Title: Dynamic prediction of photovoltaic maximum power based on sensor network multimodal data and transformer-LSTM

Authors: Kelai Zhang; Xiduan Chen; Ru Qiu

Addresses: School of Artificial Intelligence, Zhejiang Industry and Trade Vocational College, Wenzhou, 325002, China ' School of Artificial Intelligence, Zhejiang Industry and Trade Vocational College, Wenzhou, 325002, China ' School of Artificial Intelligence, Zhejiang Industry and Trade Vocational College, Wenzhou, 325002, China

Abstract: Accurate prediction of maximum power is crucial for enhancing generation efficiency in photovoltaic power generation. However, traditional forecasting methods often struggle to comprehensively reflect the real-time variations and combined effects of multiple environmental factors such as sunlight, temperature, and humidity, resulting in limited prediction accuracy. To overcome this limitation, this paper first proposes a photovoltaic data monitoring system based on wireless sensor networks. By deploying Zigbee network nodes on each photovoltaic module, the system collects and transmits environmental and operational data in real time, ensuring the reliability of photovoltaic power generation data acquisition. Building upon this foundation, this paper developed a predictive model that integrates long short-term memory networks with an enhanced transformer model. Experimental results demonstrate that the method achieves a data acquisition efficiency of 96.5% and improves model fitting accuracy by at least 5.75%. It enables faster and more precise predictions of maximum photovoltaic power.

Keywords: photovoltaic power generation system; maximum power prediction; multimodal sensor network; transformer model; long short-term memory network.

DOI: 10.1504/IJSNET.2026.153835

International Journal of Sensor Networks, 2026 Vol.51 No.1, pp.45 - 60

Received: 09 Nov 2025
Accepted: 14 Nov 2025
Published online: 27 May 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article