CN-116801389-B - Wireless energy transmission method, device, equipment and readable storage medium

CN116801389BCN 116801389 BCN116801389 BCN 116801389BCN-116801389-B

Abstract

The application provides a wireless energy transmission method, a device, equipment and a readable storage medium, which can continuously transmit wireless energy for each user so as to enable each user to train a target model locally and transmit the target model which is trained locally, judge whether the instantaneous channel state information of a wireless link is known or not, adjust the wireless energy resource allocation strategy for each user according to the instantaneous channel state information, and optimize the wireless energy resource allocation strategy for each user by adopting the resource allocation strategy based on a deep Q-learning network and a greedy algorithm if the instantaneous channel state information is known. The problem that each user participating in federal learning is forced to interrupt federal learning and cannot upload model parameters due to excessive energy consumption or overlong execution time delay is effectively avoided.

Inventors

CHEN LIMING
LIANG ZHIHONG
XU AIDONG
SUO SILIANG
Huang kaitian
HONG CHAO
Zhi Zhijun

Assignees

南方电网科学研究院有限责任公司

Dates

Publication Date: 20260505
Application Date: 20230628

Claims (7)

1. A method of wireless energy transfer, comprising: Continuously transmitting wireless energy for each user so as to train a target model locally for each user and transmit a target model which is trained locally; Judging whether the instantaneous channel state information or the statistical channel state information of a wireless link in the energy transmission system is known or not; if the instantaneous channel state information of the wireless link in the energy transmission system is known, adopting a resource allocation strategy based on a deep Q-learning network and a greedy algorithm to optimize the wireless energy resource allocation strategy for each user; if only the statistical channel state information of the wireless link in the energy transmission system is known, adopting a resource allocation strategy based on a deep Q-learning network and a golden section method to optimize the wireless energy resource allocation strategy for each user; The creation process of the resource allocation strategy based on the deep Q-learning network and the greedy algorithm comprises the following steps: Dynamically updating the wireless bandwidth obtained by each user from the hybrid access point of the energy transmission system; updating radio frequency power corresponding to wireless bandwidth obtained by each user from the hybrid access point; And updating the wireless charging time of each user at the hybrid access point according to the updated wireless bandwidth and the radio frequency power of each user obtained from the hybrid access point until a loss function of a preset first resource allocation network model converges to obtain a resource allocation strategy of the energy transmission system based on a deep Q-learning network and a greedy algorithm, wherein the preset first resource allocation network model is obtained by training the wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point and the wireless charging time of each user at the hybrid access point as training samples, and training the wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point and the resource allocation strategy corresponding to the wireless charging time of each user at the hybrid access point as sample labels.
2. The method of claim 1, wherein the creating the resource allocation policy based on the deep Q-learning network and golden section method comprises: Dynamically updating the wireless bandwidth obtained by each user from the hybrid access point of the energy transmission system; updating radio frequency power corresponding to wireless bandwidth obtained by each user from the hybrid access point; And updating the corresponding wireless charging time of each user at the hybrid access point according to the updated wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point until a loss function of a preset second resource allocation model network converges to obtain a resource allocation strategy of the energy transmission system based on a deep Q-learning network and a golden section method, wherein the preset second resource allocation network model is obtained by training the wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point and the wireless charging time of each user at the hybrid access point as training samples, and training the wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point and the resource allocation strategy corresponding to the wireless charging time of each user at the hybrid access point as sample labels.
3. The method of claim 1, wherein the continuously transmitting wireless energy for each user comprises: dividing the transmission time for transmitting wireless energy to each user into a plurality of time slot segments with equal span; when each time slot section starts, each user is subjected to wireless charging; Wherein, the The process of transmitting wireless energy to each user is as follows: Wherein, the Representing the time each user locally trains the target model; representing the time of each user transmitting the target model; Representing a wireless charging time for each user; representing a charge time ratio for each user; Representing the duration of a round of federal learning; Wherein, the Wherein, the A data set size representing each user; representing the CPU cycles required to calculate one sample of data; Representing the number of local user training rounds in a federal learning round; representing the computing power of each user; a model size representing each user; Representing the transmission rate of each user.
4. The method according to claim 1 or 2, wherein said updating the radio frequency power corresponding to the radio bandwidth obtained by each user from the hybrid access point comprises: Sequencing the radio frequency power required by each user; And updating the radio frequency power corresponding to the wireless bandwidth obtained by each user from the hybrid access point according to the radio frequency power sequencing result required by each user.
5. A wireless energy transfer apparatus, comprising: The transmission unit is used for continuously transmitting wireless energy for each user so as to enable each user to train the target model locally and transmit the target model which is trained locally; The judging unit is used for judging whether the instantaneous channel state information or the statistical channel state information of the wireless link in the energy transmission system is known; the first optimizing unit is used for optimizing the wireless energy resource allocation strategy for each user by adopting a preset resource allocation strategy based on a depth Q-learning network and a greedy algorithm when the execution result of the judging unit is that the instantaneous channel state information of the wireless link in the energy transmission system is known; The second optimizing unit is used for optimizing the wireless energy resource allocation strategy for each user by adopting the resource allocation strategy based on the deep Q-learning network and the golden section method when the execution result of the judging unit is that only the statistical channel state information of the wireless link in the energy transmission system is known; The creation process of the resource allocation strategy based on the deep Q-learning network and the greedy algorithm comprises the following steps: Dynamically updating the wireless bandwidth obtained by each user from the hybrid access point of the energy transmission system; updating radio frequency power corresponding to wireless bandwidth obtained by each user from the hybrid access point; And updating the wireless charging time of each user at the hybrid access point according to the updated wireless bandwidth and the radio frequency power of each user obtained from the hybrid access point until a loss function of a preset first resource allocation network model converges to obtain a resource allocation strategy of the energy transmission system based on a deep Q-learning network and a greedy algorithm, wherein the preset first resource allocation network model is obtained by training the wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point and the wireless charging time of each user at the hybrid access point as training samples, and training the wireless bandwidth and the corresponding radio frequency power of each user obtained from the hybrid access point and the resource allocation strategy corresponding to the wireless charging time of each user at the hybrid access point as sample labels.
6. A wireless energy transfer apparatus comprising one or more processors and a memory; Stored in the memory are computer readable instructions which, when executed by the one or more processors, implement the steps of the wireless energy transfer method of any one of claims 1 to 4.
7. A readable storage medium having stored therein computer readable instructions which, when executed by one or more processors, cause the one or more processors to implement the steps of the wireless energy transfer method of any of claims 1 to 4.

Description

Wireless energy transmission method, device, equipment and readable storage medium Technical Field The present application relates to the field of energy transmission technologies, and in particular, to a wireless energy transmission method, device, apparatus, and readable storage medium. Background The construction process of the digital power grid is the process of digitalization, intellectualization and internetworking of the traditional power grid. The digital transformation is carried out on the traditional power grid, a corresponding digital twin power grid is required to be constructed, an advanced digital technology platform is used, a powerful computing power is formed by a computing power, data, a model and an algorithm, the power grid company has super-strong sensing capability, intelligent decision capability and quick execution capability by means of sensing, analysis, decision making, business and other links of all parties related to the power grid on the basis of the Internet of things, the Internet, the boundary of the digital power grid is expanded to the social aspect from the traditional power grid, management, operation and service modes of the traditional power grid are changed, the wide configuration of energy flow, fund flow, logistics, business flow and talent flow of related industries is driven, the power is pushed to revolution and new energy system construction by the aid of electric power, the national economic system is modernized, and a new digital power grid system with safety body is constructed. The rise of artificial intelligence provides a new view angle and thought for the development of digital power grids. With the rapid development of artificial intelligence technology, deep learning has shown great potential in the fields of computer vision, signal processing, wireless communication and the like. Since in centralized deep learning, a large amount of data needs to be collected from distributed users and uploaded to a central server to support centralized training, this brings a serious risk of privacy disclosure. As a result, more and more users are reluctant to share their private data, which results in a data island situation. To address this problem, federal learning concepts have been proposed with the aim of facilitating decentralized intelligence without compromising privacy. In the federal learning framework, users can train a global model by uploading model parameters instead of private original data and cooperatively with the federal learning server, so that the safety of the private data can be ensured, and meanwhile, the communication overhead is reduced. However, energy consumption in federal learning is a key problem in wireless networks, including federal learning networks, which determines how long users in the network can operate, and the number of users participating in federal learning is directly related to the performance of the power grid system. How to solve the problem that users are forced to exit federal learning due to energy consumption when participating in federal learning is a constant concern. Disclosure of Invention The present application is directed to at least solving one of the above-mentioned technical drawbacks, and accordingly, the present application provides a wireless energy transmission method, device, apparatus and readable storage medium, which are used for solving the technical drawbacks of the prior art that users participate in federal learning and are forced to exit federal learning due to the difficulty in solving the energy consumption problem. A wireless energy transfer method, comprising: Continuously transmitting wireless energy for each user so as to train a target model locally for each user and transmit a target model which is trained locally; Judging whether the instantaneous channel state information or the statistical channel state information of a wireless link in the energy transmission system is known or not; If the instantaneous channel state information of the wireless link in the energy transmission system is known, a preset resource allocation strategy based on a deep Q-learning network and a greedy algorithm is adopted, and the wireless energy resource allocation strategy for each user is optimized. Preferably, the method further comprises: If only the statistical channel state information of the wireless link in the energy transmission system is known, adopting a resource allocation strategy based on a coupling depth Q-learning network and a golden section method to optimize the wireless energy resource allocation strategy for each user. Preferably, the creating process of the resource allocation strategy based on the deep Q-learning network and the greedy algorithm comprises the following steps: Dynamically updating the wireless bandwidth obtained by each user from the hybrid access point of the energy transmission system; updating radio frequency power corresponding to wireless bandwidth obtained