Heat supply pipe network hydraulic balance adjusting system and method integrating AI algorithm

Abstract

The invention discloses a heat supply pipe network hydraulic balance adjusting system and method integrated with an AI algorithm. The system comprises a multi-source data acquisition module, a data fusion module, an AI algorithm model module and an adjusting control module. The multi-source data acquisition module is a sensing basis and a decision-making basis source of the whole system and is used for acquiring multi-source data in real time; the data fusion module comprises a preprocessing module and an extraction fusion module. The invention belongs to the technical field of hydraulic balance adjustment of a heat supply pipe network, and aims to solve the problem that the prediction accuracy is reduced due to the fact that the valve opening is adjusted by depending on single fixed threshold data or artificial experience and dynamic factors of real-time weather are not fused in the prior art. The technical effects are that the weather data and the real-time data of the heat supply pipe network can be fused, the relevance and complementarity between different types of data are fully utilized, and the accuracy and efficiency of the hydraulic balance adjustment of the heat supply pipe network can be improved.

Description

Technical Field

[0001] This application belongs to the technical field of hydraulic balance regulation of heat supply networks, and particularly relates to a heat supply network hydraulic balance regulation system and method integrated with AI algorithms. Background Technique

[0002] With the acceleration of the urbanization process and the continuous improvement of people's living standards, the demand for heat supply is increasing day by day, and the heat supply network system has become more and more huge and complex. In this huge system, there are many heat sources, heat supply pipelines and various heat users. Due to differences in various factors such as the location, building structure, and usage requirements of each heat user, the requirements for heat supply are also different. If effective hydraulic balance regulation is lacking, the problem of uneven heat supply is likely to occur. For example, users closer to the heat source may experience overheating due to excessive water pressure, which not only causes energy waste but also may bring inconvenience to the living environment of users; while users farther from the heat source may have insufficient water pressure, resulting in the heating temperature not reaching the standard and being unable to meet their normal heating or domestic heat use requirements.

[0003] In addition, the heat supply network system is also affected by various external factors during operation, such as seasonal changes and climate changes. Under different seasons and climate conditions, the heat load of the system will change, which requires the heat supply network system to be able to perform flexible hydraulic balance regulation according to the actual situation. For example, in the cold winter, the demand for heat supply from heat users increases significantly. At this time, precise hydraulic balance regulation is required to ensure that the system can provide sufficient heat; while in the relatively warm transition season, the heating parameters need to be adjusted appropriately to avoid overheating. Therefore, in the heat supply network system, hydraulic balance regulation is the key to ensuring uniform and stable heat supply for each heat user.

[0004] However, most current heat supply network hydraulic balance regulation methods are mostly based on empirical methods, relying on single fixed threshold data or manual experience to adjust the valve opening, and cannot effectively, accurately and intelligently adjust in response to load changes in real time. At the same time, some AI solutions applied in the system only rely on historical data and do not integrate dynamic factors of real-time weather, reducing the prediction accuracy, which is likely to affect the effect of hydraulic balance regulation. Summary of the Invention

[0005] This application provides a heat supply network hydraulic balance regulation system and method integrated with AI algorithms, aiming to solve the problems in the prior art that rely on single fixed threshold data or manual experience to adjust the valve opening and do not integrate dynamic factors of real-time weather, reducing the prediction accuracy.

[0006] In the first aspect, a heat supply network hydraulic balance regulation system integrated with AI algorithms includes a multi-source data acquisition module, a data fusion module, an AI algorithm model module, and a regulation and control module;

[0007] The multi-source data acquisition module is the perception basis and decision-making basis source of the entire system, and is used to obtain multi-source data in real time;

[0008] The data fusion module includes a preprocessing module and an extraction and fusion module. The preprocessing module is used to clean the real-time data and remove noise data; the extraction and fusion module is used to fuse the extracted features with the heat supply network topology structure data to form a feature set;

[0009] The AI algorithm model module is the core part of the heat supply network hydraulic balance adjustment system, and is used to convert the multi-source data of the heat supply network into a mathematical model for prediction and adjustment;

[0010] The adjustment and control module is used to receive the adjustment strategy generated by the AI algorithm model module and control the valves in the heat supply network to perform corresponding operations.

[0011] Further, the multi-source data acquisition module includes a real-time data acquisition module, a dynamic data acquisition module, and a solid-state data acquisition module.

[0012] Further, the real-time data acquisition module is used to collect the operation parameters of temperature, pressure, and flow rate from the heat supply network system;

[0013] The dynamic data acquisition module is used to obtain the weather data for the next 12 hours and collect the historical heat consumption habits of users;

[0014] The solid-state data acquisition module is used to obtain the laying structure information of the heat supply network.

[0015] Further, the adjustment and control module further includes a monitoring and feedback unit, and the monitoring and feedback unit is used to continuously monitor the operation status of the heat supply network.

[0016] Further, the preprocessing module specifically includes the preprocessing of real-time data, the preprocessing of dynamic data, and the preprocessing of solid-state data.

[0017] In a second aspect, a heat supply network hydraulic balance adjustment method integrated with an AI algorithm, the method includes the following steps:

[0018] S1: Real-time collect and obtain the multi-source data of the heat supply network;

[0019] S2: Classify and preprocess the multi-source data and perform feature fusion to form a feature set;

[0020] S3: Build an AI algorithm model based on the GNN model and the LSTM model, and train and optimize the AI algorithm model in combination with the feature set;

[0021] S4: Receive the adjustment strategy generated by the AI algorithm model module and automatically adjust the opening degree of the valves in the heating pipe network according to the adjustment strategy;

[0022] S5: Continuously monitor the operating status of the heating pipe network based on the monitoring and feedback unit to ensure the stable operation of the system.

[0023] Furthermore, the multi-source data includes real-time data, dynamic data, and solid-state data.

[0024] Furthermore, the specific steps of S2 are as follows:

[0025] S2.1: Real-time data preprocessing: Clean the real-time data, remove the noise data, then use statistical methods to identify and process the outliers, and then convert the processed real-time data into a unified dimension and perform standardization processing;

[0026] S2.2: Dynamic data preprocessing: Clean the weather data for the next 12 hours, remove the outliers, divide it according to the time step, and organize the user's historical heat consumption habit data to form a structured data set;

[0027] S2.3: Solid-state data preprocessing: Convert the laying structure information of the heating pipe network into a graph data structure and uniquely identify the nodes and edges in the heating pipe network;

[0028] S2.4: Feature extraction and fusion: Use the sliding window method to perform time series fusion on the real-time data and the weather data for the next 12 hours.

[0029] Furthermore, the specific steps of S3 are as follows:

[0030] S3.1: Construct a GNN model: Use the solid-state data of the heating pipe network as the base graph structure;

[0031] S3.2: On the basis of the GNN model, introduce an LSTM model to capture the time dependence in the time series data;

[0032] S3.3: Use the historical data and the feature set to train the GNN model, update the model parameters through the backpropagation algorithm, and minimize the prediction error;

[0033] S3.4: Introduce a deep learning model and combine the weather data to optimize the GNN model parameters to obtain the AI algorithm model.

[0034] Compared with the prior art, the present application has at least the following beneficial effects:

[0035] This application is based on further analysis and research of existing technical problems. Through the multi-source data acquisition module, weather data can be fused with the real-time data of the heating pipe network, making full use of the relevance and complementarity between different types of data, improving the prediction accuracy and dynamic optimization effect, and providing comprehensive and high-precision data support for the accurate prediction and dynamic optimization of subsequent AI algorithms. At the same time, combined with the AI algorithm model module, the multi-source data of the heating pipe network can be transformed into a mathematical model for prediction and adjustment, so that the whole can more comprehensively understand the operation law of the heating pipe network and improve the accuracy and efficiency of the hydraulic balance adjustment of the heating pipe network. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 FIG. is a schematic diagram of the modules of a heating pipe network hydraulic balance adjustment system and method integrating an AI algorithm provided by an embodiment of the present application;

[0037] Figure 2 FIG. is a schematic flowchart of a heating pipe network hydraulic balance adjustment system and method integrating an AI algorithm provided by an embodiment of the present application; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments.

[0039] As Figure 1-2 shown, the heating pipe network hydraulic balance adjustment system integrating an AI algorithm provided by the present application includes a multi-source data acquisition module, a data fusion module, an AI algorithm model module, and a regulation and control module.

[0040] The multi-source data acquisition module is the perception basis and decision-making basis source of the entire system. Its core role is to provide comprehensive and high-precision data support for the accurate prediction and dynamic optimization of subsequent AI algorithms through the real-time acquisition and fusion of multi-source data.

[0041] The multi-source data acquisition module includes a real-time data acquisition module, a dynamic data acquisition module, and a solid-state data acquisition module; among them, the real-time data acquisition module is used to collect the operating parameters of temperature, pressure, and flow from the heating pipe network system;

[0042] The dynamic data acquisition module is used to obtain weather data for the next 12 hours and collect the historical heat consumption habits of users. The weather data includes temperature and humidity.

[0043] The solid-state data acquisition module is used to obtain the laying structure information of the heating pipe network, such as the connection relationship between the heat source and the user pipeline.

[0044] The data fusion module includes a preprocessing module and an extraction and fusion module. The preprocessing module ensures the accuracy and reliability of the data. The extraction and fusion module makes full use of the correlation and complementarity between different types of data, improving the prediction accuracy and dynamic optimization effect. The specific content is as follows:

[0045] a) Preprocessing of real-time data

[0046] First, clean the real-time data to remove noise data. The noise data includes sensor failure data. Sensor failures may cause obvious errors or unreasonable values in the data. For example, if a temperature sensor is damaged, it may display extremely high or low temperature values. In this case, the data can be screened by setting a reasonable temperature range. For example, in ambient temperature monitoring, if the temperature value exceeds the historical temperature extreme range of the area, the data may be due to sensor failure and should be excluded.

[0047] After cleaning the real-time data, use the statistical method of box plots to identify and process outliers. A box plot is a graph used to display the data distribution, which can visually show information such as the interquartile range (IQR), median, maximum, and minimum of the data. Taking temperature data as an example, plot the temperature data over a period of time as a box plot. The upper edge of the box represents the upper quartile (Q3) of the data, the lower edge represents the lower quartile (Q1) of the data, and the horizontal line in the box represents the median (Q2). In a box plot, data points that exceed 1.5 times the IQR outside the upper and lower quartile ranges are usually defined as outliers. That is, if a data point is less than Q1 - 1.5×IQR or greater than Q3 + 1.5×IQR, it is considered an outlier. For example, for a set of temperature data, Q1 is 10°C, Q3 is 20°C, and IQR is 10°C, then temperature data less than 10 - 1.5×10 = -5°C or greater than 20 + 1.5×10 = 35°C will be identified as outliers. After identifying the outliers, delete them.

[0048] Among them, the box plot generally refers to a statistical method, not limited to the box plot. For example, it can be the Z-score. The Z-score is a method for measuring the standard deviation between a data point and the mean. Similar to the box plot method, for the identified outliers, they can be deleted according to the actual situation.

[0049] Convert the real-time data after processing outliers into a unified dimension. Standardization processing helps to accelerate the convergence speed of AI algorithms and improve the prediction accuracy.

[0050] b) Preprocessing of dynamic data

[0051] According to the above preprocessing method, clean the temperature and humidity data for the next 12 hours and remove outliers. Divide the weather data by time step for subsequent time series analysis. At the same time, organize the user's historical heat consumption habit data to form a structured data set.

[0052] c) Preprocessing of solid-state data

[0053] Convert the laying structure information of the heat supply network into a graph data structure, uniquely identify the nodes and edges in the heat supply network, and extract relevant attribute information (such as location, diameter, material, etc.) for subsequent processing by the graph neural network (GNN). Among them, the nodes represent heat source entities, and the edges represent the connection relationship between the heat source and the user pipeline.

[0054] d) Feature extraction and fusion

[0055] Use the sliding window method to perform time series fusion on the real-time data (temperature, pressure, flow) and the weather data (temperature, humidity) for the next 12 hours. Align the real-time data with the weather data for subsequent time series analysis.

[0056] Fuse the heat supply network structure data with the real-time data, weather data, and user's historical heat consumption habit data. Use the nodes and edges in the heat supply network as the carriers of spatial information, and use the real-time operating parameters, weather data, user's historical heat consumption habit data, etc. as the attribute information of the nodes.

[0057] Use the principal component analysis (PCA) method to extract key features from the real-time data, weather data, and user's historical heat consumption habit data. Fuse the extracted features with the heat supply network topology structure data to form a feature set.

[0058] The AI algorithm model module is the core part of the heat supply network hydraulic balance adjustment system. It is responsible for converting the multi-source data of the heat supply network into a mathematical model for prediction and adjustment. By introducing AI algorithms, especially graph neural network (GNN) and long short-term memory network (LSTM) technologies, this module can more comprehensively understand the operation rules of the heat supply network and improve the accuracy and efficiency of heat supply network hydraulic balance adjustment. The specific content is as follows:

[0059] a) Use the solid-state data of the heat supply network as the matrix graph structure, where the nodes represent heat source entities and the edges represent the connection relationship between the heat source and the user pipeline.

[0060] b) Select a GNN model and build it based on the solid-state data of the heating network. The GNN model integrates information from neighboring nodes, enabling a more comprehensive understanding of the interactions between heat exchange stations, thereby improving prediction accuracy. The GNN's message passing mechanism transfers information between nodes and edges, updating node feature representations and facilitating the capture of the spatial impact of the heating network structure on the heat load.

[0061] c) Building on the GNN model, the LSTM model is introduced to capture temporal dependencies in time series data. The node features output by the GNN are used as the input sequence for the LSTM. The LSTM model leverages the spatial features extracted by the GNN and combines them with historical heating data and ambient temperature data to further learn how the heat load changes over time, ultimately forming a GNN model.

[0062] By setting the LSTM time step, you can ensure that the model can capture periodic characteristics such as daily and weekly changes in heat load. Setting the LSTM time step ensures that the model can capture periodic characteristics such as daily and weekly changes in heat load.

[0063] The LSTM model is a neural network that is good at processing time-sequential data and can learn the changing patterns of time series.

[0064] d) Use historical data and feature sets to train the GNN model. Update model parameters through the backpropagation algorithm to minimize prediction error. Use cross-validation and other methods to evaluate model performance and select the optimal model parameter combination to fine-tune and optimize the GNN model.

[0065] e) Introducing deep learning models

[0066] Considering the impact of varying weather data on heat load, a deep learning model is introduced to adapt to new weather data. After a GNN model is trained on a set of weather data, it is used as the source model. A small amount of data is collected from the new weather data set as the target domain data. By fine-tuning the parameters of the source model to adapt it to the target domain data, an AI algorithm model is generated. Therefore, by introducing a deep learning model and fine-tuning it to adapt to new weather data, the accuracy and efficiency of the hydraulic balance control system in the heating network can be effectively improved.

[0067] The regulation control module is responsible for receiving the regulation strategy generated by the AI algorithm model module and controlling the valves in the heating network to perform corresponding operations.

[0068] The adjustment control module automatically adjusts the opening degree of the valves in the heat supply network according to the adjustment strategy to achieve the hydraulic balance and thermal balance of the heat supply network. This is achieved through an electric actuator. The electric actuator receives signals from the control system, converts them into mechanical energy, and thus drives the opening or closing of the valves.

[0069] Meanwhile, the adjustment control module also includes a monitoring and feedback unit. The monitoring and feedback unit can continuously monitor the operating status of the heat supply network to ensure the stable operation of the system. And it will real-time feedback the new monitoring data to the AI algorithm model module for further optimization and adjustment. Through this real-time monitoring and feedback loop, the operating efficiency and stability of the heat supply network are continuously improved.

[0070] A method for adjusting the hydraulic balance of a heat supply network integrated with an AI algorithm, the method comprising the following steps:

[0071] S1: Real-time collect and obtain multi-source data of the heat supply network.

[0072] Among them, the multi-source data includes real-time data, dynamic data, and solid data. The specific data includes the operating parameters of temperature, pressure, and flow rate in the heat supply network system, weather data for the next 12 hours, collected historical heat consumption habits of users, and the laying structure information of the heat supply network. The collection of multi-source data helps to comprehensively understand the operating status of the heat supply network and provides rich information for subsequent data processing and model training.

[0073] S2: Classify, preprocess, and feature fuse the multi-source data to form a feature set. Data preprocessing is to improve the quality of the data. At the same time, through feature fusion technology, data from different sources are integrated into a feature set with higher value, providing more useful inputs for subsequent model training.

[0074] The specific steps of S2 are as follows:

[0075] S2.1: Real-time data preprocessing: Clean the real-time data, remove noise data (such as sensor failure data), then use statistical methods to identify and process outliers, and then convert the processed real-time data into a unified dimension and perform standardization processing;

[0076] S2.2: Dynamic data preprocessing: Clean the weather data for the next 12 hours, remove outliers, and divide it according to time steps, and organize the historical heat consumption habit data of users to form a structured data set;

[0077] S2.3: Solid data preprocessing: Convert the laying structure information of the heat supply network into a graph data structure and uniquely identify the nodes and edges in the heat supply network;

[0078] S2.4: Feature Extraction and Fusion: Use the sliding window method to perform time series fusion on real-time data and weather data for the next 12 hours. Perform spatial information fusion on the heat supply network structure data with real-time data, weather data, and user historical heat consumption habit data. Use the principal component analysis (PCA) method to extract key features from various types of data and fuse them with the heat supply network topology structure data to form a feature set.

[0079] S3: Construct an AI algorithm model based on the GNN model and the LSTM model, and train and optimize the AI algorithm model in combination with the feature set. Use the graph neural network (GNN) and the long short-term memory network (LSTM) to construct a complex AI algorithm model. GNN can process data with a graph structure and is suitable for representing complex networks such as heat supply networks; LSTM is good at processing time series data and can capture the dynamic changes during the operation of the heat supply network. By training and optimizing the model in combination with the feature set, a model that can accurately predict and adjust the operating state of the heat supply network can be obtained.

[0080] The specific steps of S3 are as follows:

[0081] S3.1: Construct the GNN model: Use the solid-state data of the heat supply network as the basic graph structure, where the nodes represent heat source entities and the edges represent the connection relationships between heat sources and user pipelines.

[0082] Select the GNN model and construct the GNN model according to the solid-state data of the heat supply network to integrate the information of neighboring nodes.

[0083] S3.2: On the basis of the GNN model, introduce the LSTM model to capture the time dependence in time series data. Use the node features output by the GNN as the input sequence of the LSTM to learn the variation law of heat load over time. Set the time step of the LSTM to ensure that periodic features such as daily and weekly variations of the heat load can be captured.

[0084] S3.3: Use historical data and the feature set to train the GNN model, update the model parameters through the backpropagation algorithm, and minimize the prediction error. Use methods such as cross-validation to evaluate the model performance, select the optimal combination of model parameters, and optimize the GNN model.

[0085] S3.4: Introduce a deep learning model and combine it with weather data to optimize the parameters of the GNN model to obtain the AI algorithm model.

[0086] S4: Receive the adjustment strategy generated by the AI algorithm model module. According to the adjustment strategy, automatically adjust the opening degree of the valves in the heating pipe network to achieve the hydraulic balance and thermal balance of the heating pipe network. Apply the adjustment strategy generated by the AI algorithm model to the actual heating pipe network. By automatically adjusting the opening degree of the valves, the hydraulic balance and thermal balance are achieved. This helps to improve the heating efficiency, reduce energy waste, and enhance the comfort of users.

[0087] S5: Continuously monitor the operating status of the heating pipe network based on the monitoring and feedback unit to ensure the stable operation of the system. Continuously monitor the operating status of the heating pipe network through the monitoring and feedback unit, and promptly discover and handle possible problems. At the same time, the monitoring data can also be used for the continuous learning and optimization of the model, improving the adaptive ability and stability of the system.

[0088] Real-time feedback the new monitoring data to the AI algorithm model module for further optimization and adjustment.

[0089] In the above heating pipe network hydraulic balance adjustment system and method integrated with the AI algorithm, through the multi-source data acquisition module, the weather data can be fused with the real-time data of the heating pipe network, making full use of the correlation and complementarity between different types of data, improving the prediction accuracy and dynamic optimization effect, and providing comprehensive and high-precision data support for the accurate prediction and dynamic optimization of the subsequent AI algorithm. At the same time, combined with the AI algorithm model module, the multi-source data of the heating pipe network can be transformed into a mathematical model for prediction and adjustment, so that the whole can more comprehensively understand the operating rules of the heating pipe network, and improve the accuracy and efficiency of the heating pipe network hydraulic balance adjustment.

[0090] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

Claims

1. A hydraulic balance adjustment system for a heat supply network integrated with an AI algorithm, characterized in that It includes a multi-source data acquisition module, a data fusion module, an AI algorithm model module, and an adjustment and control module; The multi-source data acquisition module is the perception basis and decision-making basis source of the entire system, and is used to obtain multi-source data in real time; The data fusion module includes a preprocessing module and an extraction and fusion module. The preprocessing module is used to clean the real-time data and remove noise data; the extraction and fusion module is used to fuse the extracted features with the heat supply network topology structure data to form a feature set; The AI algorithm model module is the core part of the heat supply network hydraulic balance adjustment system, and is used to convert the multi-source data of the heat supply network into a mathematical model for prediction and adjustment; The adjustment and control module is used to receive the adjustment strategy generated by the AI algorithm model module, and control the valves in the heat supply network to perform corresponding operations.

2. The hydraulic balance adjustment system for a heat supply network integrated with an AI algorithm according to claim 1, wherein The multi-source data acquisition module includes a real-time data acquisition module, a dynamic data acquisition module, and a solid-state data acquisition module.

3. The hydraulic balance adjustment system for a heating pipe network integrated with an AI algorithm according to claim 2, characterized in that, The real-time data acquisition module is used to collect the operating parameters of temperature, pressure, and flow rate from the heat supply network system; The dynamic data acquisition module is used to obtain the weather data for the next 12 hours and collect the historical heat consumption habits of users; The solid-state data acquisition module is used to obtain the laying structure information of the heat supply network.

4. The hydraulic balance adjustment system for a heat supply network integrated with an AI algorithm according to claim 1, characterized in that, The adjustment and control module further includes a monitoring and feedback unit, and the monitoring and feedback unit is used to continuously monitor the operating state of the heat supply network.

5. The hydraulic balance adjustment system of the heat supply network integrated with the AI algorithm according to claim 1, characterized in that The preprocessing module specifically includes the preprocessing of real-time data, the preprocessing of dynamic data, and the preprocessing of solid-state data.

6. The hydraulic balance adjustment method for a heat supply pipe network integrated with an AI algorithm, characterized in that, The method includes the following steps: S1: Collect and obtain the multi-source data of the heat supply network in real time; S2: Perform classification preprocessing and feature fusion on the multi-source data to form a feature set; S3: Build an AI algorithm model based on the GNN model and the LSTM model, and train and optimize the AI algorithm model in combination with the feature set; S4: Receive the adjustment strategy generated by the AI algorithm model module, and automatically adjust the opening degree of the valves in the heat supply network according to the adjustment strategy; S5: Continuously monitor the operating state of the heat supply network based on the monitoring and feedback unit to ensure the stable operation of the system.

7. The hydraulic balance adjustment method for a heating pipe network integrated with an AI algorithm according to claim 6, characterized in that, The multi-source data includes real-time data, dynamic data, and solid-state data.

8. The hydraulic balance adjustment method for a heat supply network integrating an AI algorithm according to claim 6, characterized in that, The specific steps of S2 are as follows: S2.1: Real-time data preprocessing: Clean the real-time data, remove noise data, then use statistical methods to identify and process outliers, and then convert the processed real-time data into a unified dimension and perform standardization processing; S2.2: Dynamic data preprocessing: Clean the weather data for the next 12 hours, remove outliers, and divide it according to time steps, and organize the historical heat consumption habit data of users to form a structured data set; S2.3: Solid-state data preprocessing: Convert the laying structure information of the heat supply network into a graph data structure, and uniquely identify the nodes and edges in the heat supply network; S2.4: Feature extraction and fusion: Use the sliding window method to perform time series fusion on the real-time data and the weather data for the next 12 hours.

9. The hydraulic balance adjustment method for a heating pipe network integrated with an AI algorithm according to claim 6, characterized in that, The specific steps of S3 are as follows: S3.1: Build a GNN model: Use the solid-state data of the heat supply network as the base graph structure; S3.2: On the basis of the GNN model, introduce the LSTM model to capture the temporal dependencies in time series data; S3.3: Use historical data and feature sets to train the GNN model, update the model parameters through the backpropagation algorithm, and minimize the prediction error; S3.4: Introduce a deep learning model and combine weather data to optimize the GNN model parameters to obtain an AI algorithm model.

Images