waveform, etc.) output by the signal discriminator are multi-dimensionally fused with the target behavior image (such as the morning washing image, the emergency water use image, etc.) determined by the phase discriminator, and the behavior state space is constructed in the time and space dimensions. The space takes time, water use amount, vibration feature, environmental factor, behavior phase, etc. as coordinate axes, each dimension corresponds to different feature parameters, for example, the time dimension is refined to hours, dates, the vibration feature dimension includes main frequency, energy distribution, and the environmental factor dimension covers temperature, humidity, etc. By associating and mapping the feature vector of the normal water use signal with the label of the target behavior image, a specific coordinate point or area is formed in the state space, and each point represents a specific water use behavior mode. For example, the water tap low-frequency vibration signal (normal water use signal) detected at 6-8 am is combined with the "morning washing image", and the corresponding state point is constructed in the behavior state space, which intuitively represents the typical water use behavior characteristics of the user in this period, and provides a structured data model for subsequent water use prediction and scheduling analysis based on the state space.

[0081] Taking the behavior state space as the reference, a long short-term memory neural network (LSTM) is used for water use prediction and scheduling analysis and management, the behavior state space data (including time sequence, vibration feature, behavior phase label, etc. Multi-dimensional features) constructed are used as the input of the LSTM model, and the processing ability of the LSTM model for long sequence dependence is used to learn the time periodicity and environmental sensitivity of the water use behavior. For example, the behavior state space data (such as vibration signal features, temperature parameters, water use amount fluctuations, etc.) of the high water use phase in the summer night in history (such as 19-21 o'clock) is input into the model, and the LSTM network is trained to capture the change mode of the water use amount in this phase. When predicting, the model generates a water use amount prediction sequence for a period of time in the future based on the current behavior state space point, combined with the future time stamp and environmental prediction data (such as weather forecast temperature), and then formulates a scheduling strategy: when it is predicted that a certain area will enter a water use peak, the prediction result output by the LSTM model will trigger the city water system to adjust the water supply pump station pressure, optimize the pipe network flow distribution, or start the standby water source, realize the closed-loop management from data modeling to scheduling execution, and improve the water use efficiency and the accuracy of system response.

[0082] Embodiment two, Figure 2 is a structural schematic diagram of a storage device provided by an embodiment of the present application, and shows a block diagram of an exemplary storage device suitable for implementing the embodiment of the present application. Figure 2The storage device shown is merely an example and should not bring any limitation to the function and use range of the embodiments of the present application. The storage device is in the form of a general computing device, and its components can include but are not limited to an input device 201, a processor 202, a memory 203 and an output device 204. The processor 202 can be one or more; the processor 202 performs various function applications and data processing of the computer device by running software programs, instructions and modules stored in the memory 203, that is, implements the above-mentioned urban water prediction and scheduling method based on user behavior modeling.

[0083] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0084] The above is only the preferred embodiment of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0085] The present application and the drawings are only exemplary descriptions of the present application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. A method for predicting and scheduling urban water use based on user behavior modeling, characterized in that, The method includes: For the user cluster topology of the target city, a vibration cascade relationship is established by dividing the water use units into multiple levels. The vibration cascade relationship is the acoustic fingerprint sensing topology corresponding to the water use unit network at different cluster granularities. By combining time-sharing behavioral phase transitions with external environmental behavioral phase transitions, behavioral phase transition points are determined and a sequence of water use behavior profiles is constructed. Combined with an acoustic fingerprint database, a scheduling decision-making device is developed for urban water systems. The acoustic fingerprint database is constructed from pipeline vibration characteristics and water flow behavior. The system receives water usage tasks through the urban water system, interprets the tasks and makes decisions based on vibration cascade relationships, performs directional sensing at the task cluster granularity, transmits back and triggers the scheduling decision-maker, performs parallel determination of signal constant and different binary classification and behavioral phase state, constructs behavioral state space and makes long and short time decisions to determine the predictive scheduling strategy. Based on the urban water system, the predicted scheduling strategy is issued and executed. Develop a scheduling decision-maker, including: Generative adversarial training was conducted using normal water usage and abnormal water usage to determine the signal discriminator; Generative adversarial training is performed using the behavior phase transition points and the behavior profile sequence to determine the behavior phase discriminator; The scheduling decision-maker is constructed based on the signal discriminator and the phase discriminator, wherein the voiceprint fingerprint database is built into the scheduling decision-maker.

2. The urban water use prediction and scheduling method based on user behavior modeling as described in claim 1, characterized in that, Deploy a voiceprint sensor array, wherein the voiceprint sensor array is distributed and deployed in the urban water supply network; Obtain the user cluster topology of the target city and establish the vibration cascade relationship between the user cluster topology and the voiceprint sensor array.

3. The urban water use prediction and scheduling method based on user behavior modeling as described in claim 2, characterized in that, By dividing the water use into multiple levels, a vibration cascade relationship is established, including: The user cluster topology is divided into multi-level water use units at the cluster granularity. For the aforementioned multi-level water use units, a first-level water use unit is determined, wherein the first-level water use unit is the smallest cluster granularity, with the household as the water use unit; For the primary water-using unit, a primary sensor array is determined in the voiceprint sensor array, wherein the primary sensor array is deployed in the primary water-using unit; Establish a cascade between the primary water-using unit and the primary sensing array as the first vibration cascade relationship.

4. The urban water use prediction and scheduling method based on user behavior modeling as described in claim 3, characterized in that, Traverse the multi-level water use units, perform matching and cascading based on the acoustic signature sensor array, and determine the Nth vibration cascading relationship, wherein the Nth vibration cascading relationship uses the largest cluster size as the water use unit; The first vibration cascade relationship up to the Nth vibration cascade relationship are added to the vibration cascade relationship.

5. The urban water use prediction and scheduling method based on user behavior modeling as described in claim 1, characterized in that, Determining the phase transition point of behavior includes: Access the water usage records of the target city, set the user water usage behavior cycle, and perform periodic time-sharing mining to identify a type of behavior phase transition point; By examining seasonal environmental evolution and sudden environmental changes, we can identify the phase transition points of two types of behaviors. By fitting the phase transition points of the first type of behavior and the phase transition points of the second type of behavior, the phase transition point of the behavior is determined.

6. The urban water use prediction and scheduling method based on user behavior modeling as described in claim 5, characterized in that, Determine the behavior phase discriminator, including: For the aforementioned behavior phase transition point, phase transition elements are determined, wherein the phase transition elements are specific identification elements for behavior switching; Based on the phase transition elements, the phase discriminator is constructed by switching between behavioral phase identification and water use behavior profiling.

7. The urban water use prediction and scheduling method based on user behavior modeling as described in claim 1, characterized in that, Performing task interpretation and decision-making based on vibration cascade relationships includes: Determine the water usage task, wherein the water usage task is any combination of global task-local task and actively initiated task-passively generated task; By interpreting the water usage task, the target water usage unit is located, wherein the target water usage unit is the optimal cluster granularity for task matching; Based on the target water-using unit, the target cascade relationship is determined by matching in the vibration cascade relationship.

8. The urban water use forecasting and scheduling method based on user behavior modeling as described in claim 7, characterized in that, The task cluster-level directional sensing is executed, and the feedback is triggered to the scheduling decision-maker to perform parallel determination of signal constant / dissimilar binary classification and behavioral phase state, including: Based on the target cascade relationship, the target sensor array is invoked to detect water ripples in the pipeline and determine the target vibration signal. The target vibration signal is transmitted back, triggering the signal discriminator to distinguish between normal water use signals and abnormal water use signals; Based on the task scenario of the water use task, a cascaded determination based on the phase discriminator is triggered to determine the target behavior profile; Based on the normal water usage signal and the target behavior profile, the behavior state space is constructed; Based on the aforementioned behavioral state space, water use prediction, scheduling analysis, and management are performed.

9. A storage device, characterized in that, The storage device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the urban water use prediction and scheduling method based on user behavior modeling as described in any one of claims 1 to 8.

Images