Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2410.20755 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910672831381504 |
|---|---|
| author | Guo, Yawen Naderi, Sonia Josephson, Colleen |
| author_facet | Guo, Yawen Naderi, Sonia Josephson, Colleen |
| contents | Solar-powered base stations are a promising approach to sustainable telecommunications infrastructure. However, the successful deployment of solar-powered base stations requires precise prediction of the energy harvested by photovoltaic (PV) panels vs. anticipated energy expenditure in order to achieve affordable yet reliable deployment and operation. This paper introduces an innovative approach to predict energy harvesting by utilizing a novel conditional Long Short-Term Memory (Cond-LSTM) neural network architecture. Compared with LSTM and Transformer models, the Cond-LSTM model reduced the normalized root mean square error (nRMSE) by 69.6% and 42.7%, respectively. We also demonstrate the generalizability of our model across different scenarios. The proposed approach would not only facilitate an accurate cost-optimal PV-battery configuration that meets the outage probability requirements, but also help with site design in regions that lack historical solar energy data. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_20755 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Provisioning for Solar-Powered Base Stations Driven by Conditional LSTM Networks Guo, Yawen Naderi, Sonia Josephson, Colleen Systems and Control Solar-powered base stations are a promising approach to sustainable telecommunications infrastructure. However, the successful deployment of solar-powered base stations requires precise prediction of the energy harvested by photovoltaic (PV) panels vs. anticipated energy expenditure in order to achieve affordable yet reliable deployment and operation. This paper introduces an innovative approach to predict energy harvesting by utilizing a novel conditional Long Short-Term Memory (Cond-LSTM) neural network architecture. Compared with LSTM and Transformer models, the Cond-LSTM model reduced the normalized root mean square error (nRMSE) by 69.6% and 42.7%, respectively. We also demonstrate the generalizability of our model across different scenarios. The proposed approach would not only facilitate an accurate cost-optimal PV-battery configuration that meets the outage probability requirements, but also help with site design in regions that lack historical solar energy data. |
| title | Provisioning for Solar-Powered Base Stations Driven by Conditional LSTM Networks |
| topic | Systems and Control |
| url | https://arxiv.org/abs/2410.20755 |