Intelligent water supply revenue system

Abstract

The invention belongs to the technical field of water supply systems, and particularly relates to an intelligent water supply revenue system. Comprising a data collection module for collecting data such as environment and water by using an intelligent water meter and transmitting the data to a central data processor; the intelligent scheduling module carries out demand prediction on water supply related factors and schedules resources by using a multi-layer adaptive neural network, the neural network comprises three sub-network layers, and the layers have different functions and can be adaptively adjusted; and the dynamic pricing and charging module adopts an improved hybrid genetic algorithm to optimize and dynamically adjust the water price based on the collected data, and relates to the steps of fitness function multi-dimensional optimization, nonlinear crossing, high-order nonlinear variation, particle swarm optimization and the like. Multiple factors can be integrated, the water supply management efficiency, the resource utilization rate and the service quality are improved, and sustainable development of the water supply industry is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of water supply systems, and in particular relates to a smart water supply revenue system. Background Art

[0002] In the water supply industry, the traditional water supply revenue model and scheduling methods have many shortcomings. Traditional water supply systems often lack effective use of real-time data and cannot accurately perform dynamic pricing and intelligent scheduling based on environmental factors, user water habits and the actual situation of the water supply network. In terms of pricing, most of them adopt a fixed water price model, which fails to fully consider the impact of factors such as water consumption peaks, troughs and environmental changes on water supply costs and demand, resulting in unreasonable resource allocation, difficulty in maximizing the benefits of water supply companies, and users may also face unreasonable charges. In terms of scheduling, due to the lack of accurate demand forecasting, it is difficult to achieve the optimal allocation of water supply resources, and it is easy to have insufficient or excessive local water supply pressure, affecting the quality of water supply services. At the same time, with the acceleration of urbanization and the improvement of people's requirements for quality of life, the demand for intelligent and precise management of water supply systems is becoming increasingly urgent, and traditional water supply revenue systems can no longer meet the needs of modern social development. Therefore, it is necessary to develop a smart water supply revenue system that can comprehensively consider multiple factors, realize dynamic pricing and billing, and intelligent scheduling, so as to improve the management efficiency, resource utilization and service quality of the water supply system and promote the sustainable development of the water supply industry. Summary of the invention

[0003] In view of the existing technical problems, the present invention proposes a smart water supply revenue system with reasonable design and simple method.

[0004] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: comprising a data collection module, an intelligent scheduling module and a dynamic pricing and billing module;

[0005] The data collection module uses a smart water meter with wireless communication function to collect environmental factors and user water use data in real time, including but not limited to water flow, water use time, and water supply network pressure, and transmits the collected real-time data to the central data processor;

[0006] The intelligent scheduling module uses a multi-level adaptive neural network to predict water supply network pressure, environmental factors, water use time and water flow according to user water demand, and schedules water supply resources;

[0007] The dynamic pricing and billing module is based on real-time collected water use data and precipitation factors, and uses an improved hybrid genetic algorithm to optimize and derive a predicted price;

[0008] The specific implementation steps of the optimization using the improved hybrid genetic algorithm are:

[0009] Step 1: The first step is to optimize the fitness function in multiple dimensions. Taking into account the prediction error, water use data, and environmental change factors, the improved fitness function is: ,in For the predicted water price, is the actual water price, is the predicted water supply network pressure, is the actual pressure, is the impact of environmental changes on water prices, expressed as the deviation of environmental parameter changes on the predicted water price. is the weight coefficient, which is used to adjust the priority of different goals;

[0010] Step 2: Introduce nonlinear crossover in the crossover process to generate a new configuration space different from the parent generation, which is specifically expressed as: ,in, are the values ​​of the parent individual at the i-th hyperparameter, is the value of the hyperparameter of the offspring individual obtained by nonlinear crossover, It is a nonlinear adjustment factor that controls the nonlinear degree of crossover;

[0011] Step 3: Then introduce high-order nonlinear mutation to change the value range of hyperparameters. The specific mutation formula is: ,in, is the value of the i-th hyperparameter in the parent individual, is the hyperparameter value of the offspring individual at the ith position obtained after the mutation operation, is a coefficient of variation , is the initial maximum coefficient of variation, is a nonlinear transformation function used to mutate the parent hyperparameters;

[0012] Step 4: In each generation of the genetic algorithm, the best individual is fine-tuned using local particle swarm optimization. The particle swarm position update formula takes into account the inertia weight and adaptive learning factor.

[0013] Step 5: Finally, when the maximum number of generations reaches the preset value of 30, the algorithm terminates.

[0014] Preferably, the dynamic pricing and billing module uses an improved hybrid genetic algorithm to optimize the weight coefficient in step 1. The method of determining is based on the contribution of each goal , which is defined as: , , ,in is the mean square error of the water price forecast error, is the mean square error of the water supply network pressure error, is the deviation of the impact of environmental changes on water prices, , , , .

[0015] As a preference, the nonlinear adjustment factor in step 2 of the optimization using the improved hybrid genetic algorithm in the dynamic pricing and billing module The specific form is: ,in is the maximum nonlinear factor value, are the values ​​of the i-th hyperparameter for parent individuals 1 and 2, respectively. Represents the difference between the two parent individuals in this hyperparameter.

[0016] Preferably, the multi-level adaptive neural network in the intelligent scheduling module is specifically implemented as follows:

[0017] Step 1: First, standardize the collected data and remove outliers to ensure data stability;

[0018] Step 2: Construct a multi-level adaptive neural network, which is divided into three sub-network levels. Each layer processes different types of information and the number of neurons in each layer and their connection mode can be adaptively adjusted according to the changes in real-time input;

[0019] Step 3: Finally, the learning rate of each layer is dynamically adjusted according to the gradient changes in the training cycle. The learning rate update formula is: ,in is the gradient of the current cycle, are the gradients of cycle 0 and the current one, respectively. is the current target loss value, is the predetermined optimal target loss value.

[0020] Preferably, the three sub-network levels of the multi-level adaptive neural network in step 2 of the intelligent scheduling module, the first layer uses a convolutional neural network to process environmental data and extract relevant features, the second layer uses an LSTM model to process time series data, which is a demand prediction network, and the third layer is a scheduling optimization network, which is combined with a support vector machine to adjust the decision of the water supply system.

[0021] Compared with the prior art, the advantages and positive effects of the present invention are that the intelligent scheduling module adopts a multi-level adaptive neural network, which can accurately predict water demand. Through the coordination of three sub-network levels, it can adaptively adjust neurons and connection methods, dynamically update learning rates and connection weights, and effectively track changes in water demand. The dynamic pricing and billing module uses improved hybrid genetic algorithm optimization, multi-dimensional optimization of fitness functions, and comprehensive consideration of multiple factors to make water prices more reasonable, which can promote the rational use of water resources, ensure corporate benefits, improve user recognition, and promote sustainable development of the industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 It is a structural flow chart of the present invention; Figure 2 Schematic diagram of fitness function iteration. DETAILED DESCRIPTION

[0024] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0025] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0026] Embodiment: Under the existing water supply model, problems continue to emerge. On the one hand, residential areas often face the problem of insufficient water pressure and unreasonable water prices during peak water use; commercial areas have high requirements for water supply stability due to concentrated water use time and concentrated equipment, and traditional methods are difficult to accurately guarantee and cannot effectively motivate them to save water; industrial areas have strong production continuity and complex and changeable water demand, and traditional water supply systems cannot adapt, resulting in increased risks of production interruptions and inaccurate cost accounting. Therefore, the present invention proposes a smart water supply revenue system. Figure 1 This is the overall structural process of the present invention.

[0027] First, in order to fully obtain water supply-related data, the data collection module uses smart water meters with wireless communication functions. Smart water meters are installed on the water supply pipes in each water use area to collect data such as water flow, water use time, and water supply network pressure in real time, and transmit the data to the central data processor through wireless communication technology. At the same time, real-time environmental data, including environmental factors such as temperature, humidity, and precipitation, are obtained, and these environmental data are associated with water use data. Through comprehensive data collection, a rich data foundation is provided for subsequent dynamic pricing and intelligent scheduling.

[0028] Then, in order to accurately predict water demand in the intelligent scheduling module, a multi-level adaptive neural network is used. Compared with the existing technology, the multi-level adaptive neural network has significant advantages in demand estimation. Traditional technologies are mostly based on simple historical data statistics or single model predictions, which are difficult to handle complex and changeable water use situations. The multi-level adaptive neural network of the present invention works together through three sub-network levels. First, the collected data is standardized, and outliers are removed to ensure the stability of the data; a multi-level adaptive neural network is constructed, which is divided into three sub-network levels. Each layer processes different types of information and the number of neurons in each layer and their connection mode can be adaptively adjusted according to the changes in real-time input. The first layer processes environmental data and uses a convolutional neural network to extract the mean, maximum, and change trend of temperature, the average level, fluctuation range, and duration of humidity, and the total amount, duration, and intensity characteristics of precipitation. These characteristics reflect changes in the environment. Specifically, first, the collected environmental data such as temperature, humidity, and precipitation are sorted into a format suitable for convolutional neural network input according to time series. Then, the convolution kernel of the convolutional layer is used to slide on the data, and a weighted sum operation is performed on the local data, and different features are extracted by setting different convolution kernels. After multiple layers of convolution and pooling, the fully connected layer integrates and maps the extracted features, and finally outputs the required temperature, humidity, and precipitation features. The second-layer LSTM model provides basic input to assist it in capturing the relationship between water demand and the environment. The precipitation features provide a decision-making basis for the third-layer scheduling optimization network to achieve optimal allocation of water supply resources, make dynamic pricing more reasonable, and promote the sustainable development of the water supply industry. The second layer uses the LSTM model to process time series data, which is a demand forecasting network. In this layer, the LSTM model mainly processes time series data related to water use, such as water flow data in different time periods, water use time distribution data, and comprehensive time series data combined with the first-layer environmental data, and the changes in water use data under different temperature, humidity, and precipitation conditions over time. The output result is the predicted value of water demand in the future, including the predicted water flow, peak and trough time points of water use, and water demand. By learning and analyzing the time series of historical water use data and related environmental data, the LSTM model can capture the long-term trend, seasonal changes, periodic fluctuations and dynamic laws related to changes in environmental factors of water demand, thereby providing accurate water demand forecasts for the intelligent scheduling module to arrange water supply resources reasonably. The third layer is the scheduling optimization network, which combines the support vector machine to adjust the decision of the water supply system. The specific implementation is to first receive the features extracted from the first layer of environmental data processing, and the water demand forecast results output by the second layer LSTM model, including water flow, water use time distribution and other forecast information.Then, a support vector machine model is constructed for these data, the kernel function of the model is determined, and the current integrated feature vector is input into the trained support vector machine model. The model outputs decision suggestions for the water supply system under the current situation, the set value of the water supply pressure in the area, the start or stop of a specific water pump, and the balance of water supply flow in each area based on the learned mapping relationship. Finally, the learning rate of each layer is dynamically adjusted according to the gradient change in the training cycle, and the learning rate update formula is:. ,in is the gradient of the current cycle, are the gradients of cycle 0 and the current one, respectively. is the current target loss value, is the predetermined optimal target loss value. It adaptively adjusts the number of neurons and the way they are connected to accurately capture the unique and dynamically changing water use patterns in different regions. Dynamically updating the learning rate and connection weights enables the model to track changes in water demand in real time and improve prediction accuracy.

[0029] Finally, there is the dynamic pricing and billing module. Conventional tiered pricing sets unit prices according to water consumption. The dynamic pricing and billing module of this application obtains predicted water prices through improved hybrid genetic algorithm optimization, taking into account multiple factors such as water use data, environmental factors, water supply network pressure and prediction errors, and its pricing rules are more flexible and scientific. For example, water prices may increase when water use is at its peak and the environment is bad, and may decrease when water prices are at their lowest and conditions are good. This can more accurately reflect changes in water supply costs, promote the rational use of water resources, and balance the interests of water supply companies and users. The first is the multi-dimensional optimization of the fitness function. Taking into account prediction errors, water use data, and environmental changes, the improved fitness function is designed as target optimization: ,in For the predicted water price, is the actual water price, is the predicted water supply network pressure, is the actual pressure, is the impact of environmental changes on water prices, expressed as the deviation of environmental parameter changes on the predicted water price. is the weight coefficient, which is used to adjust the priority of different targets. The method of determining is based on the contribution of each goal , which is defined as: , , ,in is the mean square error of the water price forecast error, is the mean square error of the water supply network pressure error, is the deviation of the impact of environmental changes on water prices, , , , ; Introduce nonlinear crossover in the crossover process to generate a new configuration space different from the parent generation, which is specifically expressed as: ,in, are the values ​​of the parent individual at the i-th hyperparameter, is the value of the hyperparameter of the offspring individual obtained by nonlinear crossover, It is a nonlinear adjustment factor that controls the nonlinear degree of crossover. The specific form is: ,in is the maximum nonlinear factor value, are the values ​​of the i-th hyperparameter for parent individuals 1 and 2, respectively. represents the difference between the two parent individuals in the hyperparameter; then introduce high-order nonlinear mutation to change the value range of the hyperparameter. The specific mutation formula is: ,in, is the value of the i-th hyperparameter in the parent individual, is the hyperparameter value of the offspring individual at the ith position obtained after the mutation operation, is a coefficient of variation , is the initial maximum coefficient of variation, is a nonlinear transformation function used to mutate the parent hyperparameters. In each generation of the genetic algorithm, the local particle swarm optimization is used to fine-tune the current best individual. The particle swarm position update formula takes into account the inertia weight and adaptive learning factor. Finally, when the maximum number of generations reaches the preset value of 30, the algorithm terminates, and the change in fitness value is as follows: Figure 2 As shown. At this time, since the fitness function is error-oriented, in order to ensure that the prediction error is minimized, the water price corresponding to the individual with the smallest fitness function in the population is the predicted water price. This predicted water price is the result of comprehensive consideration of water use data, environmental factors, water supply network pressure, and multiple iterative optimizations. It can reflect the reasonable level of future water prices as accurately as possible under the premise of meeting multi-objective optimization, and provide a scientific basis for dynamic pricing and billing. This method can effectively promote the rational use of water resources, ensure the benefits of water supply enterprises, enhance users' recognition of the rationality of charges, and promote the sustainable development of the water supply industry.

[0030] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A smart water supply revenue system, characterized in that: Including data collection module, intelligent scheduling module and dynamic pricing and billing module; The data collection module uses a smart water meter with wireless communication function to collect environmental factors and user water use data in real time, including but not limited to water flow, water use time, and water supply network pressure, and transmits the collected real-time data to the central data processor; The intelligent scheduling module uses a multi-level adaptive neural network to predict water supply network pressure, environmental factors, water use time and water flow according to user water demand, and schedules water supply resources; The dynamic pricing and billing module is based on real-time collected water use data and precipitation factors, and uses an improved hybrid genetic algorithm to optimize and derive a predicted price; The specific implementation steps of the optimization using the improved hybrid genetic algorithm are: Step 1: The first step is to optimize the fitness function in multiple dimensions. Taking into account the prediction error, water use data, and environmental change factors, the improved fitness function is: ,in For the predicted water price, is the actual water price, is the predicted water supply network pressure, is the actual pressure, is the impact of environmental changes on water prices, expressed as the deviation of environmental parameter changes on the predicted water price. is the weight coefficient, which is used to adjust the priority of different goals; Step 2: Introduce nonlinear crossover in the crossover process to generate a new configuration space different from the parent generation, which is specifically expressed as: ,in, are the values ​​of the parent individual at the i-th hyperparameter, is the value of the hyperparameter of the offspring individual obtained by nonlinear crossover, It is a nonlinear adjustment factor that controls the nonlinear degree of crossover; Step 3: Then introduce high-order nonlinear mutation to change the value range of hyperparameters. The specific mutation formula is: ,in, is the value of the i-th hyperparameter in the parent individual, is the hyperparameter value of the offspring individual at the ith position obtained after the mutation operation, is a coefficient of variation , is the initial maximum coefficient of variation, is a nonlinear transformation function used to mutate the parent hyperparameters; Step 4: In each generation of the genetic algorithm, the best individual is fine-tuned using local particle swarm optimization. The particle swarm position update formula takes into account the inertia weight and adaptive learning factor. Step 5: Finally, when the maximum number of generations reaches the preset value of 30, the algorithm terminates.

2. The intelligent water supply revenue system according to claim 1, characterized in that: The dynamic pricing and billing module adopts the improved hybrid genetic algorithm to optimize the weight coefficient in step 1 The method of determining is based on the contribution of each goal , which is defined as: , , ,in is the mean square error of the water price forecast error, is the mean square error of the water supply network pressure error, is the deviation of the impact of environmental changes on water prices, , , , .

3. The intelligent water supply revenue system according to claim 1, characterized in that: The nonlinear adjustment factor in step 2 of the optimization of the dynamic pricing and billing module using an improved hybrid genetic algorithm The specific form is: ,in is the maximum nonlinear factor value, are the values ​​of the i-th hyperparameter for parent individuals 1 and 2, respectively. Represents the difference between the two parent individuals in this hyperparameter.

4. The intelligent water supply revenue system according to claim 1, characterized in that: The multi-level adaptive neural network in the intelligent scheduling module is specifically implemented as follows: Step 1: First, standardize the collected data and remove outliers to ensure data stability; Step 2: Construct a multi-level adaptive neural network, which is divided into three sub-network levels. Each layer processes different types of information and the number of neurons in each layer and their connection mode can be adaptively adjusted according to the changes in real-time input; Step 3: Finally, the learning rate of each layer is dynamically adjusted according to the gradient changes in the training cycle. The learning rate update formula is: ,in is the gradient of the current cycle, are the gradients of cycle 0 and the current one, respectively. is the current target loss value, is the predetermined optimal target loss value.

5. The intelligent water supply revenue system according to claim 4, characterized in that: The three sub-network levels of the multi-level adaptive neural network in step 2 of the intelligent scheduling module are: the first layer uses a convolutional neural network to process environmental data and extract relevant features; the second layer uses an LSTM model to process time series data, which is a demand prediction network; and the third layer is a scheduling optimization network, which is combined with a support vector machine to adjust the decision of the water supply system.

Images