A low-complexity non-intrusive approach to predict the energy demand of buildings over short-term horizons

ABSTRACT

Reliable, non-intrusive, short-term (of up to 12 h ahead) prediction of a building's energy demand is a critical component of intelligent energy management applications. A number of such approaches have been proposed over time, utilizing various statistical and, more recently, machine learning techniques, such as decision trees, neural networks and support vector machines. Importantly, all of these works barely outperform simple seasonal auto-regressive integrated moving average models, while their complexity is significantly higher. In this work, we propose a novel low-complexity non-intrusive approach that improves the predictive accuracy of the state-of-the-art by up to $\sim 10\text{\%}$. The backbone of our approach is a K-nearest neighbours search method, that exploits the demand pattern of the most similar historical days, and incorporates appropriate time-series pre-processing and easing. In the context of this work, we evaluate our approach against state-of-the-art methods and provide insights on their performance.

KEYWORDS:

• Energy demand
• energy consumption
• forecasting
• smart buildings

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Such energy efficiency-improving approaches have long been investigated in the context of supervisory heating, ventilation and air conditioning (HVAC) control, due to their low cost compared to insulation and heating technology improvements (Wang & Ma, 2008). For instance, replacing an old thermostat with a smart one can be less costly than improving the insulation of a building, and has substantial energy efficiency improvement potential (Wang & Ma, 2008).

2 In particular, preheating generally requires additional energy when compared to ‘on-time’ heating strategies. This is the case, as more energy is needed to get the temperature above the desired level—to then let it drop to the desired one. However, the energy used for preheating can potentially be cheaper, such as energy coming from self-owned renewable generators. This, in turn, can lead to lower energy bills. Nevertheless, this is not the case in this setting and, as such, preheating would only increase the overall cost.

3 We note that the reminder of the day can be appropriately augmented according to the predictive horizon needs of the application.

4 This easing technique has been selected after experimenting with various other easing techniques and considering its low complexity (when compared, for instance, with Gaussian-processes-based easing Panagopoulos et al., 2017).

5 A discussion on adjusting these parameters is provided in Section 4 where, essentially, a grid search approach is followed to instantiate our approach for our case study data-set.