CN-121998317-A - Dynamic self-adaptive optimal scheduling method, system, medium and equipment based on mixed deep learning

CN121998317ACN 121998317 ACN121998317 ACN 121998317ACN-121998317-A

Abstract

The invention relates to the field of urban water supply networks and discloses a dynamic self-adaptive optimal scheduling method, a system, a medium and equipment based on mixed deep learning, wherein the method comprises the steps of transmitting collected data related to operation and maintenance of a water plant, a network hydraulic model and water used by users into a water consumption prediction model of the users, updating a dynamic water consumption prediction matrix to calculate the use condition of the predicted water consumption, and obtaining the dynamic water consumption matrix obtained by water consumption prediction The method comprises the steps of constructing a connection between a water meter and a hydraulic model node, dynamically updating water quantity, water supply quantity and supply pressure of a water plant in a pipe network to obtain the current pipe network operation condition so as to support daily operation and maintenance, determining the total water supply quantity of the water plant through water quantity prediction, determining the optimized water supply quantity as n hours, taking the water quantity predicted water supply quantity as the optimized water demand, optimizing the operation and maintenance scheme of the water plant, evaluating the hydraulic state after scheme generation, and outputting a dynamic optimized scheduling scheme.

Inventors

LIU SHUMING
WU XUE
XIE YIFAN
WANG ZEMIN
WANG XINGYE
WANG WENHUA

Assignees

清华大学
唐山市曹妃甸供水有限责任公司

Dates

Publication Date: 20260508
Application Date: 20260105

Claims (10)

1. A dynamic self-adaptive optimal scheduling method based on mixed deep learning is characterized by comprising the following steps: The collected data about the operation and maintenance of the water plant, the hydraulic model of the pipe network and the water consumption of the user are used as input data, the input data are transmitted to a water consumption prediction model of the user, a dynamic water consumption prediction matrix is updated, a water consumption prediction result is obtained, and the predicted water consumption condition is calculated; the dynamic water quantity matrix is constructed by combining the water quantity prediction result with the monitoring data As the real-time water quantity of the pipe network for updating, the water quantity, the water supply quantity of a water plant and the supply pressure are dynamically updated in the pipe network by constructing the connection between the water meter and the node of the hydraulic model, so that the running condition of the current pipe network is obtained to support daily operation and maintenance; determining the total water supply amount of the water plant through water quantity prediction, determining the optimized number of hours as n hours, and setting the water quantity predicted water supply amount as And (3) as the water demand for optimizing the scheduling, optimizing the operation and maintenance scheme of the water plant, and after the scheme is generated, evaluating the hydraulic state by combining the obtained pipe network operation condition to output a dynamic optimizing scheduling scheme.
2. The method for dynamically adaptively optimizing and scheduling based on mixed deep learning as set forth in claim 1, wherein the user water consumption prediction model is a water consumption prediction mixed neural network formed by combining a time sequence convolutional neural network TCN and a long-short-term memory network LSTM.
3. The method for dynamically adaptively optimizing and scheduling based on mixed deep learning as set forth in claim 2, wherein the TCN processes the time series data by using a set filter by adopting random convolution, and the time series data at the t moment of the upper layer is only derived from the data at the adjacent past moment, and the time convolution network introduces an expansion convolution and residual module; The residual error module consists of two convolution units and a nonlinear mapping unit, wherein each convolution unit comprises one-dimensional expansion causal convolution, weight normalization, correction linear unit activation function and random inactivation operation, and a convolution of 1 multiplied by 1 is added on a shortcut branch.
4. The method for dynamically adaptively optimizing scheduling based on hybrid deep learning as set forth in claim 2, wherein the long-short-term memory network LSTM is composed of a plurality of memory units, each of which has two states, a hidden state and a cell state, and each of which is composed of three internal gates, so that the long-short-term memory network can forget, input and output information.
5. The hybrid deep learning-based dynamic adaptive optimal scheduling method of claim 1, wherein the dynamic water quantity prediction matrix is: in the formula, Representing the mth user At the nth time of the future Is used for the water quantity prediction.
6. The method for dynamically adaptively optimizing and scheduling based on mixed deep learning as set forth in claim 1, wherein the uncertainty prediction is performed on the user water consumption prediction model in the water consumption prediction process, and the method comprises the following specific steps: The uncertainty of water quantity prediction is represented by assuming that the water quantity presents Gaussian distribution and predicting the mean value and standard deviation, and the scheduler is supported by the uncertainty to carry out comprehensive judgment under the condition that the water quantity change value is larger than a set threshold value.
7. The hybrid deep learning-based dynamic adaptive optimal scheduling method of claim 1, wherein the optimal scheduling includes the following objective functions and constraints: the objective function is: ; The water quantity constraint is as follows: ; the pump operating state change constraints are: in the formula, Recording electricity consumption for the optimized operation of the plant; Representing a water plant optimized scheduling search space; Supplying water to the water plant; optimizing the water supply of the scheme for the water plant; Indicating that the water pump is combined into The total water supply is Optimal energy consumption in time, k is water pump combination A water pump; Is a water pump combination Removing a water pump combination of the water pump k; The water supply amount is the water supply amount when the independent operation frequency of the water pump k is f; is the energy consumption of the water pump k when the independent operation frequency is f.
8. A hybrid deep learning based dynamic adaptive optimal scheduling system, comprising: The data collection module is used for transmitting the collected data about the operation and maintenance of the water plant, the hydraulic model of the pipe network and the water consumption of the user as input data to the water consumption prediction model of the user, updating the dynamic water consumption prediction matrix, and obtaining a water consumption prediction result so as to calculate the predicted water consumption use condition; the dynamic water conservancy simulation module is used for constructing a dynamic water volume matrix by combining water volume prediction results with monitoring data As the real-time water quantity of the pipe network for updating, the water quantity, the water supply quantity of a water plant and the supply pressure are dynamically updated in the pipe network by constructing the connection between the water meter and the node of the hydraulic model, so that the running condition of the current pipe network is obtained to support daily operation and maintenance; the optimal scheduling module is used for determining the total water supply amount of the water plant through water quantity prediction, determining the optimal time number as n hours and setting the water supply amount of the water quantity prediction as And (3) as the water demand for optimizing the scheduling, optimizing the operation and maintenance scheme of the water plant, and after the scheme is generated, evaluating the hydraulic state by combining the obtained pipe network operation condition to output a dynamic optimizing scheduling scheme.
9. A computer readable storage medium storing one or more programs, wherein the one or more programs comprise instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-7.
10. A computing device comprising one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the methods of claims 1-7.

Description

Dynamic self-adaptive optimal scheduling method, system, medium and equipment based on mixed deep learning Technical Field The invention relates to the technical field of urban water supply networks, in particular to a dynamic self-adaptive optimal scheduling method, system, medium and equipment based on mixed deep learning. Background The water supply network is an important city infrastructure, and the stable and efficient operation of the water supply network is an important foundation stone for guaranteeing the life of people. The hydraulic water quality characteristics of the water supply pipe network are the most important parameters affecting the operation of the water supply pipe network, and each water supply hydraulic water quality is not up to standard, so that user dissatisfaction and even water pollution accidents can be caused. The water supply network is comprehensively simulated by dynamically distributing water quantity to each user point and combining the states of foundation settings such as a water plant, a valve and the like. At present, the hydraulic model of a water supply network is mostly a static model, namely, the water supply network is simulated only through data in certain time periods, which is limited by various conditions such as data acquisition, model modeling and the like. Although the system can better express the hydraulic water quality condition of the whole urban water supply network, because of the static property, the simulation is limited to a certain specific working condition, and the simulation such as pipeline planning, scheme analysis, accident prediction and the like can be carried out only, so that the system is difficult to be applied to daily optimal scheduling simulation. Disclosure of Invention Aiming at the problems, the invention aims to provide a dynamic self-adaptive optimal scheduling method, a system, a medium and equipment based on mixed deep learning, which improve the real-time, dynamic and reliable characteristics of optimal scheduling of a water supply network. In order to achieve the aim, the first aspect adopts the technical scheme that the dynamic self-adaptive optimal scheduling method based on mixed deep learning comprises the steps of taking collected data related to operation and maintenance of a water plant, a pipe network hydraulics model and user water as input data, transmitting the data to a user water consumption prediction model, updating a dynamic water consumption prediction matrix to obtain a water consumption prediction result so as to calculate the predicted water consumption situation, and constructing the dynamic water consumption matrix by combining the water consumption prediction result with monitoring dataThe method comprises the steps of constructing the connection between a water meter and a hydraulic model node as real-time water quantity of a pipe network for updating, dynamically updating the water quantity, water supply quantity and supply pressure of a water plant in the pipe network to obtain the current pipe network operation condition so as to support daily operation and maintenance, determining the total water supply quantity of the water plant through water quantity prediction, determining the optimized number of hours as n hours, and setting the water supply quantity predicted by the water quantity asAnd (3) as the water demand for optimizing the scheduling, optimizing the operation and maintenance scheme of the water plant, and after the scheme is generated, evaluating the hydraulic state by combining the obtained pipe network operation condition to output a dynamic optimizing scheduling scheme. Further, the water consumption prediction model of the user is a water consumption prediction mixed neural network formed by combining a time sequence convolutional neural network TCN and a long-short-term memory network LSTM. Further, the TCN adopts random convolution, time sequence data is processed through a set filter, the time sequence data at the t moment of the upper layer only comes from the data at the nearby past moment, and the time convolution network introduces an expansion convolution and residual error module; The residual error module consists of two convolution units and a nonlinear mapping unit, wherein each convolution unit comprises one-dimensional expansion causal convolution, weight normalization, correction linear unit activation function and random inactivation operation, and a convolution of 1 multiplied by 1 is added on a shortcut branch. Furthermore, the long-term memory network LSTM is composed of a plurality of memory units, each memory unit has two states, namely a hidden state and a cell state, and each memory unit is composed of three internal gates, so that the long-term memory network can forget, input and output information. Further, the dynamic water quantity prediction matrix is: in the formula, Representing the mth userAt the nth time of the futureIs a predicted amount of wate