Search

Recommender systems continually impact multiple verticals by introducing automated intelligence to decision making. When applying such Artificial Intelligence (AI) tools to energy efficiency problems, a number of opportunities and challenges present themselves. This paper presents a modular recommender system for improving domestic household energy savings. The recommender relies upon a contextual appliance-level energy dataset from seven appliances in a household. Modularity is incorporated into the system design to create customizable sub-components that adapt to the nature of the data and the end-user’s preference, such as modules that recommend based on usage patterns, power consumption, and occupancy. Machine Learning (ML) has been used for automatic appliance profiling and rank-based methods are employed to evaluate the recommender based on relevance scores. Implementation results for generating recommendations for two weeks yield a Root Mean Square Error (RMSE) of 0.2288, Normalized Cumulative Discounted Gain (NCDG) of 0.729 for seven appliances. Future work includes evaluation on edge computing platforms and user testing through a mobile application.

Gases may be detected and identified with the aid of electronic nose (EN) that employ machine learning (ML) in conjunction with sensor arrays. In the field of smart gas sensing, sensor arrays are very important because the signals they send are used by ML algorithms to learn about the volatile organic compounds (VOCs) that have been found. EN is a powerful device which have the sense ability like human to detect the different smells. Due to its detection power, a research work is done based on EN collected data, where the system is equipped with seven Figaro Taguchi Series (TGS) sensors for detecting six different gases with their concentration levels. The built EN system for the collection of datasets, allows us to make customisable mixes of these gases and gather data automatically with our sensing equipment. The obtained dataset is used to enhance the accuracy and classification performance of the overall system by using the ML models. Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (k-NN), Decision Tree (DT) and Gradient Boosting Tree (GBT) machine learning algorithms are used in this work to boost the model performance. After a deep work on designed models, SVM performs better than the others due to its multiclass classification ability and helps to measures a 99.73% of classification accuracy while the others have less performance like RF (99.70%), k-NN (92.93%), DT (98.10%) and GBT (98.90%

With the varying applications of Artificial Intelligence (AI) on an ever expanding global scale, performance metrics and standards of computing platforms have steady elevated. High-Performance Edge Computing (HPEC) can play an instrumental role in lifting a substantial load on cloud computing solutions when it comes to deploying Deep Learning (DL) algorithms for classifying big data. Notwithstanding, the collection, pre-processing, and post-processing of big data engenders many challenges and opportunities for optimizing HPEC performance for data classification and, in turn, yield better outcomes. This is where the concept of data lakes is employed, which are raw-formatted large masses of data that are plausibly more compatible with many algorithms than structured data stores. Therefore, in this book chapter, we aim to carry out a comparative study that examines the merits, performance and efficiency metrics, and limitations of two notable HPECs centered around a DL data lake classification framework. Comparing the HPEC platforms, with similar classification accuracy of ~90%, results show that adequate performance and impressive computational efficiency, as fast as 8.19 msec per classified GAF image is achieved on the Jetson Nano.

The industrial revolution has elevated science and engineering to foster the development of Image Processing and Artificial Intelligence (AI) and put the visualization of information on an even higher pedestal. Yet, the demands of the industrial age have contributed to an ever-growing wildfire of climate change, sparking a revolution in energy efficiency research. With the aim to advance energy efficiency research from an AI standpoint, a novel transformation of raw-formatted data repositories, known as data lakes, into multi-dimensional visualizations data coupled with computationally lightweight, edge-based AI implementations are proposed as means to understand the energy consumption patterns in buildings. As a novel method of understanding energy data visually, current results comprise a Multi-Dimensional Gramian Angular Field (GAF) representation of energy data as both 2D and 3D interactive forms. Moreover, a case study on deep learning classification employed on ODROID-XU4 yields ~90% accuracy and a classification rate of 17.5 msec/image.

https://attend.ieee.org/dsc-2022/sicsa-event/ Deep Learning (DL) is increasingly empowering technology and engineering in a plethora of ways, especially when big data processing is a core requirement. Many challenges, however, arise when solely depending on cloud computing for Artificial Intelligence (AI), such as data privacy, communication latency, and power consumption. Despite the elevating popularity of edge computing, its overarching issue is not the lack of technical specifications in many edge computing platforms but the sparsity of comprehensive documentation on how to correctly utilize hardware to run ML and DL algorithms. Due to its specialized nature, installing the full version of TensorFlow, a common ML library, on an edge device is a complicated procedure that is seldom successful, due to the many dependent software libraries needed to be compatible with varying architectures of edge computing devices. Henceforth, in this paper, we present a novel technical guide on setting up the TensorFlow Lite, a lightweight version of TensorFlow and demonstrate a complete workflow of model training, validation, and testing on the ODROID-XU4. Results are presented for a case study on energy data classification using the outlined model show almost 7 times higher computational performance compared to cloud-based AI.

Data, in its purest nature, has an authority on the systems it accompanies by feeding an accurate representation of the observed reality. In energy efficiency, the underlying motivation for big data efforts revolves around the intrinsic need to understand end-user electric energy consumption and means to improve it. Hence, developing a rich, detailed, and realistic power consumption dataset entails a deliberate process of preparing the data collection environment, configuring proper Internet of Energy (IoE) sensors and managing the collected data. In this work, a novel power consumption dataset is presented in efforts to improve the state-of-the-art of energy efficiency research in buildings. The dataset is also accompanied by a two-dimensional (2D) counterpart produced using Gramian Angular Fields (GAF) that creates pictorial summaries from one-dimensional (1D) data. Data acquisition is carried out using the ODROIDXU4 edge computing hub, Home Assistant software, and a collection of smart plugs and sensors. A notable use case is presented to signify the merits of the data and its analysis tools to achieve computationally efficient classification.

The Internet of Energy (IoE) paradigm is an advancing area of research concerning the fusion of smart technology and energy efficiency [1], combing data collection, processing, and visualization. Smart energy monitoring witnesses technological advancements such as smart metering and IoE networking, allowing the expansion of smart energy networks in a smart house. In this research, we aim to understand energy behavior through big data collection and classification and improve energy efficiency using behavioral economics, deep learning-based recommender systems, and intuitive data visualizations. In specific, a specialized case study is reported on the ODROID XU4 platform [3], and a setup developed at De Montfort University (DMU) at the Energy Lab and AI Lab, it is aimed to build a novel appliance level dataset with contextual ambient environmental data. As a novel advancement in the field, the ODROID performs edge deep learning computations on the collected data, to clean it, summarize it, anonymize it, and classification, it transmits it to a cloud server for further deep processing and storage. Concluding, the proposed work provides aids in exploiting energy-efficiency technologies for improving energy efficiency via an innovative, automated energy efficiency deep learning engine.

open access proceedings With global pollution and buildings power consumption on the rise, energy efficiency research has never been more necessary. Accordingly, data visualization is one of the most sought challenges in data analysis, especially in energy efficiency applications. In this paper, a novel micro-moment Gramian angular fields time-series transformation of energy signals and ambient conditions, abbreviated as M2 GAF, is described. The proposed tool can be used by energy efficiency researchers to yield a deeper understanding of building energy consumption data and its environmental conditions. Current results show sample G2 GAF representation for three power consumption datasets. In summary, the proposed tool can unveil novel energy time-series analysis possibilities as well as original data visualization that can yield deeper insights, and in turn, improved energy efficiency.

Shape from X is an interesting area of research in computer vision community. This topic is divided into passive and active methods. Example of passive methods is shape from texture, shape from defocus and shape from the silhouette. For active methods, the important categories are shape from shading and photometric stereo. In shape from shading, the cue for shape reconstruction is shading which is the relation between intensity and shape. In this case, only one image is considered. In photometric stereo, where multiple vantage points exist, 3D reconstruction considers multiple images (at least three). Photometric stereo on its own can be categorised depending on existing information of illumination directions, illumination intensities, Lambertian surfaces or non-Lambertian surfaces. This paper presents a method employing deep learning for photometric stereo where lighting and surface conditions are unknown. The proposed method is applied to a public dataset. Based on the experimental results, this method outperforms currently existing techniques.

Shape from X is an interesting area of research in computer vision community. This topic is divided into passive and active methods. Example of passive methods is shape from texture, shape from defocus and shape from the silhouette. For active methods, the important categories are shape from shading and photometric stereo. In shape from shading, the cue for shape reconstruction is shading which is the relation between intensity and shape. In this case, only one image is considered. In photometric stereo, where multiple vantage points exist, 3D reconstruction considers multiple images (at least three). Photometric stereo on its own can be categorised depending on existing information of illumination directions, illumination intensities, Lambertian surfaces or non-Lambertian surfaces. This paper presents a method employing deep learning for photometric stereo where lighting and surface conditions are unknown. The proposed method is applied to a public dataset. Based on the experimental results, this method outperforms currently existing techniques.