Method, internet of things system, and storage medium for determining gas compensation scheme based on smart gas

A method for determining a gas compensation scheme based on smart gas, wherein the method is implemented by a smart gas management platform based on an Internet of Things system for determining a gas compensation scheme based on smart gas, comprising:

• determining, based on a planned gas supply quantity of a gas supplier, a predicted gas supply quantity for a future preset time period using a preset manner;
• predicting a gas demand quantity for the future preset time period based on a gas usage quantity of a historical user;
• predicting a gas supply deviation rate of the gas supplier by a deviation rate prediction model based on a gas pipeline design map, weather information for the future preset time period, and the gas demand quantity for the future preset time period, the deviation rate prediction model being a machine learning model;
• determining a gas supply quantity for the future preset time period based on the predicted gas supply quantity and the gas supply deviation rate;
• determining a gas gap for the future preset time period based on the gas supply quantity and the gas demand quantity;
• and
• determining the gas compensation scheme based on the gas gap, operational requirements of an end of a pipeline, and operational parameters of the end of the pipeline, wherein:
• the deviation rate prediction model includes a pipeline feature extraction layer and a deviation rate prediction layer, the pipeline feature extraction layer is configured to process the gas pipeline design map to determine a pipeline feature map, and the deviation rate prediction layer is configured to process the pipeline feature map, the weather information for the future preset time period, and the gas demand quantity for the future preset time period to determine the gas supply deviation rate of the gas supplier;
• the deviation rate prediction model is obtained by jointly training of the pipeline feature extraction layer and the deviation rate prediction layer based on samples and labels, the samples include historical sample gas pipeline design maps for at least one historical sample area, weather information for a historical sample time period, and a gas demand quantity for the historical sample time period, and labels include a historical sample gas deviation rate corresponding to the historical sample area of the historical sample time period;
• wherein the jointly training includes:
• inputting the historical sample gas pipeline design maps into an initial pipeline feature extraction layer and obtaining an output of the initial pipeline feature extraction layer, wherein the output of the initial pipeline feature extraction layer includes a pipeline feature map that reflects features of gas pipelines, the pipeline feature map includes nodes and edges, the nodes represent gas pipelines, node attributes of the nodes reflect relevant features of corresponding gas pipelines, the node attributes include reliability of the pipelines and a standard flow rate interval of the pipelines, the edges represent that the gas pipelines are adjacent and connected, edge attributes of the edges reflect relevant features of corresponding pathways, and the edge attributes include a degree of bending at a node connection;
• inputting the weather information for the historical time period, the gas demand quantity for the historical time period, and the output of the initial pipeline feature extraction layer into an initial deviation rate prediction layer and obtaining an output of the initial deviation rate prediction layer, wherein the output of the initial deviation rate prediction layer includes a gas supply deviation rate of the gas supplier, and the gas supply deviation rate of the gas supplier is a degree of deviation between a quantity of gas actually supplied by the gas supplier to an entire preset area and the predicted gas supply quantity for the future preset time period;
• constructing a loss function based on the output of the initial deviation rate prediction layer and the labels;
• updating parameters of the initial pipeline feature extraction layer and the initial deviation rate prediction layer iteratively based on the loss function until meeting a preset condition;
• and
• obtaining the deviation rate prediction model.