A ML-Based Model for Evaluating the Power Consumption of Network Devices

The power consumption from Information and Communications Technology (ICT) Industry has been regarded as the key factor of global climate change. Furthermore, it has become the most important environmental issue as well. This negative situation has made ICT industry pay more and more attention to power consumption related issues and evaluate the results through Environmental, Social, Governance (ESG). The scope of network run by Communication Service Provider (CSP) is extremely wide including various network devices. Moreover, among these network devices, there are many legacy ones which can't provide query interface for checking power consumption. This situation makes CSPs can't get present power consumption of network devices. CSPs need to find their own ways to evaluating the power consumption of network devices. However, the ways CSPs found normally can't consider environmental factors and make result imprecise. This research referred to suggestions from past researches and use Machine Learning Based (ML-Based) algorithm to build model. We applied the model to Very-high-bit-rate Digital Subscriber Line 2 (VDSL 2) devices and predict the power consumption of these devices. According to the research result, we compared the prediction to the real power consumption and found that the model we proposed has high accuracy within online field. Talking about the construction of the model, we only need key parameters from network devices to evaluate power consumption. With this model, not only can we resolve the problem of many old network devices lack for query interface for checking power consumption but also can we increase the visibility of power status from telecommunications network. In addition, the model can be regarded as the basic of energy conservation either.