A fog based load forecasting strategy for smart grids using big electrical data

Abstract

Internet of things (IoT) enables the smart electrical grids (SEGs) to support a lot of tasks throughout the generation, transmission, distribution and consumption of energy. Large amounts of data are generated through these smart devices, which can be sent to the cloud for further processing. Hence, appropriate actions can be taken based on the cloud processing. However, sending the complete captured data directly to the cloud would lead to resource wastage. To overcome the cloud challenges, fog computing tier acts as a bridge in the middle between the IoT devices embedded in SEG and the cloud. Thus, a 3-tiers architecture is proposed to replace 2-tiers one. The added fog tier offers a place for collecting, computing, storing smart meter data before transmitting them to the cloud. In this paper, a new electrical load forecasting (ELF) strategy has been proposed based on the pre-mentioned three tiers architecture. The proposed strategy consists of two phases; which are; (i) data pre-processing phase (DP2) and (ii) load prediction phase (LP2). Both phases take place at a cloud data center (CDC) on the collected data, which is received from all fogs connected to the entire cloud. The aim of the data preprocessing phase is to; (i) reject outliers from the collected data, and (ii) feature selection. The main contribution of this paper lied on the feature selection, which elects only the effective features for the next load prediction phase. Features selection is an essential pre-processing operation to reduce the dataset dimensions from all irrelevant features leading to enable the ELF from taking fast and correct decision for other electrical subsystems. A new feature selection methodology called fuzzy based feature selection (FBFS) is presented, which involves two stages; feature ranking stage (FRS) and feature selection stage (FS2). FBFS is a hybrid method based on both features selection methods filter and wrapper providing fast and more accurate significant features. Then, during LP2, the filtered data is used to give a fast and accurate load prediction. Experimental results have shown that the proposed FBFS promotes the load forecasting efficiency in terms of precision, recall, accuracy and F-measure compared with recent features selection methodologies. FBFS provides the highest precision, recall, and accuracy results. On the other hand, it provides the lowest error value.