CN-121688874-B - Power equipment control method and system based on Internet of things

CN121688874BCN 121688874 BCN121688874 BCN 121688874BCN-121688874-B

Abstract

The invention belongs to the technical field of equipment control, and particularly relates to an electric equipment control method and system based on the Internet of things, comprising the following steps of S1, acquiring real-time operation data of nodes of all electric equipment in the Internet of things, and constructing a second-order state tensor for each node, wherein the second-order state tensor represents the operation state of the node; the method comprises the steps of obtaining a low-dimensional state descriptor by a first-order state tensor, projecting the second-order state tensor to a low-dimensional subspace defined by a base vector, obtaining the low-dimensional state descriptor, adaptively updating the base vector according to historical state covariance matrixes of all nodes, calculating Wasserstein distances of low-dimensional state descriptor subsequences among nodes, generating a weighted transfer matrix for parameter aggregation among the nodes by combining with communication link quality, and starting local optimization if the condition number of the local state covariance matrixes of the nodes exceeds a first threshold. The invention improves the convergence speed and precision of the optimization algorithm, and realizes more stable cooperative control on the power equipment group through a secondary regulation mechanism combining local optimization and global correction.

Inventors

ZHANG WENJING
Hang Chencong
WANG PENGFEI
SONG LEI
ZHU YAHUI
DAI KE
DAI LINLIN
LI DONGYA
XU HAIBO
LI YU
LIU XINGLI

Assignees

陕西斯锐明天智能设备有限公司
国网河南省电力公司新蔡县供电公司

Dates

Publication Date: 20260505
Application Date: 20260209

Claims (10)

1. The power equipment control method based on the Internet of things is characterized by comprising the following steps of: The method comprises the steps of S1, acquiring real-time operation data of nodes of all power equipment in the Internet of things, constructing a second-order state tensor representing the operation state of the nodes for each node, projecting the second-order state tensor to a low-dimensional subspace defined by a basis vector to obtain a low-dimensional state descriptor, and adaptively updating the basis vector according to a historical state covariance matrix of all the nodes; S2, if the condition number of the local state covariance matrix of the node exceeds a first threshold, starting local optimization, namely aggregating control parameters of neighbor nodes by using a weighted transfer matrix, combining self Riemann gradient, and updating the control parameters of the node through exponential mapping; s3, decoding the updated control parameters into specific control instructions and executing the specific control instructions.
2. The method for controlling power equipment based on the internet of things according to claim 1, wherein acquiring real-time operation data of nodes of each power equipment in the internet of things, constructing a second-order state tensor representing an operation state of each node, comprises: Collecting voltage U, current I, frequency f, active power P and reactive power Q of nodes of each power device in 50 continuous sampling time points as a real-time operation data sequence; Taking the collected data sequences of the five physical quantities as five-dimensional state vectors of the nodes; and calculating a covariance matrix of the five-dimensional state vector of the node at 50 sampling time points, and taking the covariance matrix as a second-order state tensor.
3. The power equipment control method based on the internet of things according to claim 1, wherein the base vector is adaptively updated according to the historical state covariance matrix of all the nodes, and the method comprises the following steps: Collecting historical state covariance matrixes of all nodes in the past 100 periods, and calculating an arithmetic average value of the historical state covariance matrixes to obtain a global average covariance matrix; performing eigenvalue decomposition on the global average covariance matrix; And selecting the feature vector corresponding to the first 10 maximum feature values as the updated base vector.
4. The power equipment control method based on the internet of things according to claim 1, wherein calculating the wasperstein distance of the low-dimensional state description subsequence between nodes, and combining the quality of the communication link, generating a weighted transfer matrix for parameter aggregation between nodes, comprises: acquiring a bidirectional communication packet loss rate between a node i and a neighbor node j And average round trip delay Calculating a communication link quality factor, the communication link quality factor ; Obtaining Wasserstein distance of low-dimensional state description subsequence between node i and neighbor node j ; According to the formula Calculating the original weight of the neighbor node j to the node i ; And carrying out normalization processing on the original weights of all neighbor nodes of the node i to obtain an element of the ith row in the weighted transfer matrix.
5. The method for controlling power equipment based on the internet of things according to claim 1, wherein if the condition number of the local state covariance matrix of the node exceeds a first threshold, starting local optimization, wherein aggregating control parameters of neighbor nodes by using a weighted transfer matrix, and updating the control parameters of the node by exponential mapping in combination with self Riemann gradient comprises: When the condition number of the local state covariance matrix of the node i exceeds a first threshold, the node performs the following operations: obtaining aggregation parameters by using control parameters of all neighbor nodes of weighting transfer matrix aggregation node ; Calculating self-control parameters Riemann gradient of (F) ; The formula combined by exponential mapping and logarithmic mapping: Updating the control parameter, wherein alpha is the learning rate, wherein, For the index map to be an index map, In order to make a logarithmic mapping, Is the updated control parameter.
6. The method for controlling power equipment based on the internet of things according to any one of claims 1 to 5, wherein if the average value of the distance between the global nodes exceeds the second threshold, starting global correction, namely solving the sparse optimization model by the node with the highest centrality in the network to obtain a global correction parameter, and sending the global correction parameter to each node to adjust the control parameter of the node, wherein the method comprises the following steps: When the arithmetic average of the Wasserstein distances between the global nodes exceeds a second threshold, performing the following operations: Calculating the intermediation centrality of all nodes in the network topology, and selecting the node with the highest intermediation centrality as a central node; the center node builds a LASSO regression model: wherein A is a global state feature matrix, b is an ideal state vector, x is a global correction parameter to be solved, and lambda is a regularization coefficient; solving the LASSO regression model to obtain a sparse global correction parameter x, and issuing the global correction parameter.
7. Electric power equipment control system based on thing networking, its characterized in that includes following module: The projection and aggregation module is used for acquiring real-time operation data of nodes of all electric equipment in the Internet of things, constructing a second-order state tensor representing the operation state of the nodes for each node, projecting the second-order state tensor to a low-dimensional subspace defined by a basis vector to obtain a low-dimensional state descriptor, and adaptively updating the basis vector according to a historical state covariance matrix of all the nodes; the correction module is used for starting local optimization if the condition number of the local state covariance matrix of the node exceeds a first threshold value, aggregating the control parameters of neighbor nodes by using a weighted transfer matrix, combining the self Riemann gradient, and updating the control parameters of the node through exponential mapping; And the issuing module is used for decoding the updated control parameters into specific control instructions and executing the specific control instructions.
8. The power equipment control system based on the internet of things according to claim 7, wherein acquiring real-time operation data of nodes of each power equipment in the internet of things, constructing a second-order state tensor representing an operation state of each node, comprises: Collecting voltage U, current I, frequency f, active power P and reactive power Q of nodes of each power device in 50 continuous sampling time points as a real-time operation data sequence; Taking the collected data sequences of the five physical quantities as five-dimensional state vectors of the nodes; and calculating a covariance matrix of the five-dimensional state vector of the node at 50 sampling time points, and taking the covariance matrix as a second-order state tensor.
9. The power equipment control system based on the internet of things according to claim 7, wherein the base vector is adaptively updated according to the historical state covariance matrix of all nodes, comprising: Collecting historical state covariance matrixes of all nodes in the past 100 periods, and calculating an arithmetic average value of the historical state covariance matrixes to obtain a global average covariance matrix; performing eigenvalue decomposition on the global average covariance matrix; And selecting the feature vector corresponding to the first 10 maximum feature values as the updated base vector.
10. The power equipment control system based on the internet of things according to claim 7, wherein calculating the wasperstein distance of the low-dimensional state description subsequence between nodes, in combination with the communication link quality, generates a weighted transfer matrix for parameter aggregation between nodes, comprises: acquiring a bidirectional communication packet loss rate between a node i and a neighbor node j And average round trip delay Calculating a communication link quality factor, the communication link quality factor ; Obtaining Wasserstein distance of low-dimensional state description subsequence between node i and neighbor node j ; According to the formula Calculating the original weight of the neighbor node j to the node i ; And carrying out normalization processing on the original weights of all neighbor nodes of the node i to obtain an element of the ith row in the weighted transfer matrix.

Description

Power equipment control method and system based on Internet of things Technical Field The invention belongs to the technical field of equipment control, and particularly relates to an electric equipment control method and system based on the Internet of things. Background Along with the integration of the internet of things technology in a modern power system, power equipment nodes are interconnected, and the sensing and the regulation of the running state of a power grid are realized. However, the operation data generated by each equipment node has the characteristics of high dimensionality, isomerism and strong dynamic property, and the control method based on the simplified model is difficult to express the complex operation state of the equipment and the nonlinear association of the equipment. Moreover, the communication environment of the Internet of things is complex, uncertainty such as time delay and packet loss exists, and higher requirements are put on the instantaneity of a control strategy. The existing power equipment control method mostly adopts a centralized architecture or a traditional distributed cooperative strategy. The centralized method faces large communication overhead, calculation bottleneck and single point fault risk, and is difficult to adapt to the expansion requirement of a large-scale equipment network. While the distributed method improves the flexibility of the system, the control precision is often reduced due to insufficient information representation when processing high-dimensional data, and the dynamic difference of node states and the actual quality of communication links are mostly not fully considered, so that the efficiency and the control effect of parameter coordination are affected. And simply vectorizing the device operational data will lose structural information inherent in the device. In the aspect of coordination among nodes, how to effectively measure the differences among different node state sequences and perform self-adaptive parameter aggregation according to the differences is a key for improving the distributed control performance. The conventional gradient optimization algorithm is adopted in the Euclidean space, so that not only can the inherent manifold constraint of the data be destroyed, but also the problem of low convergence speed and even non-convergence can be caused. Therefore, a new power equipment control method capable of adapting to dynamic changes and performing efficient optimization is needed to realize safe and stable operation of a power system in the environment of the internet of things. Disclosure of Invention The invention provides a power equipment control method and a system based on the Internet of things, which are used for solving the technical problems of the defects of a centralized architecture in power equipment control in the prior art and the defects of the traditional distributed method in the aspects of high-data processing, dynamic difference adaptation, manifold constraint maintenance, optimization efficiency and the like. In a first aspect, the present invention provides a method for controlling electric power equipment based on the internet of things, including the following steps: The method comprises the steps of S1, acquiring real-time operation data of nodes of all power equipment in the Internet of things, constructing a second-order state tensor representing the operation state of the nodes for each node, projecting the second-order state tensor to a low-dimensional subspace defined by a basis vector to obtain a low-dimensional state descriptor, and adaptively updating the basis vector according to a historical state covariance matrix of all the nodes; S2, if the condition number of the local state covariance matrix of the node exceeds a first threshold, starting local optimization, namely aggregating control parameters of neighbor nodes by using a weighted transfer matrix, combining self Riemann gradient, and updating the control parameters of the node through exponential mapping; s3, decoding the updated control parameters into specific control instructions and executing the specific control instructions. Further, acquiring real-time operation data of nodes of each power equipment in the internet of things, and constructing a second-order state tensor for each node, wherein the second-order state tensor represents the operation state of the node comprises the following steps: Collecting voltage U, current I, frequency f, active power P and reactive power Q of nodes of each power device in 50 continuous sampling time points as a real-time operation data sequence; Taking the collected data sequences of the five physical quantities as five-dimensional state vectors of the nodes; and calculating a covariance matrix of the five-dimensional state vector of the node at 50 sampling time points, and taking the covariance matrix as a second-order state tensor. Further, the base vector is adaptively updated accordi