In this paper, we propose a potential solution to enhance energy-harvesting and spectral efficiency through the combination of double Intelligent Reflecting Surfaces (double IRS) and cooperative NOMA. A machine learning (ML) algorithm has been adapted as a powerful tool for optimizing the double IRS phase shifts to enhance the performance of the wireless communication system. Primarily, we aim to optimize the double IRS phase shifts for aligning the reflection angle of the electromagnetic wave towards the intended users, which, in turn, enhances the spectral efficiency. Moreover, we have optimized the Wireless Energy Transmission (WET) time to ensure a sufficient time slot for harvesting energy wirelessly by users’ devices. The performance of our proposed scheme has been compared with the traditional optimization algorithms numerically to reveal the effectiveness of ML in optimizing double IRS phase shifts and maximizing the sum rate. Numerical results indicate that: The achievable rate of the machine learning-based IRS phase shift optimization is superior to all benchmarks, i.e., SDP-based method Random phase, and Single IRS. Furthermore, it demonstrates improved wireless energy transfer (WET) efficiency, indicating better energy harvesting and enhanced system sustainability.