Monte Carlo simulation-based independent metering and partitioned pressure regulation control system for water supply pipe network

Abstract

The invention relates to the technical field of intelligent pressure regulation control, and particularly discloses a Monte Carlo simulation-based water supply network independent metering subarea pressure regulation control system which divides a water supply network into a plurality of independent metering subareas. Pressure data of pressure supply points and most unfavorable points in each independent metering subarea and flow data of water inlet and outlet points are collected as training data, one day is divided into three stages including a peak period, a valley period and a non-peak and non-valley period according to the change rule of the water consumption, different Monte Carlo models are established by using the training data in corresponding time intervals, and the water consumption is calculated. And calling the Monte Carlo model of the corresponding time interval to simulate the pressure regulation and control scheme of each pressure supply point according to the target pressure value of the most unfavorable point transmitted by the user. Therefore, the water supply pressure can be dynamically adjusted according to the water supply peak and valley, so that the water supply pressure better conforms to the water use habit of a user, the loss caused by overlarge pressure of a water supply pipeline is reduced, and water supply disputes caused by insufficient pressure are reduced.

Description

Technical Field

[0001] The present application relates to the field of intelligent pressure regulation control technology, and more specifically, to an independent metering and zoning pressure regulation control system for a water supply network based on Monte Carlo simulation. Background Art

[0002] Water supply network pressure regulation technology plays a vital role in urban and regional water supply systems. Its core goal is to ensure the water supply needs and service quality of users at various water supply points by reasonably controlling the water pressure in the pipeline. The implementation of pressure regulation technology mainly relies on a variety of equipment and methods. By real-time monitoring of water pressure and flow in the pipeline network, the pressure regulating equipment can dynamically adjust the water supply pressure to maintain the stable operation of the system and improve water safety.

[0003] With the acceleration of urbanization and the depletion of water resources, water supply demand is becoming increasingly complex. This background has prompted the water supply industry to urgently need more efficient pressure regulation technology to cope with the severe water supply challenges in the future.

[0004] The current pressure regulation technology of water supply networks has some defects and deficiencies. Existing pressure regulation systems mostly use traditional mechanisms, relying on manual monitoring and regular manual adjustments, resulting in slow response times. When water demand increases sharply or emergencies occur (such as pipe ruptures or high-demand periods), the lack of a real-time data feedback mechanism makes it impossible for the pressure regulation system to quickly adapt to changes, often resulting in water supply interruptions or low water pressure for users at the end of the network, affecting water supply reliability. Another type of pressure regulation scheme based on hydraulic models requires a large amount of resources to be invested in the early stage to obtain geographic information data of the network, remote water volume, remote pressure, valves and other information, in order to conduct data analysis and further provide a pressure regulation scheme. This scheme requires large resource investment, high implementation difficulty, and few successful cases.

[0005] Therefore, an optimized independent metering and zoning pressure regulation control system for the water supply network based on Monte Carlo simulation is expected to optimize the testing process to ensure that the production rhythm will not be significantly slowed down. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a water supply network independent metering partition pressure regulation control system based on Monte Carlo simulation, which divides the water supply network into several independent metering partitions, collects the pressure data of the internal pressure supply point and the most unfavorable point of each independent metering partition, and the flow data of the water inlet and outlet points as training data, and divides a day into three stages: peak period, trough period and non-peak trough period according to the change law of water consumption, and uses the training data in the corresponding time interval to establish different Monte Carlo models, so that according to the target pressure value of the most unfavorable point input by the user, the Monte Carlo model of the corresponding time interval is called to simulate the pressure control scheme of each pressure supply point. In this way, the water supply pressure can be dynamically adjusted according to the peak and trough of water supply, so that it is more in line with the user's water use habits, thereby reducing the loss of water supply pipelines caused by excessive pressure and reducing water supply disputes caused by insufficient pressure.

[0007] Accordingly, according to one aspect of the present application, a water supply network independent metering zone pressure regulation control system based on Monte Carlo simulation is provided, which includes: A training data acquisition module, used to acquire training data of each independent metering partition collected by the sensor group; A Monte Carlo model training module, for dividing the training data of each independent metering partition into time intervals and then training the Monte Carlo model respectively to obtain a peak interval Monte Carlo model, a trough interval Monte Carlo model and a non-peak trough interval Monte Carlo model; A user setting information acquisition module is used to obtain the target pressure value and current time point of the most unfavorable point set by the user; A target Monte Carlo model retrieving module, used to determine the time interval to which the current time point belongs, and retrieve the Monte Carlo model of the corresponding time interval to obtain the target Monte Carlo model; A target pressure value generating module, used for inputting the target pressure value of the most unfavorable point set by the user into the trained target Monte Carlo model to obtain the target pressure value of each water plant and the target pressure value of each pump station; The pressure remote regulation module is used to determine the pressure remote regulation strategy based on the comparison between the target pressure value of each water plant and the real-time pressure value of each water plant and based on the comparison between the target pressure value of each pumping station and the real-time pressure value of each pumping station.

[0008] Compared with the prior art, the independent metering and zoning pressure regulation control system for water supply networks based on Monte Carlo simulation provided by this application can greatly reduce resource investment. It is not necessary to obtain all the geographic information data of the water supply networks, nor is it necessary to install a large number of remote transmission equipment. It only needs to obtain the pressure data of the pressure supply points in the zones, the pressure at the most unfavorable points, the flow rate at the water inlet points in the zones, and the flow rate at the water outlet points in the zones, and then data simulation can be performed to provide a pressure regulation plan. In addition, this application relies on the division method of independent metering zones, and can independently analyze and regulate the pressure of a single zone, making the implementation more flexible. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a flow chart of an independent metering and zoned pressure regulating control system for a water supply network based on Monte Carlo simulation according to an embodiment of the present application.

[0011] Figure 2 It is a block diagram of an independent metering and zoned pressure regulation control system for a water supply network based on Monte Carlo simulation according to an embodiment of the present application.

[0012] Figure 3 The block diagram is a Monte Carlo model training module in a water supply network independent metering and zoned pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application.

[0013] Figure 4 The present invention is a flow chart of Monte Carlo simulation in a water supply network independent metering zone pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application.

[0014] Figure 5 It is a block diagram of a model training unit in a water supply network independent metering and zoned pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application.

[0015] Figure 6 The block diagram is a target pressure value inference subunit in a water supply network independent metering zone pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application.

[0016] Figure 7 It is a block diagram of a compensatory enhanced interactive aggregation secondary subunit in an independent metering and zoned pressure regulation control system for a water supply network based on Monte Carlo simulation according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein. Example 1

[0018] In view of the above technical problems, in order to reduce the impact of climate factors on the pressure regulation scheme, this application establishes Monte Carlo models by month and time period, and implements pressure regulation control based on independent metering partitions. Specifically, first collect the pressure data of the internal pressure supply points of each independent metering partition, the flow data of the water inlet and outlet points, and the pressure data of the most unfavorable point as training data, and divide the day into three stages: peak period, trough period and non-peak trough period according to the change law of water consumption each month. Use the training data in the corresponding time interval to establish different Monte Carlo models to simulate the relationship between the water supply and each pressure value, so as to simulate the pressure value scheme of each pressure supply point based on the target pressure value of the most unfavorable point input by the user, using the Monte Carlo model of the corresponding time interval. In this way, the water supply pressure can be dynamically adjusted according to the peak and trough of water supply to make it more in line with the user's water use habits, thereby reducing the loss of water supply pipelines caused by excessive pressure and reducing water supply disputes caused by insufficient pressure.

[0019] Figure 1 It is a flow chart of an independent metering and zoned pressure regulating control system for a water supply network based on Monte Carlo simulation according to an embodiment of the present application. Figure 2 FIG. 1 is a block diagram of a water supply network independent metering and zoned pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the independent metering partition pressure regulating control system 100 of the water supply network based on Monte Carlo simulation includes: a training data acquisition module 110, which is used to acquire the training data of each independent metering partition collected by the sensor group; a Monte Carlo model training module 120, which is used to divide the training data of each independent metering partition into time intervals and then train the Monte Carlo model respectively to obtain a peak interval Monte Carlo model, a trough interval Monte Carlo model and a non-peak trough interval Monte Carlo model; a user setting information acquisition module 130, which is used to acquire the target pressure value and current time point of the most unfavorable point set by the user; a target ... The Monte Carlo model retrieval module 140 is used to determine the time interval to which the current time point belongs, and to retrieve the Monte Carlo model of the corresponding time interval to obtain the target Monte Carlo model; the target pressure value generation module 150 is used to input the target pressure value of the most unfavorable point set by the user into the trained target Monte Carlo model to obtain the target pressure value of each water plant and the target pressure value of each pump station; the pressure remote adjustment module 160 is used to determine the pressure remote adjustment strategy based on the comparison between the target pressure value of each water plant and the real-time pressure value of each water plant and based on the comparison between the target pressure value of each pump station and the real-time pressure value of each pump station.

[0020] In the above-mentioned independent metering partition pressure regulation control system 100 for water supply network based on Monte Carlo simulation, the training data acquisition module 110 is used to obtain the training data of each independent metering partition collected by the sensor group, wherein the training data includes the time queue of instantaneous flow data of the water inlet point, the time queue of instantaneous flow data of the water outlet point, the time queue of pressure data of each water plant, the time queue of pressure data of each pump station and the time queue of pressure data of the most unfavorable point, wherein the water supply network is divided into several independent metering partitions, and the most unfavorable point in the independent metering partition is the point with the lowest pressure in the independent metering partition. It should be understood that by dividing the water supply network into several independent metering partitions, it is possible to accurately monitor and effectively control the water supply pressure of each independent metering partition, thereby achieving a more detailed and accurate pressure regulation management goal. In addition, Monte Carlo simulation is a simulation method based on probability statistics, which simulates the actual physical process through random sampling and can effectively obtain an approximate solution to the problem. In the technical solution of the present application, in order to achieve accurate pressure regulation control using the Monte Carlo simulation method, first, by collecting the flow data of the water inlet and outlet points of each independent metering partition, the pressure data of each pressure supply point and the pressure data of the most unfavorable point, a training data set is constructed to train the Monte Carlo model, so that the Monte Carlo model can simulate the pressure distribution of each pressure supply point in the independent metering partition under a specific water use mode, thereby achieving accurate pressure regulation control of each independent metering partition of the water supply network.

[0021] In the above-mentioned independent metering partition pressure regulation control system 100 for water supply network based on Monte Carlo simulation, the Monte Carlo model training module 120 is used to divide the training data of each independent metering partition into time intervals and then train the Monte Carlo model respectively to obtain the peak interval Monte Carlo model, the valley interval Monte Carlo model and the non-peak valley interval Monte Carlo model. Specifically, Figure 3 FIG. 1 is a block diagram of a Monte Carlo model training module in a water supply network independent metering zone pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application. Figure 3 As shown, the Monte Carlo model training module 120 includes: a time interval determination unit 121, which is used to determine the time intervals of the peak period interval, the trough period interval and the non-peak and trough period interval of each independent metering partition based on the training data of each independent metering partition; a training data segmentation unit 122, which is used to segment the training data based on the peak period interval, the trough period interval and the non-peak and trough period interval of each day to obtain peak period interval training data, trough period interval training data and non-peak and trough period interval training data; a model training unit 123, which is used to determine the time intervals of the peak period interval, the trough period interval and the non-peak and trough period interval based on the peak period interval, the trough period interval and the non-peak and trough period interval of each day; The peak period interval training data, the trough period interval training data and the non-peak trough period interval training data are used to train the Monte Carlo model respectively to obtain the peak period interval Monte Carlo model, the trough period interval Monte Carlo model and the non-peak trough period interval Monte Carlo model, wherein the input parameters of the Monte Carlo model are the instantaneous flow data of the water inlet point, the instantaneous flow data of the water outlet point, the pressure data of each water plant, the pressure data of each pump station, and the pressure data of the most unfavorable point, and the output parameters are the target pressure value of each water plant, the target pressure value of each pump station, and the target pressure value of the most unfavorable point.

[0022] That is, it takes into account the different water demand and water supply pressure changes in different time periods. For example, the peak period is the time period of the day when water demand is the highest, such as when people get up in the morning to wash, and go home to cook and take a bath in the evening. During this time period, the water supply system faces the greatest load, and it is necessary to ensure that the water demand of all users can be met. The trough period is the time period of the day when water demand is the lowest, such as late at night. During this time period, the load on the water supply system is small, and the water supply pressure can be appropriately reduced to save energy. The off-peak trough period is the time period between the peak period and the trough period, and the water demand is relatively stable. During this time period, the water supply system needs to maintain a moderate pressure to cope with possible demand fluctuations. Therefore, in order to improve the adaptability and accuracy of pressure regulation control, this application further divides a day into different time periods, including peak period, trough period and non-peak trough period, and divides the training data according to the corresponding time interval to form peak period interval training data, trough period interval training data and non-peak trough period interval training data of the corresponding time interval, so as to establish Monte Carlo models for the water use pattern and pressure change characteristics of each time interval, and more accurately simulate and predict the pressure distribution of the pressure supply point of the water supply network in different time periods, so as to achieve more accurate pressure regulation control. Through this time-divided simulation method, the pressure of the pressure supply point of the water supply network can be more accurately modeled and controlled, and more intelligent water supply regulation can be achieved, thereby further improving the overall management level of the water supply system.

[0023] Figure 4 FIG. 1 is a flow chart of Monte Carlo simulation in a water supply network independent metering zone pressure regulating control system based on Monte Carlo simulation according to an embodiment of the present application. Figure 4 As shown, first, data preparation is performed. The flow data of the water inlet point, the pressure data of each pressure supply point, and the pressure data of the most unfavorable point of a certain independent metering district (District Metered Area, referred to as DMA) within one month are collected. In the specific implementation, the above data can be collected in real time through sensors installed at each key node. In addition, in order to ensure the accuracy and completeness of the data, the data needs to be preprocessed, including data cleaning (removing outliers and missing values), data standardization (converting data of different dimensions to the same scale) and data segmentation (dividing the data into training sets and validation sets for model training and performance evaluation).

[0024] Then, a probability model of water supply and pressure variables for each month is established, and the structure of the Monte Carlo model is defined. Specifically, the input parameters of the Monte Carlo model are the instantaneous flow data of the water inlet point, the instantaneous flow data of the water outlet point, the pressure data of each water plant, the pressure data of each pump station, and the pressure data of the most unfavorable point, and its output parameters are the target pressure value of each water plant, the target pressure value of each pump station, and the target pressure value of the most unfavorable point. At the same time, considering that flow data and pressure data usually have different distribution characteristics, this application selects the normal distribution function to describe the fluctuation of flow data in the actual water supply network, and selects the uniform distribution function to describe the distribution of pressure data in the actual water supply network.

[0025] Next, simulate N data for the peak period, trough period, and non-peak and trough period of each month. Specifically, use a random number generator to extract a large number of random samples from a predefined probability distribution function to construct a sample set. For example, generate 10,000 groups of random samples, and for each group of samples, include the flow rate at the water inlet, the flow rate at the water outlet, the pressure of each water plant, the pressure of each pump station, and the pressure at the most unfavorable point.

[0026] After random sampling, random numbers are randomly sampled from the probability distribution of each variable and substituted into the Monte Carlo model for simulation operation to determine the target pressure value of each pressure supply point (i.e., each water plant, each pump station, and the most unfavorable point) under given flow and pressure conditions, and the simulation results of each group of random samples are statistically analyzed to calculate the average value and standard deviation to evaluate the stability and reliability of the model. Subsequently, the genetic algorithm is used to optimize the target pressure value of each pressure supply point, with the goal of minimizing the total energy consumption while keeping the pressure value of the most unfavorable point within a reasonable range. Through the iterative calculation of the genetic algorithm, the model parameters are gradually optimized to find the best pressure regulation strategy.

[0027] Finally, the model is validated. The validation set data is used to validate the iteratively optimized Monte Carlo model, and the model prediction results are compared with the actual operation data to verify the effectiveness and accuracy of the model. If the model prediction results are in good agreement with the actual data, it means that the model has high reliability and practicality. Through the above steps, the peak interval Monte Carlo model, the trough interval Monte Carlo model, and the non-peak trough interval Monte Carlo model suitable for different time intervals can be trained.

[0028] In the above-mentioned independent metering partition pressure regulation control system 100 for water supply network based on Monte Carlo simulation, the user setting information acquisition module 130 is used to obtain the target pressure value and current time point of the most unfavorable point set by the user. It should be understood that in actual applications, the most unfavorable point is the point with the lowest pressure in the independent metering partition. Ensuring that the pressure value of the most unfavorable point reaches the set target value can ensure that all users in the entire area can obtain sufficient water supply pressure. Therefore, in the actual application stage of the Monte Carlo model, this application will select the corresponding Monte Carlo model for pressure regulation control based on the target pressure value and current time point set by the user.

[0029] In the above-mentioned independent metering and zoning pressure regulation control system 100 for the water supply network based on Monte Carlo simulation, the target Monte Carlo model calling module 140 is used to determine the time interval to which the current time point belongs, and to call the Monte Carlo model of the corresponding time interval to obtain the target Monte Carlo model. For example, if the current time point is at a peak period, the system will call the peak period Monte Carlo model as the target Monte Carlo model to perform real-time pressure regulation control on the water supply network. Correspondingly, if the current time point is at a trough period or a non-peak trough period, the system will respectively call the trough period or non-peak trough period Monte Carlo model for pressure regulation control. In this way, the pressure of the water supply network can be adjusted in real time according to the water use pattern and pressure change characteristics of the current time period to ensure that appropriate water supply pressure can be provided at any time point.

[0030] In the above-mentioned independent metering and zoning pressure regulation control system 100 for the water supply network based on Monte Carlo simulation, the target pressure value generation module 150 is used to input the target pressure value of the most unfavorable point set by the user into the trained target Monte Carlo model to obtain the target pressure value of each water plant and the target pressure value of each pump station. It should be understood that during the training stage, the target Monte Carlo model has learned the pressure distribution law of each pressure supply point in the water supply network within the corresponding time period. Therefore, in actual operation, it is only necessary to input the target pressure value of the most unfavorable point, and the target Monte Carlo model can calculate the target pressure value that each water plant and pump station should reach based on the learned pressure distribution law, so as to achieve pressure balance regulation of the entire water supply network.

[0031] In the above-mentioned independent metering and zoning pressure regulation control system 100 for the water supply network based on Monte Carlo simulation, the remote pressure regulation module 160 is used to determine the remote pressure regulation strategy based on the comparison between the target pressure value of each water plant and the real-time pressure value of each water plant and the comparison between the target pressure value of each pump station and the real-time pressure value of each pump station. That is, by real-time monitoring of the operating status of each water plant and pump station, the output pressure is automatically adjusted to ensure that the target setting is achieved. If the real-time pressure value is lower than the target pressure value, the output pressure is increased, otherwise the output is reduced. This dynamic adjustment mechanism ensures the efficient and stable operation of the water supply network while reducing energy waste.

[0032] In summary, the independent metering partition pressure regulation control system of the water supply network based on Monte Carlo simulation is explained, which divides the water supply network into several independent metering partitions, collects the pressure data of the internal pressure supply point and the most unfavorable point of each independent metering partition, and the flow data of the water inlet and outlet points as training data, and divides a day into three stages: peak period, trough period and non-peak trough period according to the change law of water consumption, and uses the training data in the corresponding time interval to establish different Monte Carlo models, so that according to the target pressure value of the most unfavorable point input by the user, the Monte Carlo model of the corresponding time interval is called to simulate the pressure regulation scheme of each pressure supply point. In this way, the water supply pressure can be dynamically adjusted according to the peak and trough of water supply, so that it is more in line with the user's water use habits, thereby reducing the loss of water supply pipelines caused by excessive pressure and reducing water supply disputes caused by insufficient pressure. Example 2

[0033] Considering that the traditional Monte Carlo model training method requires a large amount of random sampling and calculation, especially for high-dimensional problems, the amount of calculation may be too large, which leads to a long training process. Based on this, the present application proposes an improved training method, which uses a neural network model based on deep learning to mine the complex nonlinear relationship between the pressure data of the pressure supply point, the flow data of the water inlet and outlet points, and the pressure data of the most unfavorable point, thereby accelerating the training process of the Monte Carlo model and improving the efficiency of model training.

[0034] Figure 5 FIG. 1 is a block diagram of a model training unit in a water supply network independent metering zone pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application. Figure 5As shown, the model training unit 123 includes: a full-parameter transient aggregation subunit 1231, which is used to aggregate the instantaneous flow data of the water inlet point, the instantaneous flow data of the water outlet point, the pressure data of each water plant, the pressure data of each pump station and the pressure data of the most unfavorable point according to the parameter sample dimension to obtain an independent metering partition full-parameter transient aggregation vector; a parameter association feature deep mining subunit 1232, which is used to perform parameter association feature deep mining on the independent metering partition full-parameter transient aggregation vector to obtain an independent metering partition full-parameter transient fully connected deep implicit aggregation representation vector; a target pressure The force value inference subunit 1233 is used to generate the inferred target pressure value of each water plant and the inferred target pressure value of each pumping station based on the complementary interactive fusion characteristics between the independent metering partition full-parameter transient aggregation vector and the independent metering partition full-parameter transient fully connected deep implicit aggregation representation vector; the loss function value calculation subunit 1234 is used to train the Monte Carlo model based on the cross entropy loss function value between the inferred target pressure value of each water plant and the inferred target pressure value of each pumping station and the actual target pressure value of each water plant and the actual target pressure value of each pumping station.

[0035] Specifically, the full-parameter transient aggregation subunit 1231 is used to aggregate the instantaneous flow data of the water inlet point, the instantaneous flow data of the water outlet point, the pressure data of each water plant, the pressure data of each pump station and the pressure data of the most unfavorable point according to the parameter sample dimension to obtain the full-parameter transient aggregation vector of the independent metering partition. That is, by aggregating different types of parameter data according to the parameter sample dimension, the multi-source heterogeneous data is integrated into a unified representation vector form, which facilitates the analysis and processing of the correlation between parameters on the same dimension.

[0036] Specifically, the parameter association feature deep mining subunit 1232 is used to perform parameter association feature deep mining on the instantaneous aggregation vector of all parameters of the independent metering partition to obtain the instantaneous fully connected deep implicit aggregation representation vector of all parameters of the independent metering partition. In a specific example of the present application, an embedded encoder based on a multi-layer perceptron is used to perform multi-layer full connection encoding on the instantaneous aggregation vector of all parameters of the independent metering partition to obtain the instantaneous fully connected deep implicit aggregation representation vector of all parameters of the independent metering partition. That is, in order to capture the deep correlation feature representation between the multi-dimensional parameters of the water supply network, the present application uses an embedded encoder based on a multi-layer perceptron to perform feature deep mining on the instantaneous aggregation vector of all parameters of the independent metering partition. Through multi-layer nonlinear transformation, the original instantaneous aggregation vector of all parameters of the independent metering partition is mapped to a high-dimensional implicit feature space to learn the intrinsic correlation structure and feature pattern between the parameters, and generate an instantaneous fully connected deep implicit aggregation representation vector of all parameters of the independent metering partition containing deeper information.

[0037] Specifically, the target pressure value inference subunit 1233 is used to generate the inferred target pressure value of each water plant and the inferred target pressure value of each pump station based on the complementary interactive fusion characteristics between the independent metering partition full parameter transient aggregation vector and the independent metering partition full parameter transient full connection depth implicit aggregation representation vector. Figure 6 FIG. 1 is a block diagram of a target pressure value inference subunit in a water supply network independent metering partition pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application. Figure 6 As shown, the target pressure value inference subunit 1233 includes: a compensatory enhanced interactive aggregation secondary subunit 12331, which is used to perform compensatory enhanced interactive aggregation on the independent metering partition full-parameter transient fully connected depth implicit aggregation representation vector and the independent metering partition full-parameter transient aggregation vector to obtain the independent metering partition full-parameter transient residual aggregation enhanced representation vector; a decoding estimation secondary subunit 12332, which is used to input the independent metering partition full-parameter transient residual aggregation enhanced representation vector into a decoder-based target parameter estimation module to obtain the inferred target pressure value of each water plant and the inferred target pressure value of each pump station.

[0038] Specifically, the compensatory enhanced interactive aggregation secondary subunit 12331 is used to perform compensatory enhanced interactive aggregation on the independent metering partition full parameter transient fully connected deep implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector to obtain the independent metering partition full parameter transient residual aggregation enhanced representation vector. It should be understood that, considering that the multi-layer fully connected encoding can capture deeper parameter correlation features, it will also cause some detail information in the original data to be lost. Therefore, in order to comprehensively utilize the complementary information in the independent metering partition full parameter transient fully connected deep implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector, the present application further performs feature interactive fusion on the two to improve the accuracy of feature representation.

[0039] Figure 7 : is a block diagram of a secondary subunit of a compensatory enhanced interactive aggregation in a water supply network independent metering and zoning pressure regulation control system based on Monte Carlo simulation according to an embodiment of the present application. Figure 7As shown, the compensatory enhanced interactive aggregation secondary subunit 12331 includes: a common feature extraction tertiary subunit 123311, which is used to input the independent metering partition full parameter transient fully connected depth implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector into a common feature extraction network to obtain an independent metering partition full parameter transient residual common feature representation vector; a feature compensation modulation tertiary subunit 123312, which is used to extract the independent metering partition full parameter transient fully connected depth implicit aggregation representation vector based on the independent metering partition full parameter transient fully connected depth implicit aggregation representation vector, the independent metering partition full parameter transient aggregation vector relative to the independent metering partition full parameter transient residual common feature representation vector The specific features of the independent metering partition full parameter transient fully connected deep implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector are feature compensated and modulated to obtain an independent metering partition full parameter transient fully connected deep feature compensation vector and an independent metering partition full parameter transient feature compensation vector; the feature fusion tertiary subunit 123313 is used to cascade the independent metering partition full parameter transient fully connected deep feature compensation vector, the independent metering partition full parameter transient feature compensation vector and the independent metering partition full parameter transient residual common feature representation vector to obtain the independent metering partition full parameter transient residual aggregation enhanced representation vector.

[0040] In a specific example of the present application, the common feature extraction network performs point-wise fusion on the independent metering partition full-parameter transient fully connected deep implicit aggregation representation vector and the independent metering partition full-parameter transient aggregation vector, and then uses a tanh function-based neural network layer to extract common features to obtain the independent metering partition full-parameter transient residual common feature representation vector.

[0041] The above common feature extraction three-level sub-unit 123311 can be expressed by the formula: in, represents the fully parameterized transient fully connected deep implicit aggregation representation vector of the independent metering partition, represents the transient aggregation vector of all parameters of the independent metering partition, represents the common characteristic vector of transient residual errors of all parameters in independent measurement partitions, and Represent the weight parameter matrix and bias term respectively, That is, a common feature extraction network is constructed using a neural network unit to learn the common pattern between the independent metering partition full parameter transient fully connected deep implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector, and generate an independent metering partition full parameter transient residual common feature representation vector.

[0042] In a specific example of the present application, the feature compensation modulation three-level subunit 123312 is used to: input the independent metering partition full parameter transient full connection depth implicit aggregation representation vector, the independent metering partition full parameter transient aggregation vector and the independent metering partition full parameter transient residual common feature representation vector into a vector probabilization unit based on the Sigmoid function to obtain a probabilistic independent metering partition full parameter transient full connection depth implicit aggregation representation vector, a probabilistic independent metering partition full parameter transient aggregation vector and a probabilistic independent metering partition full parameter transient residual common feature representation vector; which can be expressed by the formula:

[0043]

[0044]

[0045] in, represents the sigmoid function, represents the probabilistic independent measurement partition full parameter transient fully connected deep implicit aggregation representation vector, represents the transient aggregation vector of all parameters of the probabilistic independent metering partition, Represents the common feature representation vector of transient residuals of all parameters of probabilistic independent measurement partitions. That is, by normalizing them separately, the value range of the feature vector is ensured to be between 0 and 1, which is convenient for subsequent data processing.

[0046] Next, the specific features of the implicit aggregation representation vector of the transient full-connected depth of all parameters of the probabilistic independent metering partition relative to the transient residual common feature representation vector of the probabilistic independent metering partition are calculated to obtain the transient full-connected depth specific feature compensation representation vector of the transient full parameters of the independent metering partition; the specific features of the transient aggregation vector of the transient full parameters of the probabilistic independent metering partition relative to the transient common feature representation vector of the transient full parameters of the probabilistic independent metering partition are calculated to obtain the transient specific feature compensation representation vector of the transient full parameters of the independent metering partition; the above process can be expressed by the formula:

[0047]

[0048] in, , and They represent the first eigenvalue, the first eigenvalue of the transient aggregation vector of all parameters of the probabilistic independent measurement partition The eigenvalue and the common characteristic feature representation vector of the transient residual of all parameters of the probabilistic independent measurement partition eigenvalues, represents the logarithmic function with base 2, represents the natural exponential function, represents the normalized exponential function, and They represent the independent metering partition full parameter transient fully connected depth specific feature compensation representation vector and the independent metering partition full parameter transient specific feature compensation representation vector respectively. That is, by respectively calculating the feature difference between the probabilistic independent metering partition full parameter transient fully connected depth implicit aggregation representation vector and the probabilistic independent metering partition full parameter transient aggregation vector relative to the probabilistic independent metering partition full parameter transient residual common feature representation vector, the unique feature information of the independent metering partition full parameter transient fully connected depth implicit features and the original full parameter transient aggregation features in addition to the common features is mined.

[0049] Then, the position point multiplication between the instantaneous full-connected depth specific feature compensation representation vector of all parameters of the independent metering partition and the instantaneous full-connected depth implicit aggregation representation vector of all parameters of the independent metering partition is calculated to obtain the instantaneous full-connected depth feature compensation vector of all parameters of the independent metering partition; the position point multiplication between the instantaneous specific feature compensation representation vector of all parameters of the independent metering partition and the instantaneous aggregation vector of all parameters of the independent metering partition is performed to obtain the instantaneous feature compensation vector of all parameters of the independent metering partition; the above process can be expressed by the formula:

[0050]

[0051] in, represents the Hadamard product, and They represent the independent metering partition full parameter transient fully connected deep feature compensation vector and the independent metering partition full parameter transient feature compensation vector respectively. That is, the original independent metering partition full parameter transient fully connected deep implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector are compensated and strengthened by positional point multiplication to ensure that the unique information of the two is retained as much as possible during the fusion process.

[0052] In a specific example of the present application, the feature fusion three-level subunit 123313 is used to cascade the independent metering partition full parameter transient fully connected deep feature compensation vector, the independent metering partition full parameter transient feature compensation vector and the independent metering partition full parameter transient residual common feature representation vector to obtain the independent metering partition full parameter transient residual aggregation enhanced representation vector, which can be expressed by the formula:

[0053] in, Indicates cascade operation, Represents the independent metering partition full parameter transient residual aggregation enhanced representation vector. That is, the compensated enhanced independent metering partition full parameter transient fully connected depth implicit aggregation representation vector, the independent metering partition full parameter transient aggregation vector and the independent metering partition full parameter transient residual common feature representation vector are fused to comprehensively consider the common information between the two and their respective unique information, and generate the independent metering partition full parameter transient residual aggregation enhanced representation vector, thereby retaining the high-level features while retaining the detailed information of the original data, and realizing a comprehensive understanding and accurate modeling of the correlation pattern between multi-dimensional parameters.

[0054] Specifically, the decoding estimation secondary subunit 12332 is used to input the independent metering partition full parameter transient residual aggregation enhanced representation vector into the decoder-based target parameter estimation module to obtain the inferred target pressure value of each water plant and the inferred target pressure value of each pump station. In a specific example of the present application, the decoder adopts an LSTM architecture and uses the memory capacity of the LSTM model to capture and utilize the parameter association information in the independent metering partition full parameter transient residual aggregation enhanced representation vector to more accurately predict and estimate the target pressure value of each water plant and pump station.

[0055] Specifically, the loss function value calculation subunit 1234 is used to train the Monte Carlo model based on the cross entropy loss function value between the inferred target pressure value of each water plant and the inferred target pressure value of each pump station and the real target pressure value of each water plant and the real target pressure value of each pump station. That is, the Monte Carlo model is trained using the cross entropy loss function as the training objective function to optimize the model parameters. The cross entropy loss function guides the model to continuously adjust parameters during the training process to minimize the prediction error by measuring the difference between the inferred target pressure value and the real target pressure value of each water plant and each pump station inferred by the target parameter estimation module. By minimizing the cross entropy loss function value, the prediction result of the model is closer to the real target pressure value, thereby achieving accurate control and management of the pressure of each pressure supply point.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A water supply network independent metering and zoning pressure regulation control system based on Monte Carlo simulation, characterized in that: include: A training data acquisition module, used to acquire training data of each independent metering partition collected by the sensor group; A Monte Carlo model training module, for dividing the training data of each independent metering partition into time intervals and then training the Monte Carlo model respectively to obtain a peak interval Monte Carlo model, a trough interval Monte Carlo model and a non-peak trough interval Monte Carlo model; A user setting information acquisition module is used to obtain the target pressure value and current time point of the most unfavorable point set by the user; A target Monte Carlo model retrieving module, used to determine the time interval to which the current time point belongs, and retrieve the Monte Carlo model of the corresponding time interval to obtain the target Monte Carlo model; A target pressure value generating module, used for inputting the target pressure value of the most unfavorable point set by the user into the trained target Monte Carlo model to obtain the target pressure value of each water plant and the target pressure value of each pump station; The pressure remote regulation module is used to determine the pressure remote regulation strategy based on the comparison between the target pressure value of each water plant and the real-time pressure value of each water plant and based on the comparison between the target pressure value of each pumping station and the real-time pressure value of each pumping station.

2. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 1 is characterized in that: The training data includes a time queue of instantaneous flow data at a water inlet, a time queue of instantaneous flow data at a water outlet, a time queue of pressure data at each water plant, a time queue of pressure data at each pump station, and a time queue of pressure data at the most unfavorable point, wherein the water supply network is divided into a number of independent metering partitions, and the most unfavorable point in the independent metering partition is the point with the lowest pressure in the independent metering partition.

3. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 2 is characterized in that: The Monte Carlo model training module includes: A time interval determination unit, configured to determine the time intervals of the peak period interval, the trough period interval, and the non-peak and trough period interval of each independent metering partition every day based on the training data of each independent metering partition; A training data segmentation unit, configured to segment the training data based on the peak period interval, the trough period interval and the non-peak and trough period interval of each day to obtain peak period interval training data, trough period interval training data and non-peak and trough period interval training data; A model training unit is used to train the Monte Carlo model based on the peak period interval training data, the trough period interval training data and the non-peak trough period interval training data to obtain a peak period interval Monte Carlo model, a trough period interval Monte Carlo model and a non-peak trough period interval Monte Carlo model, wherein the input parameters of the Monte Carlo model are the instantaneous flow data of the water inlet point, the instantaneous flow data of the water outlet point, the pressure data of each water plant, the pressure data of each pump station, and the pressure data of the most unfavorable point, and its output parameters are the target pressure value of each water plant, the target pressure value of each pump station, and the target pressure value of the most unfavorable point.

4. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 3 is characterized in that: The model training unit comprises: A full-parameter transient aggregation subunit, used for aggregating the instantaneous flow data of the water inlet point, the instantaneous flow data of the water outlet point, the pressure data of each water plant, the pressure data of each pump station and the pressure data of the most unfavorable point according to the parameter sample dimension to obtain a full-parameter transient aggregation vector of an independent metering partition; A parameter-related feature deep mining subunit is used to perform parameter-related feature deep mining on the transient aggregate vector of all parameters of the independent metering partition to obtain a transient fully connected deep implicit aggregate representation vector of all parameters of the independent metering partition; A target pressure value inference subunit, used to generate the inferred target pressure value of each water plant and the inferred target pressure value of each pump station based on the complementary interactive fusion characteristics between the independent metering partition full parameter transient aggregation vector and the independent metering partition full parameter transient fully connected deep implicit aggregation representation vector; The loss function value calculation subunit is used to train the Monte Carlo model based on the cross entropy loss function value between the inferred target pressure value of each water plant and the inferred target pressure value of each pump station and the real target pressure value of each water plant and the real target pressure value of each pump station.

5. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 4 is characterized in that: The parameter association feature deep mining subunit is used to: A multi-layer perceptron-based embedded encoder is used to perform multi-layer full-connection encoding on the independent metering partition full-parameter transient aggregation vector to obtain the independent metering partition full-parameter transient fully-connected deep implicit aggregation representation vector.

6. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 5 is characterized in that: The target pressure value inference subunit includes: A compensatory enhanced interactive aggregation secondary subunit is used to perform a compensatory enhanced interactive aggregation on the independent metering partition full parameter transient fully connected depth implicit aggregation representation vector and the independent metering partition full parameter transient aggregation vector to obtain an independent metering partition full parameter transient residual aggregation enhanced representation vector; The decoding estimation secondary subunit is used to input the independent metering partition full parameter transient residual aggregation enhanced representation vector into the decoder-based target parameter estimation module to obtain the inferred target pressure value of each water plant and the inferred target pressure value of each pump station.

7. The independent metering and zoning pressure regulation control system for water supply network based on Monte Carlo simulation according to claim 6 is characterized in that: The compensatory enhanced interactive polymerization secondary subunit comprises: The third-level subunit for common feature extraction is used to input the independent metering partition full-parameter transient fully connected depth implicit aggregation representation vector and the independent metering partition full-parameter transient aggregation vector into the common feature extraction network to obtain the independent metering partition full-parameter transient residual common feature representation vector; The feature compensation modulation tertiary subunit is used to perform feature compensation modulation on the independent metering partition full parameter instantaneous full connection depth implicit aggregation representation vector and the independent metering partition full parameter instantaneous aggregation vector based on the specific features of the independent metering partition full parameter instantaneous aggregation vector relative to the independent metering partition full parameter instantaneous residual common feature representation vector to obtain the independent metering partition full parameter instantaneous full connection depth feature compensation vector and the independent metering partition full parameter instantaneous feature compensation vector; The feature fusion three-level sub-unit is used to cascade the independent metering partition full parameter transient fully connected deep feature compensation vector, the independent metering partition full parameter transient feature compensation vector and the independent metering partition full parameter transient residual common feature representation vector to obtain the independent metering partition full parameter transient residual aggregation enhanced representation vector.

8. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 7 is characterized in that: After the common feature extraction network performs point-wise fusion on the transient fully connected deep implicit aggregation representation vector of all parameters of the independent metering partition and the transient aggregation vector of all parameters of the independent metering partition, it uses a neural network layer based on the tanh function to extract common features to obtain the transient residual common feature representation vector of all parameters of the independent metering partition.

9. The independent metering and zoning pressure regulating control system for water supply network based on Monte Carlo simulation according to claim 8 is characterized in that: The characteristic compensation modulation three-level subunit is used for: Input the independent metering partition full parameter transient fully connected depth implicit aggregation representation vector, the independent metering partition full parameter transient aggregation vector and the independent metering partition full parameter transient residual common feature representation vector into a vector probabilization unit based on the Sigmoid function to obtain a probabilistic independent metering partition full parameter transient fully connected depth implicit aggregation representation vector, a probabilistic independent metering partition full parameter transient aggregation vector and a probabilistic independent metering partition full parameter transient residual common feature representation vector; Calculate the specific features of the probabilistic independent measurement partition full parameter transient full connection depth implicit aggregation representation vector relative to the probabilistic independent measurement partition full parameter transient residual common feature representation vector to obtain the independent measurement partition full parameter transient full connection depth specific feature compensation representation vector; Calculate the specific features of the transient aggregate vector of all parameters of the probabilistic independent metering partition relative to the transient residual common feature representation vector of all parameters of the probabilistic independent metering partition to obtain a transient specific feature compensation representation vector of all parameters of the independent metering partition; Calculate the position point multiplication between the independent metering partition full parameter instantaneous full connection depth specific feature compensation representation vector and the independent metering partition full parameter instantaneous full connection depth implicit aggregation representation vector to obtain the independent metering partition full parameter instantaneous full connection depth feature compensation vector; The independent metering partition full parameter transient specific feature compensation representation vector is multiplied by the position point between the independent metering partition full parameter transient aggregation vector to obtain the independent metering partition full parameter transient feature compensation vector.

Images