A smart energy-saving system and method based on an active decision-making supply-demand balance model

Abstract

This invention belongs to the field of intelligent control technology for water supply networks, specifically disclosing an intelligent energy-saving system and method based on an active decision-making supply-demand balance model. The system includes: a pressure data processing module for cleaning historical pressure data and extracting the mean and variance characteristics of pressure loss; a pressure loss prediction module that uses a neural network model to predict the future mean and variance of pressure loss and identify peak water usage periods accordingly; and an outlet pressure game decision module that establishes a dynamic game model between the pump station and users, where the pump station aims to minimize outlet pressure, and user complaints are triggered by a probability distribution of end-point pressure. The system calculates the optimal outlet pressure value that maximizes energy saving while ensuring user satisfaction by solving the game equilibrium. This invention enables refined and intelligent control of water supply pressure, achieving significant energy savings while ensuring water supply stability during peak periods.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for water supply networks, specifically relating to an intelligent energy-saving system and method based on an active decision-making supply and demand balance model. Background Technology

[0002] Urban municipal water supply systems are critical infrastructure for ensuring urban water supply. However, traditional water supply systems generally suffer from high energy consumption during operation. The root cause lies in the fact that pump stations typically employ fixed pressure or simple empirical pressure control strategies. Under this strategy, the outlet pressure of the pump station remains constant or undergoes only coarse segmented adjustments throughout the 24 hours of the day. Whether during off-peak hours such as early morning when water consumption is extremely low, or during peak hours when water consumption is concentrated in the morning and evening, the system supplies water at the same or similar high pressure.

[0003] This control method has significant drawbacks: maintaining high pressure during periods of low water demand results in substantial energy waste, directly increasing the operating costs of water supply companies; maintaining unnecessary high pressure over a long period increases the operational load on pipe networks and valves, accelerating equipment aging and even increasing the risk of pipe network leakage; it is difficult to find a dynamic and precise balance between the conflicting goals of "energy conservation and consumption reduction" and "ensuring user satisfaction with water use." Simply reducing pressure may lead to user complaints during peak hours, while maintaining high pressure inevitably results in excessive energy consumption.

[0004] Therefore, how to intelligently and dynamically adjust the water pressure of the pump room in real time according to the actual water demand of users, and maximize energy saving and consumption reduction while ensuring user experience, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] Purpose of the invention: The present invention aims to solve the problems existing in the above-mentioned background technology, and provide an intelligent energy-saving system and method based on an active decision-making supply and demand balance model. It can accurately predict changes in users' water demand, and on this basis, calculate the optimal water pressure through a game model, thereby achieving significant energy saving and energy consumption reduction of the water supply system without reducing users' water satisfaction.

[0006] Technical Solution: The intelligent energy-saving system based on an active decision-making supply-demand balance model of the present invention includes a pressure data processing module, configured to acquire historical data of pump station outlet pressure and user terminal pressure, perform outlier processing, and calculate historical time series data of pressure loss, wherein the historical time series data of pressure loss includes the average pressure loss representing the average water load of users and the variance of pressure loss representing the intensity of water fluctuations; a pressure loss prediction module, configured to predict the average pressure loss and variance of pressure loss within a predetermined future time period based on the historical time series data of the average pressure loss and variance of pressure loss, and identify peak water consumption periods based on the predicted variance of pressure loss; and an outlet pressure game decision module, configured to establish a dynamic game model between the pump station and users, wherein the pump station aims to minimize outlet pressure, and user complaints are triggered based on the received terminal pressure according to a preset complaint probability distribution. By solving the game model, an optimal outlet pressure value for balancing energy saving and user water satisfaction is determined, and this value is output to control the pump station.

[0007] To further improve the above technical solution, the pressure data processing module uses the quartile method for outlier handling. Specifically, within a preset translation time window, it calculates the first quartile Q1 and the third quartile Q3, and then handles outliers exceeding the specified values. Data points within a certain range are identified as outliers and filled using forward padding, where IQR is the interquartile range.

[0008] Furthermore, the pressure loss prediction module includes two independent neural network models: a first LSTM neural network model for predicting the future mean based on historical data of the pressure loss mean; and a second LSTM neural network model for predicting the future variance based on historical data of the pressure loss variance.

[0009] Furthermore, both the first LSTM neural network model and the second LSTM neural network model are long short-term memory neural network models. The long short-term memory neural network model includes an input layer for receiving historical stress loss data, an LSTM layer containing multiple LSTM units, and a fully connected output layer using a linear activation function, and is trained using mean squared error as the loss function.

[0010] Furthermore, the pressure loss prediction module identifies peak water usage periods by: determining one or more local maxima as peak reference points in the predicted future pressure loss variance; multiplying the value of the peak reference points by a preset threshold coefficient to generate a dynamic threshold; and marking the periods when the predicted pressure loss variance exceeds the dynamic threshold as peak water usage periods.

[0011] Furthermore, the water pressure game decision module is configured to: model the pressure loss as a random variable that follows a normal distribution with the mean and variance of the prediction; and calculate the total number of complaints by performing Monte Carlo simulations on multiple simulated users, wherein the simulation includes randomly generating a pressure loss value from the normal distribution for each simulated user.

[0012] Furthermore, in the Monte Carlo simulation, the complaint behavior of each simulated user is modeled using a Sigmoid function, which includes a shift parameter that is activated only during identified peak water usage periods to increase the end-pressure threshold that triggers complaints during peak periods.

[0013] Furthermore, the total number of complaints is obtained by performing Bernoulli trials on the complaint events of each simulated user and summing the results.

[0014] Furthermore, the water pressure game decision module solves the game model iteratively. The iterative method includes: if the number of user complaints calculated based on the current water pressure and the predicted pressure loss is greater than zero, then the water pressure is increased according to the following formula: ;in, The preset adjustment step size, The total number of complaints is given; and this process is repeated until the calculated number of complaints is zero or the water pressure reaches the preset upper limit.

[0015] The intelligent energy-saving method using the above-mentioned intelligent energy-saving system based on the active decision-making supply-demand balance model includes the following steps:

[0016] Data processing steps: Obtain historical data on pump station outlet pressure and user terminal pressure, perform outlier processing, and calculate historical time series data of pressure loss. This pressure loss data includes the mean pressure loss representing the average water load of users and the variance of pressure loss representing the intensity of water fluctuations.

[0017] Loss prediction steps: Based on the historical time series data of the pressure loss mean and variance, a pre-trained neural network model is used to predict the pressure loss mean and pressure loss variance within a predetermined time period in the future, and the peak water consumption period is identified based on the predicted pressure loss variance.

[0018] Game theory decision-making and control steps: Establish a dynamic game model between the pump station and the user, in which the pump station aims to minimize the outlet water pressure, and the user's complaint behavior is triggered based on the received terminal pressure according to a preset complaint probability distribution. By solving the game model, an optimal outlet water pressure value that balances energy saving and consumption reduction with user water satisfaction is obtained, and the pump station is controlled according to this value.

[0019] Beneficial Effects: Compared with existing technologies, the advantages of this invention are as follows: This invention learns historical water usage patterns through LSTM neural networks, enabling accurate prediction of future water load and fluctuation intensity, transforming the system from passive response to proactive prediction, and achieving forward-looking intelligent control; it dynamically adjusts the outlet water pressure based on the prediction results, significantly reducing pipeline pressure during low water usage periods such as late night and midday, while ensuring sufficient supply during peak periods, resulting in significant energy savings compared to traditional constant pressure control strategies; the user complaint prediction mechanism based on game theory and probability models quantifies the abstract "user satisfaction" into a calculable complaint probability, ensuring that user water experience is effectively guaranteed while saving energy and reducing consumption; through game decision-making, it automatically finds the optimal balance between energy consumption and satisfaction, with "zero complaints" as the game equilibrium goal, automatically maintaining or increasing the pressure to a safe level during peak water usage periods, ensuring that user water experience is not affected, and adapting to seasonal or long-term changes in user water usage habits, exhibiting high intelligence and robustness. Attached Figure Description

[0020] Figure 1 This is an architecture diagram of an intelligent energy-saving system based on an active decision-making supply and demand balance model in an embodiment of the present invention.

[0021] Figure 2 This is a comparison chart of the actual and predicted values ​​of the pressure loss variance in an embodiment of the present invention.

[0022] Figure 3 This is a distribution map of peak and normal periods based on pressure loss variance identification in an embodiment of the present invention.

[0023] Figure 4 This is a comparison chart of the original fixed outlet water pressure and the optimized dynamic outlet water pressure in the embodiments of the present invention. Detailed Implementation

[0024] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments described.

[0025] Example 1: The intelligent energy-saving system based on the proactive decision-making supply-demand balance model provided in this example, such as... Figure 1 As shown, the overall architecture includes a pressure data processing module, a pressure loss prediction module, and an outlet pressure game decision module, which together constitute an intelligent control system.

[0026] 1. Pressure data processing module

[0027] This module forms the system's data foundation, responsible for cleaning and feature extraction of the raw data. It samples the effluent pressure per second. and end pressure The data is processed using a translation window method, with a window size of 10 minutes. The maximum threshold is calculated for the data within each window using the quartile method. and minimum threshold Outlier detection is performed, and the specific method is as follows:

[0028] ;

[0029] in, It is the first quartile (25th percentile). It is the third quartile (75th percentile). Values ​​exceeding the threshold are considered outliers. or Outliers are marked as missing values, and forward imputation is used to fill them in:

[0030] ;

[0031] in, This represents the most recent valid data point before the missing value.

[0032] For identified outliers, a forward filling method is used, where the outlier is replaced with the value of the nearest valid data point preceding it. After data cleaning, the outlier is determined based on the effluent pressure. and end pressure Calculate the pressure loss per second, pressure loss value. This reflects the user's water consumption and pressure loss caused by the water transmission process during that time period. The specific calculation formula is:

[0033] ;

[0034] Treating pressure loss as a random variable and setting the window size to 1 minute, calculate the pressure loss, mean, and variance. The mean pressure loss reflects the average water load per minute, and the specific calculation formula is:

[0035] ;

[0036] in, The number of data points per minute is generally... The pressure loss variance reflects the intensity of water usage fluctuations per minute, and is calculated using the following formula:

[0037] ;

[0038] Finally, the module outputs historical time series data of the mean and variance of pressure loss.

[0039] 2. Pressure Loss Prediction Module

[0040] Pressure loss prediction consists of two core steps: neural network prediction and peak period identification. Neural networks are established to predict the mean and variance of pressure loss.

[0041] (1) Neural network prediction

[0042] An LSTM neural network was constructed using the mean and variance of the pressure loss. Historical data was divided into multiple data segments using a sliding window, with each segment containing 11 data points. These segments constituted the dataset for the LSTM neural network. For each prediction point, the input layer of the LSTM neural network consisted of the pressure loss data from the previous day's corresponding time point (10 minutes prior), indicating that the model would use pressure loss data from the past 10 minutes (10 time steps) to predict the pressure loss at that point. The LSTM layer contained 20 LSTM units, and the output of the last time step was used as the predicted value. A fully connected layer with a single neuron was added at the top of the model, using a linear activation function to map the features extracted by the LSTM layer to the final predicted value. The model used mean squared error as the loss function, which effectively measures the deviation between the predicted and actual values. The specific formula is as follows:

[0043] ;

[0044] in, Indicates the predicted value. Represents the actual value.

[0045] The mean squared error (MSE) was used as the loss function for training until the model could accurately reproduce the fluctuation patterns of historical data. (Refer to...) Figure 2 The figure shows a high degree of fit between the predicted values ​​(orange dots) and the actual values ​​(green dots) of the pressure loss variance.

[0046] After training a neural network using pressure loss data, the previous day's data is used as input. The trained neural network is then used to predict the pressure loss data for the following day, yielding the predicted values ​​for the mean and variance of pressure loss. The specific formulas are as follows:

[0047] ;

[0048] ;

[0049] in, It is a neural network trained using mean pressure loss data. It is a neural network trained using pressure loss variance data.

[0050] (2) Peak period identification and smoothing processing

[0051] Peak detection is performed on the predicted data. The larger of the two largest local maxima is taken as the peak value of the daily water consumption peak, denoted as . The range exceeding a certain peak value is considered a peak time period and is marked as a peak to generate a dynamic threshold. The specific formula is as follows:

[0052] ;

[0053] in, express Time markers, The threshold coefficient is typically set to [value missing]. Any period in which the prediction variance exceeds this dynamic threshold is marked by the system as a peak water usage period, as per [reference]. Figure 3 The figure clearly shows the distribution of the identified peak periods (red dots) and normal periods (blue dots).

[0054] 3. Water pressure game decision-making module

[0055] By employing a dynamic game theory model between the pump station and users, the water pressure is reduced while ensuring user satisfaction with water usage. The specific process includes the following steps:

[0056] (1) Game environment

[0057] The pump house can supply water at a certain outlet pressure, and the pressure regulation range of the pump house is set as follows: .in, To ensure a good water experience for residents, the original outlet water pressure before optimization was set. This serves as the system's safety lower limit, preventing excessively drastic water pressure drops and increasing the risk of system complaints. It is based on the average pressure loss output from the pressure loss prediction module. and variance A stochastic pressure loss model is constructed, in which the pressure loss follows a normal distribution, reflecting the statistical characteristics of the pressure loss during the prediction period. The specific form is as follows:

[0058] ;

[0059] The terminal pressure reaching the user after pressure loss is:

[0060] ;

[0061] Users can decide whether to file a complaint based on the pressure at the end of the chain. The lower the pressure at the end, the higher the probability of a user filing a complaint. The sigmoid function is used to characterize the distribution function of the user's complaint probability, and its specific expression is as follows:

[0062] ;

[0063] in, This indicates that the user has filed a complaint. It is an indicator function for peak hours. This is a shift parameter for the peak-hour distribution function, typically set to 0.05, to prevent users from spending too much time during peak water usage periods throughout the day. Introducing this shift parameter during peak hours ensures a better user experience during these times.

[0064] (2) Pressure loss and user complaints

[0065] In each round of the game, a Monte Carlo simulation is performed to... Each virtual user generates a stress loss, which is a random variable that follows a normal distribution.

[0066] ;

[0067] Pressure loss is throttled to ensure its reasonableness. Based on the current pump house outlet pressure and pressure loss Calculate the terminal pressure reaching each user. and the probability of user complaints :

[0068] ;

[0069] ;

[0070] Complaint events are generated using Bernoulli trials; random variables... Indicates user Whether to file a complaint, and Follows a 0-1 distribution. The total number of user complaints can be expressed as follows:

[0071] ;

[0072] Each decision-making moment involves determining the final water pressure through game theory. The time to complete a full game equilibrium calculation is within 1 minute, which can meet the regulation requirements of the water supply system.

[0073] (3) Pressure regulation and equilibrium judgment

[0074] When there are user complaints At that time, the pump station increases the outlet water pressure proportionally:

[0075] ;

[0076] in, The pressure adjustment step size is defined as small steps, which allows for a lower outlet water pressure while meeting users' water needs. This process is repeated until the game reaches equilibrium with zero complaints. Or the outlet water pressure reaches the preset upper limit Pmax, that is The water pressure obtained at this point is the optimal solution for the current period, which is the lowest energy consumption pressure while ensuring user satisfaction.

[0077] like Figure 4 As shown, after optimization, the outlet water pressure (red dot) is significantly lower than the original fixed pressure (blue dot) during off-peak periods such as late night and midday, achieving deep energy saving; while during morning and evening peak periods, the pressure can quickly rise to a safe level, ensuring the demand for centralized water use.

[0078] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. An intelligent energy-saving system based on an active decision-making supply-demand balance model, characterized in that, include: The pressure data processing module is configured to acquire historical data of pump station outlet pressure and user terminal pressure, perform outlier processing, and calculate historical time series data of pressure loss. The historical time series data of pressure loss includes the mean pressure loss representing the average water load of users and the variance of pressure loss representing the intensity of water fluctuations. The pressure loss prediction module is configured to predict the mean and variance of pressure loss within a predetermined future time period based on historical time series data of the mean and variance of pressure loss using a pre-trained neural network model. It then identifies peak water usage periods based on the predicted variance. The identification of peak water usage periods by the pressure loss prediction module includes: determining one or more local maxima as peak reference points in the predicted future pressure loss variance; multiplying the value of the peak reference points by a preset threshold coefficient to generate a dynamic threshold; and marking periods where the predicted pressure loss variance exceeds the dynamic threshold as peak water usage periods. The water pressure game decision module is configured to establish a dynamic game model between the pump station and users. The pump station aims to minimize the water pressure at the outlet, while user complaints are triggered based on the received terminal pressure according to a preset complaint probability distribution. By solving the game model, an optimal water pressure value is determined to balance energy conservation and user satisfaction with water usage, and this value is output to control the pump station. The water pressure game decision module calculates the total number of complaints by performing Monte Carlo simulations on multiple simulated users. In the Monte Carlo simulation, the complaint behavior of each simulated user is modeled using a Sigmoid function. The Sigmoid function includes a shift parameter that is activated only during identified peak water usage periods, used to increase the terminal pressure threshold that triggers complaints during peak periods.

2. The intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 1, characterized in that, The pressure data processing module uses the quartile method for outlier handling. Specifically, within a preset translation time window, it calculates the first quartile Q1 and the third quartile Q3, and then handles outliers exceeding the specified values. Data points within a certain range are identified as outliers and filled using forward padding, where IQR is the interquartile range.

3. The intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 1, characterized in that, The pressure loss prediction module includes two independent neural network models: a first LSTM neural network model for predicting the future mean based on historical data of the pressure loss mean; and a second LSTM neural network model for predicting the future variance based on historical data of the pressure loss variance.

4. The intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 3, characterized in that, Both the first LSTM neural network model and the second LSTM neural network model are long short-term memory neural network models. The long short-term memory neural network model includes an input layer for receiving historical stress loss data, an LSTM layer containing multiple LSTM units, and a fully connected output layer using a linear activation function, and is trained using mean squared error as the loss function.

5. The intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 1, characterized in that, The water pressure game decision module is configured to: model the pressure loss as a random variable that follows a normal distribution with the mean and variance of the prediction; and the Monte Carlo simulation includes randomly generating pressure loss values ​​from the normal distribution for each simulated user.

6. The intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 5, characterized in that, The total number of complaints was obtained by performing Bernoulli trials on each simulated user's complaint events and summing the results.

7. The intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 1, characterized in that, The water outlet pressure game decision module solves the game model iteratively. The iterative method includes: if the number of user complaints calculated based on the current water outlet pressure and the predicted pressure loss is greater than zero, then the water outlet pressure is increased according to the following formula: ;in, The preset adjustment step size, The total number of complaints is given; and this process is repeated until the calculated number of complaints is zero or the water pressure reaches the preset upper limit.

8. The intelligent energy-saving method for an intelligent energy-saving system based on an active decision-making supply-demand balance model according to claim 1, characterized in that, Includes the following steps: Data processing steps: Obtain historical data on pump station outlet pressure and user terminal pressure, perform outlier processing, and calculate historical time series data of pressure loss. This pressure loss data includes the mean pressure loss representing the average water load of users and the variance of pressure loss representing the intensity of water fluctuations. Loss prediction steps: Based on the historical time series data of the pressure loss mean and variance, a pre-trained neural network model is used to predict the pressure loss mean and pressure loss variance within a predetermined time period in the future, and the peak water consumption period is identified based on the predicted pressure loss variance. Game theory decision-making and control steps: Establish a dynamic game model between the pump station and the user, in which the pump station aims to minimize the outlet water pressure, and the user's complaint behavior is triggered based on the received terminal pressure according to a preset complaint probability distribution. By solving the game model, an optimal outlet water pressure value is determined to balance energy saving and consumption reduction with user water satisfaction, and the pump station is controlled according to this value.

Images