| CPC G06Q 10/04 (2013.01) [G16Y 10/35 (2020.01); G16Y 20/30 (2020.01); G16Y 40/35 (2020.01)] | 6 Claims |
|
1. A method for regulating a gas pressure at an intelligent gas gate station based on an Internet of Things, wherein the method is performed by at least one processor of an intelligent gas management platform, and the intelligent gas management platform comprises an intelligent customer service management sub-platform, an intelligent operation management sub-platform, and an intelligent gas data center, the method comprising:
obtaining, by the at least one processor of the intelligent gas data center, gas terminal information from an intelligent gas object platform through an intelligent gas sensor network platform, wherein the gas terminal information includes gas terminal flow and a gas terminal distribution feature;
predicting, by the at least one processor of the intelligent operation management sub-platform, gas gate station flow by analyzing the gas terminal flow and the gas terminal distribution feature based on a flow model, wherein the flow model is a machine learning model, and the flow model includes a first embedding layer, a second embedding layer, and a first output layer, wherein an output of the first embedding layer and an output of the second embedding layer are input to the first output layer, an input of the first embedding layer includes gas terminal flow at a plurality of time points, and an output of the first embedding layer includes a flow feature vector; an input of the second embedding layer includes the gas terminal distribution feature, and an output of the second embedding layer includes a distribution feature vector; and an input of the first output layer includes the flow feature vector and the distribution feature vector, and an output of the first output layer includes the gas gate station flow; wherein the first embedding layer, the second embedding layer, and the first output layer are obtained through joint training based on a plurality of first training samples and first labels, including:
generating the plurality of first training samples and the first labels, wherein the first training samples include sample gas terminal flow and sample gas terminal distribution features and the first labels include the sample gas gate station flow;
inputting the sample gas terminal flow into an initial first embedding layer to obtain the flow feature vector output by the initial first embedding layer;
inputting the sample gas terminal distribution features into an initial second embedding layer to obtain the distribution feature vector output by the initial second embedding layer;
inputting the flow feature vector and the distribution feature vector into an initial first output layer to obtain the gas gate station flow output by the initial first output layer;
updating the initial first embedding layer, the initial second embedding layer, and the initial first output layer simultaneously by constructing a loss function based on the sample gas gate station flow and the gas gate station flow output by the initial first output layer; and
obtaining a trained first embedding layer, a trained second embedding layer, and a trained first output layer by updating parameters of the initial first embedding layer, the initial second embedding layer, and the initial first output layer;
determining, by the at least one processor of the intelligent operation management sub-platform, gas gate station pressure of each gas gate station based the gas gate station flow, gas gate station feature and gas terminal target pressure through a pressure model, obtaining a pressure sum by aggregating gas gate station pressures of gas gate stations, and determining a pressure regulation scheme of the each gas gate station based on the pressure sum and a total pressure threshold, wherein the pressure regulation scheme is a scheme for regulating the gas pressure; wherein the pressure model is a machine learning model, and the pressure model includes an input layer, a third embedding layer and a second output layer; an input of the input layer includes the gas gate station flow and the gas gate station feature, and an output of the input layer includes gate station flow data; an input of the third embedding layer includes the gas terminal target pressure, and an output of the third embedding layer includes a gas terminal pressure distribution; and an input of the second output layer includes the gate station flow data and the gas terminal pressure distribution, and an output of the second output layer includes the gas gate station pressure, wherein the input layer, the third embedding layer, and the second output layer are obtained through joint training based on a plurality of second training samples and second labels, including:
generating the plurality of second training samples and the second labels, wherein the second training samples include sample gas gate station flow, sample gas gate station features, sample gas terminal target pressure, and sample distribution feature vector, and the second labels include a sample gas gate station pressure;
inputting the sample gas gate station flow and the sample gas gate station features into an initial input layer to obtain the gate station flow data output by the initial input layer;
inputting the sample gas terminal target pressure and the sample distribution feature vector into an initial third embedding layer to obtain the gas terminal pressure distribution output by the initial third embedding layer;
inputting the gate station flow data and the gas terminal pressure distribution into an initial second output layer to obtain the gas terminal pressure output by the initial second output layer;
updating the initial input layer, the initial third embedding layer, and the initial second output layer simultaneously by constructing a loss function based on the sample gas gate station pressure and the gas gate station pressure output by the initial second output layer; and
obtaining a trained input layer, a trained third embedding layer, and a trained second output layer by updating parameters of the initial input layer, the initial third embedding layer, and the initial second output layer;
wherein the flow model and the pressure model are obtained through joint training based on a plurality of third training samples and third labels, including:
generating the plurality of third training samples and the third labels, wherein the third training samples include the sample gas terminal flow, the sample gas terminal distribution features, the sample gas gate station features, and the sample gas terminal target pressure, and the third labels include the sample gas gate station pressure;
inputting the sample gas terminal flow and the sample gas terminal distribution feature into an initial flow model to obtain the gate station flow data output by the initial flow model;
inputting the gate station flow data into an initial pressure model together with the sample gas gate station features and the sample gas terminal target pressure to obtain the gas gate station pressure output by the initial pressure model;
updating the initial flow model and the initial pressure model simultaneously by constructing a loss function based on the sample gas gate station pressure and the gas gate station pressure output by the initial pressure model; and
obtaining a trained flow model and a trained pressure model by updating parameters of the initial flow model and the initial pressure model; and
regulating, by the at least one processor of the intelligent operation management sub-platform, the gas gate station pressure using the trained flow model and the trained pressure model.
|