Optimization control method for integrated energy system based on physical-informed neural network

Patent metadata

Field	Value
Publication number	US-2025199488-A1
Application number	US-202418948572-A
Country	US
Kind code	A1
Filing date	Nov 15, 2024
Priority date	Dec 18, 2023
Publication date	Jun 19, 2025
Grant date	—

Abstract

The present disclosure discloses an optimization control method for an integrated energy system based on a physical-informed neural network, which comprises the following steps: S 1 , constructing an a solar-electricity-heat-gas integrated energy system optimization control model; S 2 , generating a node connection relation matrix based on the network topology structure of the integrated energy system; S 3 , constructing a deep graph neural network model with physical-informed fusion; S 4 , constructing a loss function of the deep graph neural network model with physical-informed fusion; and S 5 , training a physical-informed neural network model according to the historical operation data to be used for system optimization control. The present disclosure can effectively deal with the influence of uncertainty of renewable energy and unexpected situations on the energy system, thereby ensuring the safe and stable operation of the integrated energy system.

First claim

What is claimed is:

1 . An optimization control method for an integrated energy system based on a physical-informed neural network, comprising following steps:
step S 1 , constructing a solar-electricity-heat-gas integrated energy system optimization control model; step S 2 , generating a node connection relation matrix based on a network topology structure of the integrated energy system; step S 3 , constructing a deep graph neural network model with physical-informed fusion based on the solar-electricity-heat-gas integrated energy system optimization control model constructed in the step S 1 and the node connection relation matrix constructed in the step S 2 ; wherein the step S 3 comprises the following sub-steps: sub-step S 31 , adopting a graph neural network structure with a graph convolutional network as a core for processing devices in the integrated energy system and an interrelationship therebetween; and regarding each device in the integrated energy system as a node in a graph, and regarding a physical connection between the devices as an edge:

n
i

∈
N

,

e

i
,
j

∈
E

where n i represents a node represented by a device in the integrated energy system; N represents a set constituted by all nodes; e i,j represents an edge represented by a branch between adjacent nodes i,j in the integrated energy system; and E represents a set constituted by all edges;
sub-step S 32 , assigning a feature vector to each node for representing a state variable and control strategy of the node; and determining, by the node, the feature vector V based on a state variable and a control variable of a corresponding device:

V
=

[

P
in

P
out

Q
in

Q
out

G
in

G
out

SOC

X

]

where P in and P out represent electric powers that are input into and output from the node, respectively; Q in and Q out represent thermal powers that are input into and output from the node, respectively; G in and G out represent natural gas flows that are input into and output from the node, respectively; SOC represents a battery capacity of the node; X is a node start-stop state, indicating whether the node accesses the graph convolutional network, wherein X=0 indicates that the node does not access the graph convolutional network, and X=1 indicates that the node accesses the graph convolutional network; and
configuring, by the each node, values of the state variable and the control variable based on a current state and an adopted control strategy, and assigning 0 to a corresponding position for the node that does not comprise a variable;
sub-step S 33 , assigning a weight coefficient to each edge for representing an attribute of each branch; and learning branch characteristics using a feedforward fully connected neural network and outputting an edge weight coefficient W e of a uniform order;
sub-step S 34 , updating the feature vector of each node by feature fusion of adjacent nodes;

V

i
,
z

l
+
1

=

α
⁢

(

∑

j
∈

N
i

e

i
,
j

l

⁢

W
l

⁢

V

j
,
z

l

+

b
l

)

where V i,z l+1 represents the feature vector of a node i at a next layer; N i represents a set of nodes adjacent to the node i, comprising the node i; V j,z l represents a feature vectors of a node j at a current layer; e i,j l represents an edge weighted value between nodes i,j; W 1 and b l represent learnable weight matrix and bias coefficient of the current layer; a represents an activation function; and ReLU function is selected except for a last layer, and a linear activation function is selected for the last layer; and
sub-step S 35 , resetting a new feature vector for each node after completing the feature fusion of nodes, and repeating the step S 34 until a network computing of all layers is completed, to realize a construction of the deep graph neural network model;
step S 4 , constructing a loss function of the deep graph neural network model with physical-informed fusion; and
step S 5 , training the deep graph neural network model with physical-informed fusion according to historical operation data, and performing an optimization control for the integrated energy system using the trained deep graph neural network model with physical-informed fusion, wherein said optimization control further comprises:
collecting physical parameters of each node in the integrated energy in real time, and filtering and standardizing the collected physical parameters to ensure a quality and consistency of the p

Assignees

• Univ Zhejiang

Inventors

• Lin Xiaojie
• Yuan Chongshuo
• Zhong Wei
• Wu Yanling
• Zhou Yi

Classifications

• G05B13/042

  in which a parameter or coefficient is automatically adjusted to optimise the performance · CPC title

• G06Q50/06

  Energy or water supply · CPC title

• G05B13/027 Primary

  using neural networks only · CPC title