Vol.:(0123456789) 1 3 Journal of Ambient Intelligence and Humanized Computing (2020) 11:209–236 https://doi.org/10.1007/s12652-019-01299-x ORIGINAL RESEARCH A fog based load forecasting strategy based on multi‑ensemble classifcation for smart grids Asmaa H. Rabie 1  · Shereen H. Ali 2  · Ahmed I. Saleh 1  · Hesham A. Ali 1 Received: 7 December 2018 / Accepted: 12 April 2019 / Published online: 20 April 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019 Abstract Internet of things (IoT) improves the development and operation of smart electrical grids (SEGs). Overcoming the cloud challenges, 2-tier architecture is replaced by 3-tier one for including a fog computing tier that acts as a bridge between the IoT devices embedded in SEG and cloud. Load forecasting is an essential process for the electrical system operation and planning as it provides intelligence to energy management. This paper completes the electrical load forecasting (ELF) strategy introduced (Rabie et al. in Cluster Comput 22(1):241–270, 2019). ELF consists of two phases which are; (1) data pre-processing Phase (DP 2 ) and (2) load prediction phase (LP 2 ). Both phases can be performed in the cloud tier on the stored data which is sent from fogs to cloud at cloud servers. DP 2 aims to perform feature selection and outlier rejection using data mining techniques. The main contribution of this paper focuses on LP 2 providing multi-ensemble load prediction (MELP) method which can be learned by the fltered data from DP 2 to give fast and accurate predictions. MELP can deal with big electrical data based on Map-Reduce method. It mainly consists of two levels which are; (1) local ensemble level (LEL) in map phase and (2) global ensemble level (GEL) in reduce phase. In LEL, the ensemble classifcation principle is applied at every device in map phase. In GEL, the perfect and fnal decision for load prediction is taken in reduce phase based on global judger (GJ) method from many local predictions which are the results of all devices in map phase. The conducted experi- mental results have shown that the proposed MELP outperforms recent prediction methods in terms of accuracy, precision, recall, F1-measure, and run time. It is concluded that the proposed MELP method can deal with big electrical data. It has a good impact in maximizing system reliability, resilience, and stability as it introduces fast and accurate load predictions. Keywords IoT · FOG · Cloud · Load forecasting 1 Introduction SEGs have already achieved wide adoption in information sensing, transmission, and processing (Rabie et al. 2015; Saleh et al. 2016). IoT has been growing rapidly owing to recent advancements in communications and sensor tech- nologies (Ozger et al. 2018). Hence, IoT helps SEGs to support a lot of tasks throughout the generation, transmis- sion, distribution, and consumption of energy (Ozger et al. 2018; Mahajan and Patil 2016). IoT reduced human data entry eforts by using diferent types of sensors which are used to collect electrical data from SEGs. An emerging wave of IoT requires mobility support, geo-distribution, location awareness, and low latency. To meet these requirements, a new platform called fog computing is needed to store and process IoT data locally at IoT devices (Barik et al. 2019; Atlam et al. 2018). Although the integration of IoT with cloud overcomes a lot of issues such as performance, secu- rity, privacy, and reliability, cloud brings many challenges. One challenge is related to how subsequently analysis of huge amount of data that is collected from smart meters and other IoT devices to determine decisions regarding various actions. Besides, sending whole captured data directly to cloud requires excessively high network bandwidth (Atlam et al. 2018). To address these challenges, fog computing plays an essential role for extending cloud computing to the edge of the network. * Asmaa H. Rabie asmaa91hamdy@yahoo.com 1 Computers and Systems Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt 2 Communications and Electronics Engineering Department, Delta Higher Institute for Engineering &Technology, Mansoura, Egypt