"Establishment of a long short-term memory (LSTM) model using the Hilbert-Huang transform (HHT) to forecast the current outputs of biophotovoltaic dev

“使用希尔伯特-黄变换（HHT）建立长短期记忆（LSTM）模型来预测生物光伏开发的当前输出

• 批准号: 2801696
• 金额: --
• 依托单位: University of Cambridge
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2801696
• 关键词: Establishment; long; short; term; memory

Striving towards renewable energy technologies, biophotovoltaic devices (BPVs), which generate electricity usingsunlight and microbes like cyanobacteria, can pose a great potential. However, several challenges such as a low poweroutput and optimisation difficulties (due to many unpredictable parameters) still need to be met for commercialisation.With the goal to overcome some of these challenges, Tonny Okedi (as a PhD student in Dr Adrian Fisher's researchgroup) decomposed BPVs' current outputs using the Hilbert-Huang transform (HHT). He concluded that the obtainedoscillatory subcomponents, so-called intrinsic model functions (IMFs), may create the basis for a long short-term memory(LSTM) model which can more accurately predict the BPVs' current outputs at various conditions. Building upon Tonny'swork, my PhD research may aim to determine the IMFs' individual physical meanings through experiments. Moreover,the individual IMFs could be modelled and subsequently combined to establish an LSTM model that forecasts the BPVs'current outputs. Lastly, by further examining the BPVs' sensitivities with regards to, for example, light intensities andinorganic carbon levels, a more accurate LSTM model can be developed that estimates current outputs for a wider rangeof operating conditions. As a result, the obtained LSTM model of BPVs could represent a great advancement withregards to BPVs' optimisation and, consequently, their commercialisation.

致力于可再生能源技术的生物光伏设备（BPV）利用阳光和蓝藻等微生物发电，具有巨大的潜力。然而，商业化仍需要克服一些挑战，例如低功率输出和优化困难（由于许多不可预测的参数）。为了克服其中一些挑战，Tonny Okedi（作为 Adrian Fisher 博士研究组的博士生）使用希尔伯特-黄变换（HHT）分解了 BPV 的电流输出。他得出的结论是，所获得的振荡子组件，即所谓的内在模型函数（IMF），可以为长短期记忆（LSTM）模型奠定基础，该模型可以更准确地预测各种条件下 BPV 的电流输出。在托尼的工作的基础上，我的博士研究可能旨在通过实验确定 IMF 各自的物理意义。此外，可以对各个 IMF 进行建模，然后将其组合起来建立一个 LSTM 模型来预测 BPV 的当前输出。最后，通过进一步检查 BPV 对光强度和无机碳水平等的敏感性，可以开发更准确的 LSTM 模型，该模型可以估计更广泛操作条件下的电流输出。因此，所获得的 BPV LSTM 模型可能代表着 BPV 优化及其商业化方面的巨大进步。