Whole-house water purification system control optimization method and platform based on dynamic regulation

Abstract

The present application provides a whole-house water purification system control optimization method and platform based on dynamic adjustment, which relate to the field of control optimization technology. A plurality of state variables are determined by analyzing the operation changes of the whole-house water purification system; a variable dynamic network is established and data fitting is performed to construct a dynamic model of the whole-house water purification system to obtain control decisions; reinforcement learning is performed on the dynamic model based on state variables and control action information, a predictive controller is constructed, and control decisions are dynamically optimized to obtain control optimization decisions; the control optimization decisions are executed for feedback dynamic adjustment, and the whole-house water purification system is intelligently controlled and optimized based on a dynamic adjustment parameter set, which solves the technical problems of low control efficiency, slow response speed and poor stability of water quality changes in the whole-house water purification system, and difficulty in meeting the efficient and accurate needs of modern families for safe water use, and achieves the technical effect of improving the control efficiency, response speed and stability of the water purification system and ensuring the safety of family water use.

Description

Technical Field

[0001] The present invention relates to the field of control optimization technology, and in particular to a control optimization method and platform for a whole-house water purification system based on dynamic regulation. Background Art

[0002] With the increase in household water demand and the increasing attention paid to water safety issues, whole-house water purification systems have gradually become an important device for modern families. However, with the changes in lifestyles and the diversification of water pollution sources, traditional whole-house water purification systems face many challenges in control and management. Current whole-house water purification systems usually rely on fixed filtering and processing algorithms and work according to preset parameters such as water source quality and water consumption. While providing basic water purification functions, these systems often lack flexibility and real-time adjustment capabilities and cannot adapt to changing water quality needs.

[0003] In summary, the existing whole-house water purification system has technical problems such as low efficiency in controlling water quality changes, slow response speed and poor stability, making it difficult to meet the efficient and accurate needs of modern families for safe water use. Summary of the invention

[0004] The present application provides a whole-house water purification system control optimization method and platform based on dynamic adjustment, which is used to solve the technical problems that the existing whole-house water purification system has low control efficiency of water quality changes, slow response speed and poor stability, and is difficult to meet the efficient and accurate needs of modern families for safe water use.

[0005] In view of the above problems, the present application provides a whole-house water purification system control optimization method and platform based on dynamic regulation.

[0006] In a first aspect, the present application provides a whole-house water purification system control optimization method based on dynamic regulation, the method comprising:

[0007] Based on the operation change analysis of the whole-house water purification system, a plurality of state variables are determined, and the plurality of state variables have a plurality of control action information; a variable dynamic network is established according to the plurality of state variables, and the variable dynamic network is used to perform data fitting, so as to construct a dynamic model of the whole-house water purification system and obtain control decisions; reinforcement learning is performed on the dynamic model based on the plurality of state variables and the plurality of control action information, and a predictive controller is constructed; control decisions are dynamically optimized through the predictive controller to obtain control optimization decisions; the control optimization decisions are executed to perform feedback dynamic adjustment, and intelligent control optimization of the whole-house water purification system is performed based on a dynamic adjustment parameter set.

[0008] In the second aspect, the present application provides a whole-house water purification system control optimization platform based on dynamic regulation, the platform comprising:

[0009] A system operation analysis module is used to perform operation change analysis based on the whole-house water purification system and determine multiple state variables, wherein the multiple state variables have multiple control action information; a dynamic model construction module is used to establish a variable dynamic network based on the multiple state variables, use the variable dynamic network to perform data fitting, build a dynamic model of the whole-house water purification system, and obtain control decisions; a predictive controller construction module is used to perform reinforcement learning on the dynamic model based on the multiple state variables and the multiple control action information, and build a predictive controller; a control optimization module is used to dynamically optimize the control decision through the predictive controller to obtain a control optimization decision; a feedback adjustment module is used to execute the control optimization decision for feedback dynamic adjustment, and perform intelligent control optimization on the whole-house water purification system based on the dynamic adjustment parameter set.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The control optimization method of the whole-house water purification system based on dynamic adjustment provided in the present application determines a plurality of state variables by performing operation change analysis based on the whole-house water purification system, wherein the plurality of state variables have a plurality of control action information; establishes a variable dynamic network according to the plurality of state variables, uses the variable dynamic network to perform data fitting, constructs a dynamic model of the whole-house water purification system, and obtains control decisions; performs reinforcement learning on the dynamic model based on the plurality of state variables and the plurality of control action information, and constructs a predictive controller; dynamically optimizes the control decision through the predictive controller to obtain a control optimization decision; executes the control optimization decision to perform feedback dynamic adjustment, and performs intelligent control optimization of the whole-house water purification system based on a dynamic adjustment parameter set, thereby solving the technical problems of low control efficiency, slow response speed and poor stability of water quality changes in the existing whole-house water purification system, and difficulty in meeting the efficient and accurate needs of modern families for safe water use, and achieves the technical effect of improving the control efficiency, response speed and stability of the water purification system and ensuring the safety of family water use. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flow chart of a whole-house water purification system control optimization method based on dynamic regulation is provided for this application.

[0013] Figure 2 A structural schematic diagram of a whole-house water purification system control optimization platform based on dynamic regulation is provided for this application.

[0014] Explanation of the reference numerals: system operation analysis module 11 , dynamic model construction module 12 , prediction controller construction module 13 , control optimization module 14 , feedback adjustment module 15 . DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0016] Embodiment 1, as Figure 1 As shown, the present application provides a whole-house water purification system control optimization method based on dynamic regulation, the method comprising:

[0017] Step S100: performing an operation change analysis based on the whole-house water purification system to determine a plurality of state variables, wherein the plurality of state variables have a plurality of control action information.

[0018] Specifically, a comprehensive analysis of the operation changes of the whole-house water purification system is performed to determine multiple state variables and obtain corresponding control action information. Among them, the state variables refer to key parameters that can reflect the current operating status of the system, including water quality indicators (such as turbidity, residual chlorine content), flow, pressure value and temperature, etc. These state variables provide basic data for the dynamic adjustment of the system, which in turn affects the control strategy of the water purification process. In this process, data collection is first performed, and the changes of state variables are monitored in real time using intelligent sensors to ensure the accuracy and timeliness of the acquired data. Subsequently, by analyzing these data, the key change patterns in the operation of the system are identified. This process includes statistical analysis and trend prediction of the data to determine when corresponding control actions need to be taken. For example, when the turbidity in the water is monitored to exceed the preset threshold, the system will automatically identify this change and generate corresponding control action information, such as increasing the filtration rate or replacing the filter element. Among them, a state variable may have multiple control action information, and a control action information may also affect multiple state variables, and multiple state variables have multiple control action information. Through this in-depth analysis of state variables and intelligent association of control actions, the efficient operation of the whole-house water purification system in different usage scenarios is ensured, thereby meeting the family's demand for safe drinking water. Through systematic analysis of operational changes, not only is basic data for dynamic adjustment provided for the whole-house water purification system, but a solid foundation is also laid for subsequent control decisions.

[0019] Step S200: establishing a variable dynamic network according to the multiple state variables, using the variable dynamic network to perform data fitting, constructing a dynamic model of the whole-house water purification system, and obtaining a control decision.

[0020] Optionally, after determining multiple state variables, the relationship between these state variables is further systematically represented by establishing a variable dynamic network. A variable dynamic network refers to a graphical model that displays the correlation between each state variable in the form of nodes and edges to facilitate understanding and analysis of their mutual influence. The establishment of a variable dynamic network is intended to provide a structured basis for subsequent data fitting and dynamic model construction. Specifically, the nodes of the variable dynamic network are first constructed using the collected real-time data. Each node represents a state variable, and the edges represent the relationship between the variables, such as causal relationship or correlation. Next, by applying data fitting techniques, such as linear regression or nonlinear models, the state variables are associated with system outputs (such as water quality, flow, etc.). For example, when it is found that there is a strong correlation between water pressure and water flow, the dynamic network will form a corresponding mathematical model through fitting analysis. This model can accurately describe how water flow is affected by changes in water pressure, thereby providing an important basis for control decisions. Through the above steps, the dynamic model of the whole-house water purification system is finally obtained. This dynamic model can not only reflect the current state of the system in real time, but also predict future operating trends based on the input state variables, providing a scientific basis for control decisions. By analyzing the dynamic model, the system can generate corresponding control strategies to achieve intelligent optimization and precise management of the water purification process, ensuring safe and efficient household water use.

[0021] Step S300: Perform reinforcement learning on the dynamic model based on the multiple state variables and the multiple control action information to construct a predictive controller.

[0022] Exemplarily, after the construction of the dynamic model is completed, the dynamic model is trained by using a reinforcement learning method based on the multiple state variables and control action information to build an efficient predictive controller. Reinforcement learning is a machine learning technique that aims to optimize decision-making strategies through interaction with the environment to maximize the long-term benefits of the system. Specifically, the dynamic model is first regarded as an environment, in which the state variables are the current state of the system, and the control action information represents the actions that the system can take. In this process, the controller gradually learns how to make optimal decisions in different situations by exploring the impact of different control actions on the system state. The key steps of this process include defining a reward function to measure the impact of each control action on the system performance. For example, when the operation selected by the controller successfully reduces the turbidity of the water quality, the system will give a positive reward, otherwise it will give a penalty to guide the learning process. Through repeated trial and error and learning, the controller can gradually master the best control strategy under a specific state and then form a predictive model. This predictive controller can predict the future system state based on real-time monitoring data and make corresponding control decisions. For example, when the water quality is detected to be degraded, the predictive controller will immediately recommend increasing the filtration rate or adjusting other control parameters to optimize the water purification effect. Ultimately, the constructed predictive controller not only has real-time response capabilities, but can also adapt to the diverse needs of household water use through continuous learning and optimization, thereby ensuring the efficient operation and water quality safety of the whole-house water purification system under various conditions.

[0023] Step S400: dynamically optimizing the control decision through the predictive controller to obtain a control optimization decision.

[0024] Furthermore, the controller dynamically optimizes the control decision to obtain the best control optimization decision. Among them, dynamic optimization refers to the continuous adjustment of the control strategy based on the real-time monitored state variables and the output of the prediction model to maximize the system performance. After receiving the real-time data from the whole-house water purification system, the predictive controller first evaluates the current state, including key indicators such as water quality parameters, flow rate and pressure. On this basis, the controller applies optimization algorithms such as model predictive control (MPC) or genetic algorithms to analyze the potential effects of different control actions and generate a series of possible control decisions. In this process, the controller not only considers the current state, but also predicts possible future states to ensure the effectiveness and foresight of the decision. For example, when the predictive controller detects that the water quality is gradually deteriorating, it may calculate the pros and cons of multiple control options such as increasing the filtration intensity, extending the water purification time, or adjusting the additive dosage. Through dynamic optimization, the controller will select the decision that has the greatest impact on the system performance and form a control optimization decision. Dynamic optimization of control decisions through predictive controllers not only improves the response speed and intelligence level of the whole-house water purification system, but also ensures water quality safety and user satisfaction, providing a more efficient water solution for modern families.

[0025] Step S500: Execute the control optimization decision to perform feedback dynamic adjustment, and perform intelligent control optimization on the whole-house water purification system based on the dynamic adjustment parameter set.

[0026] Specifically, after obtaining the control optimization decision, feedback dynamic adjustment is implemented by executing the decision to further optimize the intelligent control of the whole-house water purification system. The feedback dynamic adjustment refers to adjusting the system parameters in real time according to the execution effect of the control optimization decision to ensure the optimal performance of the system under various operating conditions. Specifically, after executing the control optimization decision, the system will immediately implement the corresponding control actions, such as adjusting the operating speed of the pump, changing the working mode of the filter, or modifying the dosage of the agent, etc. These actions are designed to quickly respond to the current water quality requirements and ensure the effectiveness of the water purification process. In this process, the system uses dynamic adjustment parameter sets, which contain pre-set optimal operating conditions and response strategies, covering adjustment plans under different water quality changes. For example, if the real-time monitoring data shows that the water quality has not improved significantly after the decision is executed, the system will analyze the effect of the current control action through the feedback mechanism and compare it with the expected effect. If it is found that the water quality does not meet the target standard, the system will automatically adjust the relevant parameters, such as increasing the filtration time or increasing the water pressure, to enhance the purification effect. In addition, feedback dynamic adjustment also includes continuous monitoring of state variables during the execution process to ensure that the system can flexibly respond to environmental changes. Through this real-time adjustment, not only can the current operating status be optimized, but also future control strategies can be continuously learned and improved, thereby improving overall operating efficiency. In summary, by executing control optimization decisions and performing feedback dynamic adjustment, the whole-house water purification system achieves intelligent control optimization, ensuring that families can obtain safe and high-quality water under various conditions of use, which not only improves the flexibility and stability of the system, but also enhances user trust and satisfaction.

[0027] Furthermore, according to the plurality of state variables, a variable dynamic network is established, and step S200 of the present application further includes:

[0028] Step S210: monitoring the multiple state variables according to the operation cycle of the whole-house water purification system, and determining multiple operation time nodes, wherein the multiple operation time nodes correspond to the multiple state variables.

[0029] Step S220: Arrange the multiple state variables at the multiple running time nodes in time sequence to obtain a running time sequence.

[0030] Step S230: performing correlation analysis on the multiple state variables based on the running time series, and extracting multiple dynamic relationships.

[0031] Step S240: construct the variable dynamic network by taking the multiple dynamic relationships as edges and the multiple state variables as points.

[0032] Optionally, multiple state variables are comprehensively monitored, and the monitoring process is carried out according to the operation cycle of the whole-house water purification system. The key to this stage is to determine multiple operation time nodes, which not only represent the state of the system at different times, but also have a clear correspondence with the state variables. For example, the values ​​of state variables such as water quality parameters, flow and pressure at specific time points can be recorded during monitoring. Next, the multiple operation time nodes determined are arranged in time sequence to form a continuous operation time series, which provides a time dimension for subsequent analysis and ensures that the operation changes of the system in different time periods can be effectively captured. In this way, the system can reveal the law of state variables changing over time and lay the foundation for subsequent analysis. Then, the correlation analysis of state variables is performed based on the obtained operation time series. This process adopts dynamic correlation analysis methods, such as cross-correlation, Granger causality test or mutual information, to quantify the dependency between state variables. Through this analysis, multiple dynamic relationships can be extracted to identify which state variables have significant correlations with each other. For example, the relationship between water pressure and flow may indicate that when water pressure increases, flow will also increase, thereby suggesting possible directions for optimizing control strategies. Finally, the extracted multiple dynamic relationships are used as edges and multiple state variables are used as points to construct the variable dynamic network. At this time, the nodes represent the key state variables in the system, and the edges represent the dynamic relationships between them. This graphical representation not only facilitates the understanding of the interaction between the state variables, but also provides an intuitive structured framework for subsequent data fitting and dynamic model construction. By establishing a variable dynamic network, a solid foundation is provided for the intelligent control optimization of the whole-house water purification system, and important support is provided for subsequent dynamic adjustment and control decisions.

[0033] Furthermore, the multiple dynamic relationships are used as edges and the multiple state variables are used as points to construct a variable dynamic network. Step S240 of the present application further includes:

[0034] Step S241: weighting the multiple dynamic relationships based on the multiple state variables to obtain multiple weight coefficients.

[0035] Step S242: Continuously updating the multiple weight coefficients based on the running time sequence to determine multiple dynamic weight coefficients.

[0036] Step S243: construct a sliding window, and use the sliding window in combination with the multiple dynamic weight coefficients to connect the multiple state variables according to the multiple dynamic relationships to construct the variable dynamic network.

[0037] Specifically, multiple dynamic relationships are weighted based on the multiple state variables to obtain multiple weight coefficients, which reflect the strength and importance of the relationship between the state variables, ensuring that their mutual influence can be accurately reflected when constructing the dynamic network. For example, if a state variable has a significant impact on water quality, its corresponding edge weight will be relatively high. Then, the multiple weight coefficients are continuously updated based on the operating time series. Among them, the continuous update is due to the fact that the state variables of the system and their relationships will change over time, so the edge weights of the dynamic network need to be constantly adjusted. By introducing the latest monitoring data, the system can re-evaluate the weight coefficients to ensure that the dynamic network always reflects the current operating status. For example, if the water quality changes significantly over a period of time, the weight of the state variable related to water quality will increase to reflect its importance in the current situation.

[0038] Furthermore, in order to achieve dynamic adjustment, a sliding window is constructed, and the sliding window method is used to effectively connect multiple state variables and dynamically adjust the weight of the edge. The sliding window is a time series analysis tool that allows the system to analyze relevant data in each time window and update it in real time by moving in continuous time periods. In the specific implementation, a sliding window of appropriate size is first defined, and the length of the window can be set according to the system operation cycle and the frequency of data changes. In each sliding window, the state variables in the current time period are analyzed, including calculating correlation, causal relationship and fitting function, etc. These analyses provide the basis for subsequent dynamic connection. For example, when the window slides to a specific time period, the system will extract the state variable data in the time period and use the dynamic weight coefficient to evaluate the strength of the relationship between these variables. In this way, the connection between state variables will be adjusted according to their actual operating status and the degree of mutual influence. The process of connection is the key. The system connects the state variables according to multiple dynamic relationships to form directed edges, where the weight of each edge reflects the strength and directionality of the connection. As time goes by and data is updated, the correlation, causal relationship or fitting function in each window will be updated in real time according to the new data, ensuring that the constructed variable dynamic network always reflects the true state of the current system. This dynamic network can not only reflect the interactive relationship between the state variables, but also reveal the evolution process of the system in different time periods, thereby improving the intelligent control capability of the whole-house water purification system.

[0039] Through the sliding window, multiple state variables are successfully connected according to dynamic relationships, and a variable dynamic network that changes with time is constructed. This network will provide an important basis for subsequent intelligent control decisions. This method ensures the timeliness and accuracy of the variable dynamic network, enabling it to timely reflect the dynamic evolution process between different state variables of the system.

[0040] Furthermore, the variable dynamic network is used to perform data fitting to construct a dynamic model of the whole-house water purification system. Step S200 of the present application also includes:

[0041] Step S250: performing characteristic analysis based on the whole-house water purification system according to the operating rules to determine a physical characteristic data set of the whole-house water purification system.

[0042] Step S260: traverse the variable dynamic network and match it with the physical characteristic data set, identify the variable dynamic network according to the matching result, and generate multiple physical characteristic identification nodes.

[0043] Step S270: dynamically calculating the multiple state variables according to the multiple physical characteristic identification nodes to generate parameters to be fitted.

[0044] Step S280: performing data fitting on the parameters to be fitted based on the running time series, and constructing the dynamic model according to the fitting results.

[0045] Exemplarily, a characteristic analysis is performed based on the operation law of the whole-house water purification system to determine the physical characteristic data set of the system. This data set includes physical characteristics such as flow, pressure, temperature and water quality. These characteristics reflect the behavior mode of the system under normal operation and provide an important basis for subsequent data fitting. Then, the variable dynamic network is traversed and matched with the physical characteristic data set. The matching process is intended to identify the association between each state variable and the physical characteristic in the dynamic network to ensure that each state variable can reflect the corresponding physical characteristic. The variable dynamic network is identified according to the matching results, and multiple physical characteristic identification nodes are generated. These identification nodes not only clarify the specific physical meaning of each state variable, but also provide a structured framework for subsequent dynamic calculations. After the identification node is established, the system dynamically calculates multiple state variables according to multiple physical characteristic identification nodes to generate parameters to be fitted. These parameters to be fitted represent the response characteristics of the system under specific conditions and can play a key role in the construction of dynamic models. For example, if there is an obvious correlation between flow and water quality, the system will generate corresponding parameters to be fitted based on this relationship. Finally, the generated parameters to be fitted are fitted based on the running time series. This fitting process uses appropriate statistical or machine learning algorithms to optimize the model to better reflect the dynamic characteristics of the whole-house water purification system. According to the fitting results, the dynamic model is constructed, which can predict the output effect of the system under different input conditions in real time and provide a scientific basis for subsequent control decisions.

[0046] Furthermore, data fitting is performed on the parameters to be fitted based on the running time series, and the dynamic model is constructed according to the fitting results. Step S280 of the present application also includes:

[0047] Step S281: Building a model structure based on the physical property data set and determining the initial structure information of the model.

[0048] Step S282: traverse the whole house water purification system using the physical characteristic data for analysis to determine the physical environment information.

[0049] Step S283: setting a plurality of initial values ​​for the parameters to be fitted according to the physical environment information and the operating time sequence.

[0050] Step S284: Use the gradient descent method to iteratively fit the multiple initial values ​​until the multiple initial values ​​are traversed and stop, and output the fitting result.

[0051] Step S285: Perform data integration based on the initial structure information of the model and the fitting results to construct the dynamic model.

[0052] Further, a model structure is built based on the physical property data set to determine the initial structure information of the model. The model structure can be a linear regression model, an ARIMA model, a difference equation model or a nonlinear model. Next, the physical property data is used to traverse the whole house water purification system for analysis to determine the physical environment information. This analysis process can provide important background information about the system operation status, such as physical environment parameters such as temperature, pressure and flow rate. This information will provide a basis for the subsequent initial value setting. After mastering the physical environment information, multiple initial values ​​are set for the fitting parameters according to this information and the running time series. The set initial values ​​can provide a reasonable starting point for each parameter based on historical data, such as the correlation between flow and water quality. The rationality of these initial values ​​is crucial to the efficiency and accuracy of the fitting process. Subsequently, the gradient descent method is used to fit the multiple initial values ​​iteratively. The gradient descent method is a common optimization algorithm that adjusts the parameters by calculating the gradient of the objective function and gradually approaches the optimal solution. The iterative process will continue until all initial values ​​are traversed and it is judged that the output fitting result has reached a convergence state, that is, the error drops below the preset threshold. Finally, based on the initial structure information of the model and the obtained fitting results, data integration is performed to construct a dynamic model. By combining the model structure with the fitting results, a model that can accurately reflect the dynamic characteristics of the whole-house water purification system is finally formed, providing a reliable theoretical basis for subsequent intelligent control and optimization decisions. By constructing a dynamic model of the whole-house water purification system, the system's adaptability to changing environments and prediction accuracy are significantly improved.

[0053] Furthermore, the step S200 of obtaining the control decision in the present application further includes:

[0054] Step S290: Based on the dynamic model, the state of the whole-house water purification system is predicted according to the operation cycle to generate multiple prediction variables.

[0055] Step S2100: Retrieve a historical operation data record log, and set a plurality of state control targets based on the plurality of state variables in combination with the historical operation data record log.

[0056] Step S2110: performing predictive control of multiple time steps according to the multiple state control objectives in combination with the multiple prediction variables to obtain multiple prediction results.

[0057] Step S2120: Perform rolling optimization based on the multiple prediction results to make the control decision.

[0058] Next, control decisions for the whole-house water purification system are generated by combining the dynamic model with historical data. First, the system is predicted according to the set operating cycle based on the dynamic model. In this process, the dynamic model uses the latest state variables and model parameters to predict the future system state and generate multiple prediction variables, such as water quality parameters (TDS value, turbidity), flow rate and pressure, which are the basis for the subsequent control strategy formulation to ensure that each control decision is based on the latest system performance prediction. Subsequently, the historical operation data record logs are retrieved, which record the performance of the system under different conditions in detail. Multiple state control targets are set based on historical performance and multiple current prediction variables. For example, water quality targets such as TDS value less than 100ppm are set to ensure water quality safety, flow targets to ensure sufficient water flow rate to meet peak water demand, pressure control targets to ensure stable system water pressure, avoid equipment damage or insufficient water flow caused by water pressure problems, and filter management targets to maximize the service life of the filter while maintaining water quality.

[0059] Then, according to the set multiple state control targets, multiple predictive variables are combined to perform predictive control with multiple time steps. In the process of predictive control, not only the next upcoming state is predicted, but also adjustments are made based on the prediction results of each time step to achieve continuous system state optimization. This continuous prediction and adjustment process ensures that the system can continue to operate in the optimal state under changing environmental and operating conditions. Finally, based on the multiple prediction results obtained, rolling optimization is performed. Rolling optimization is a dynamic decision-making process in which the system updates the control strategy based on the latest prediction data after each prediction cycle. This approach keeps the control strategy up to date and effectively responds to any changes in external conditions and system performance.

[0060] Through the above steps, the whole-house water purification system can achieve intelligent operation management, continuously adjust the control strategy according to the prediction results and actual operation conditions, optimize system performance, ensure water safety and efficiency, and extend the service life of system components.

[0061] Furthermore, based on the multiple state variables and the multiple control action information, the dynamic model is subjected to reinforcement learning to construct a predictive controller. Step S300 of the present application further includes:

[0062] Step S310: constructing a reward function based on the multiple state variables and the multiple control action information.

[0063] Step S320: Continuously analyzing the multiple state variables in combination with the multiple control actions to generate a multi-dimensional state data set.

[0064] Step S330: The multi-dimensional state data set is used as input data and input into the dynamic model to interact with the reward function to obtain a continuous control data set.

[0065] Step S340: Randomly sampling the continuous control data sets to obtain first continuous control data, second continuous control data, ... Nth continuous control data, where N is the total number of continuous control data sets and N is an integer greater than 1.

[0066] Step S350: Based on the first continuous control data, the second continuous control data...the Nth continuous control data, the dynamic model is subjected to reinforcement learning to generate a first control action, a second control action...the Mth control action, where M is an integer greater than 1 and less than or equal to N.

[0067] Step S360: Conduct training based on the first control action, the second control action, ... the Mth control action to construct the predictive controller.

[0068] Specifically, a reward function is constructed based on the multiple state variables and control action information. The reward function is the core of reinforcement learning, which defines the feedback that the system should receive under different control actions. The reward function sets goals for the whole-house water purification system, such as optimizing water quality, minimizing energy consumption, and extending the service life of the filter element. Specifically, if the system operation causes the water quality to meet or exceed the preset standard (such as the TDS value is less than 100ppm), the reward function will give a positive reward; conversely, if the water quality does not meet the standard or the energy consumption is too high, a negative reward will be given. Next, multiple state variables and control actions are continuously analyzed to generate a multidimensional state data set, which includes various possible state combinations and corresponding system outputs, such as water quality parameters, energy consumption, filter element status, etc. These data are used to train the dynamic model to ensure that the model can fully understand and respond to changes in the system state. Next, the multidimensional state data set is used as input and input into the dynamic model combined with the reward function for interaction. In this step, the model continuously adjusts its internal parameters through interaction with the input data to maximize the cumulative reward. This process is the stage where the model learns how to adjust its behavior based on actual operation feedback. After that, random sampling is performed according to the obtained continuous control data set to obtain a series of continuous control data points, which represent the expected system response under specific operations and are used as training data to further optimize the model. This step ensures that the most representative examples are extracted from a large number of possible operations, improving learning efficiency. Finally, reinforcement learning is performed on the dynamic model based on these continuous control data, and the model is trained through these data to identify and implement the optimal control actions. After each round of training, the model generates a series of control actions, each of which is designed to drive the system to or maintain the optimal state. After training, the predictive controller can calculate the optimal control strategy based on the current and predicted system states, and adjust the operation of the water purification system in real time to achieve the optimal water treatment effect.

[0069] Through the above steps, the predictive controller can not only respond to changes in system status in real time, but also foresee future changes in status and adjust the control strategy in advance, thereby significantly improving the intelligence level and operational efficiency of the whole-house water purification system.

[0070] Through the technical solutions of the above embodiments, the whole-house water purification system control optimization method based on dynamic adjustment provided by the present application solves the technical problems that the existing whole-house water purification system has low control efficiency, slow response speed and poor stability in water quality changes, and is difficult to meet the efficient and accurate needs of modern families for safe water use. It achieves the technical effect of improving the control efficiency, response speed and stability of the water purification system and ensuring the safety of household water use.

[0071] Embodiment 2 is based on the same inventive concept as the whole-house water purification system control optimization method based on dynamic adjustment in the previous embodiment. Figure 2As shown, the present application provides a whole-house water purification system control optimization platform based on dynamic regulation, the platform comprising:

[0072] The system operation analysis module 11 is used to perform operation change analysis based on the whole-house water purification system and determine a plurality of state variables, wherein the plurality of state variables have a plurality of control action information;

[0073] A dynamic model building module 12 is used to establish a variable dynamic network according to the multiple state variables, use the variable dynamic network to perform data fitting, build a dynamic model of the whole-house water purification system, and obtain a control decision;

[0074] A prediction controller construction module 13, configured to perform reinforcement learning on the dynamic model based on the multiple state variables and the multiple control action information to construct a prediction controller;

[0075] A control optimization module 14, used to dynamically optimize the control decision through the predictive controller to obtain a control optimization decision;

[0076] The feedback adjustment module 15 is used to execute the control optimization decision to perform feedback dynamic adjustment, and to perform intelligent control optimization on the whole-house water purification system based on the dynamic adjustment parameter set.

[0077] Furthermore, the dynamic model building module 12 is also used to perform the following steps:

[0078] The multiple state variables are monitored according to the operation cycle of the whole-house water purification system to determine multiple operation time nodes, and the multiple operation time nodes correspond to the multiple state variables;

[0079] Arrange the multiple state variables according to the time sequence at the multiple running time nodes to obtain a running time sequence;

[0080] Performing correlation analysis on the multiple state variables based on the running time series to extract multiple dynamic relationships;

[0081] The variable dynamic network is constructed by taking the multiple dynamic relationships as edges and the multiple state variables as points.

[0082] Furthermore, the dynamic model building module 12 is also used to perform the following steps:

[0083] Based on the multiple state variables, weights are assigned to the multiple dynamic relationships to obtain multiple weight coefficients;

[0084] Continuously updating the plurality of weight coefficients based on the running time sequence to determine a plurality of dynamic weight coefficients;

[0085] A sliding window is constructed, and the plurality of state variables are connected according to the plurality of dynamic relationships by using the sliding window in combination with the plurality of dynamic weight coefficients to construct the variable dynamic network.

[0086] Furthermore, the dynamic model building module 12 is also used to perform the following steps:

[0087] Based on the characteristics analysis of the whole-house water purification system according to the operating rules, the physical characteristics data set of the whole-house water purification system is determined;

[0088] Traversing the variable dynamic network to match with the physical characteristic data set, identifying the variable dynamic network according to the matching result, and generating a plurality of physical characteristic identification nodes;

[0089] Dynamically calculating the multiple state variables according to the multiple physical characteristic identification nodes to generate parameters to be fitted;

[0090] Data fitting is performed on the parameters to be fitted based on the running time series, and the dynamic model is constructed according to the fitting results.

[0091] Furthermore, the dynamic model building module 12 is also used to perform the following steps:

[0092] Building a model structure based on the physical property data set and determining initial model structure information;

[0093] Utilizing the physical characteristic data to traverse the whole house water purification system for analysis and determine physical environment information;

[0094] Setting a plurality of initial values ​​for the parameters to be fitted according to the physical environment information and the operating time sequence;

[0095] Using a gradient descent method to iterate the fitting of the multiple initial values ​​until the multiple initial values ​​are traversed and then outputting the fitting result;

[0096] The dynamic model is constructed by performing data integration based on the initial structure information of the model and the fitting results.

[0097] Furthermore, the dynamic model building module 12 is also used to perform the following steps:

[0098] Based on the dynamic model, the state of the whole-house water purification system is predicted according to the operation cycle to generate multiple prediction variables;

[0099] Retrieving a historical operation data record log, and setting a plurality of state control targets based on the plurality of state variables in combination with the historical operation data record log;

[0100] Perform predictive control of multiple time steps according to the multiple state control targets and the multiple prediction variables to obtain multiple prediction results;

[0101] The control decision is made by performing rolling optimization based on the multiple prediction results.

[0102] Furthermore, the prediction controller building module 13 is also used to perform the following steps:

[0103] Constructing a reward function based on the plurality of state variables and the plurality of control action information;

[0104] Continuously analyzing the plurality of state variables in combination with the plurality of control actions to generate a multi-dimensional state data set;

[0105] The multi-dimensional state data set is used as input data, inputted into the dynamic model to interact with the reward function, and a continuous control data set is obtained;

[0106] Randomly sampling the continuous control data set to obtain first continuous control data, second continuous control data, ... Nth continuous control data, where N is the total number of continuous control data sets and N is an integer greater than 1;

[0107] Based on the first continuous control data, the second continuous control data ... the Nth continuous control data, the dynamic model is reinforced to generate a first control action, a second control action ... the Mth control action, where M is an integer greater than 1 and less than or equal to N;

[0108] The predictive controller is constructed by training based on the first control action, the second control action, ... the Mth control action.

[0109] Through the above-mentioned detailed description of the whole-house water purification system control optimization method based on dynamic adjustment in this specification, those skilled in the art can clearly know the whole-house water purification system control optimization platform based on dynamic adjustment in this embodiment. For the platform disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0110] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A whole-house water purification system control optimization method based on dynamic regulation, characterized in that: The method comprises: Based on the operation change analysis of the whole-house water purification system, a plurality of state variables are determined, wherein the plurality of state variables have a plurality of control action information; Establish a variable dynamic network according to the multiple state variables, use the variable dynamic network to perform data fitting, build a dynamic model of the whole-house water purification system, and obtain a control decision; Performing reinforcement learning on the dynamic model based on the multiple state variables and the multiple control action information to construct a predictive controller; Dynamically optimizing the control decision by using the predictive controller to obtain a control optimization decision; Execute the control optimization decision to perform feedback dynamic adjustment, and perform intelligent control optimization on the whole-house water purification system based on the dynamic adjustment parameter set; Wherein, establishing a variable dynamic network according to the multiple state variables includes: The multiple state variables are monitored according to the operation cycle of the whole-house water purification system to determine multiple operation time nodes, and the multiple operation time nodes correspond to the multiple state variables; Arrange the multiple state variables according to the time sequence at the multiple running time nodes to obtain a running time sequence; Performing correlation analysis on the multiple state variables based on the running time series to extract multiple dynamic relationships; Using the multiple dynamic relationships as edges and the multiple state variables as points to construct the variable dynamic network; The multiple dynamic relationships are used as edges, and the multiple state variables are used as points to construct a variable dynamic network, including: Based on the multiple state variables, weights are assigned to the multiple dynamic relationships to obtain multiple weight coefficients; Continuously updating the plurality of weight coefficients based on the running time sequence to determine a plurality of dynamic weight coefficients; A sliding window is constructed, and the plurality of state variables are connected according to the plurality of dynamic relationships by using the sliding window in combination with the plurality of dynamic weight coefficients to construct the variable dynamic network.

2. The whole-house water purification system control optimization method based on dynamic adjustment according to claim 1, characterized in that: The variable dynamic network is used to perform data fitting to construct a dynamic model of the whole-house water purification system, and the method includes: Based on the characteristics analysis of the whole-house water purification system according to the operating rules, the physical characteristics data set of the whole-house water purification system is determined; Traversing the variable dynamic network to match with the physical characteristic data set, identifying the variable dynamic network according to the matching result, and generating a plurality of physical characteristic identification nodes; Dynamically calculating the multiple state variables according to the multiple physical characteristic identification nodes to generate parameters to be fitted; Data fitting is performed on the parameters to be fitted based on the running time series, and the dynamic model is constructed according to the fitting results.

3. The whole-house water purification system control optimization method based on dynamic adjustment as claimed in claim 2, characterized in that: Performing data fitting on the parameters to be fitted based on the running time series, and constructing a dynamic model according to the fitting results, the method comprising: Building a model structure based on the physical property data set and determining initial model structure information; Utilizing the physical characteristic data to traverse the whole house water purification system for analysis and determine physical environment information; Setting a plurality of initial values ​​for the parameters to be fitted according to the physical environment information and the operating time sequence; Using a gradient descent method to iterate the fitting of the multiple initial values ​​until the multiple initial values ​​are traversed and then outputting the fitting result; The dynamic model is constructed by performing data integration based on the initial structure information of the model and the fitting results.

4. The whole-house water purification system control optimization method based on dynamic regulation according to claim 1, characterized in that: Obtaining the control decision, the method comprises: Based on the dynamic model, the state of the whole-house water purification system is predicted according to the operation cycle to generate multiple prediction variables; Retrieving a historical operation data record log, and setting a plurality of state control targets based on the plurality of state variables in combination with the historical operation data record log; Perform predictive control of multiple time steps according to the multiple state control targets and the multiple prediction variables to obtain multiple prediction results; The control decision is made by performing rolling optimization based on the multiple prediction results.

5. The whole-house water purification system control optimization method based on dynamic regulation according to claim 1, characterized in that: Reinforcement learning is performed on the dynamic model based on the multiple state variables and the multiple control action information to construct a predictive controller, the method comprising: Constructing a reward function based on the plurality of state variables and the plurality of control action information; Continuously analyzing the plurality of state variables in combination with the plurality of control actions to generate a multi-dimensional state data set; The multi-dimensional state data set is used as input data, inputted into the dynamic model to interact with the reward function, and a continuous control data set is obtained; Randomly sampling the continuous control data set to obtain first continuous control data, second continuous control data, ... Nth continuous control data, where N is the total number of continuous control data sets and N is an integer greater than 1; Based on the first continuous control data, the second continuous control data ... the Nth continuous control data, the dynamic model is reinforced to generate a first control action, a second control action ... the Mth control action, where M is an integer greater than 1 and less than or equal to N; The predictive controller is constructed by training based on the first control action, the second control action, ... the Mth control action.

6. The whole-house water purification system control optimization platform based on dynamic adjustment is characterized by: For implementing the whole-house water purification system control optimization method based on dynamic regulation according to any one of claims 1 to 5, the platform comprises: A system operation analysis module, used to perform operation change analysis based on the whole-house water purification system and determine a plurality of state variables, wherein the plurality of state variables have a plurality of control action information; A dynamic model building module, used to establish a variable dynamic network according to the multiple state variables, use the variable dynamic network to perform data fitting, build a dynamic model of the whole-house water purification system, and obtain control decisions; A prediction controller construction module, used for performing reinforcement learning on the dynamic model based on the multiple state variables and the multiple control action information to construct a prediction controller; A control optimization module, used to dynamically optimize the control decision through the predictive controller to obtain a control optimization decision; The feedback adjustment module is used to execute the control optimization decision to perform feedback dynamic adjustment, and to perform intelligent control optimization on the whole-house water purification system based on the dynamic adjustment parameter set.

Images