Scholarlink

<p>Anaerobic digestion (AD) is a crucial bioenergy source widely applied in wastewater treatment. However, its efficiency improvement is hindered by complex microbial communities and sensitivity to feedstock properties. We, thus, propose a hybrid quantum-classical machine learning (Q-CML) regression algorithm using a quantum circuit learning (QCL) strategy. Combining a variational quantum circuit with a classical optimiser, this approach predicts biogas production from operational data of 18 full-scale mesophilic AD sites in the UK. Tailored for noisy intermediate-scale quantum (NISQ) devices, the low-depth QCL model outperforms conventional regression methods (<i>R</i>²: 0.53) and matches the performance of a classical multi-layer perceptron (MLP) regressor (<i>R</i>²: 0.959) with significantly fewer parameters and better scalability. Comparative analysis highlights the advantages of quantum superposition and entanglement in capturing intricate correlations in AD data. This study positions Q-CML as a cutting-edge solution for optimising AD processes, boosting energy recovery, and driving the circular economy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantum machine learning regression optimisation for full-scale sewage sludge anaerobic digestion

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月5日

查看全文
  • Yomna Mohamed,
  • Ahmed Elghadban,
  • Hei I Lei,
  • Amelie Andrea Shih,
  • Po-Heng Lee

摘要

Anaerobic digestion (AD) is a crucial bioenergy source widely applied in wastewater treatment. However, its efficiency improvement is hindered by complex microbial communities and sensitivity to feedstock properties. We, thus, propose a hybrid quantum-classical machine learning (Q-CML) regression algorithm using a quantum circuit learning (QCL) strategy. Combining a variational quantum circuit with a classical optimiser, this approach predicts biogas production from operational data of 18 full-scale mesophilic AD sites in the UK. Tailored for noisy intermediate-scale quantum (NISQ) devices, the low-depth QCL model outperforms conventional regression methods (R²: 0.53) and matches the performance of a classical multi-layer perceptron (MLP) regressor (R²: 0.959) with significantly fewer parameters and better scalability. Comparative analysis highlights the advantages of quantum superposition and entanglement in capturing intricate correlations in AD data. This study positions Q-CML as a cutting-edge solution for optimising AD processes, boosting energy recovery, and driving the circular economy.