Fisayo Caleb Sangogboye

OccuRE: An Occupancy REasoning Platform for Occupancy-Driven Applications

Occupant behavior determines a large share of the energy consumption of buildings. Software applications driven by information about occupant behavior provide a mean to optimize this share. However, existing systems for sensing occupancy behavior provide technology-specific APIs statically coupled to the type of computed occupancy information. Software platforms for developing applications for buildings do also not provide abstractions for occupancy behavior. Therefore, technology lock in and lack of proper abstractions wreck the development of occupancy-driven applications. In this paper we present the design, implementation and evaluation of OccuRE, a stream-based Occupancy REasoning platform. OccuRE provides a technology agnostic API for accessing occupancy information to significantly improve portability. The platform uses a component-based computation model with dynamic composition to calculate and reason about occupancy behavior. Together these elements avoid that developers need to deal with technology-specific processing of sensor data to ease application development. Through micro-benchmarks we show that OccuRE successfully and efficiently computes occupancy information for technology-heterogeneous building instrumentations. We use the development of three prototype applications to demonstrate that the API of OccuRE (i) enables several types of occupancy-driven applications, (ii) that the applications -- by using the interface -- achieve portability in regards to occupancy information computation and (iii) that the application code avoids handling sensor data processing.

Demand response in commercial buildings with an Assessable impact on occupant comfort

Electricity grids are facing challenges due to peak consumption and renewable electricity generation. In this context, demand response offers a solution to many of the challenges, by enabling the integration of consumer side flexibility in grid management. Commercial buildings are good candidates for providing flexible demand due to their volume and the stability of their loads. However, existing technologies and strategies for demand response in commercial buildings fail to enable services with an assessable impact on load changes and occupant comfort. In this paper we propose the ADRALOC system for Automated Demand Response with an Assessable impact on Loads and Occupant Comfort. This enhances the quality of demand response services from a grid management perspective, as these become predictable and trustworthy. At the same time building managers and owners can participate without worrying about the comfort of occupants. We present results from a case study in a real office building where we illustrate the advantages of the system (i.e., load sheds of 3kW within comfort limits). Presenting a better system for demand response in commercial buildings is a step towards enabling a higher penetration of intelligent smart grid solutions in commercial buildings.

Improving occupancy presence prediction via multi-label classification

Heating and cooling of commercial buildings accounts for a large proportion of worldwide energy consumption. There exists an opportunity to reduce energy waste by improving the scheduling of heating, ventilation, and air conditioning (HVAC) based on occupancy. However, to enable this potential, we require more accurate methods for predicting occupancy to deliver the required level of comfort when rooms are occupied. This paper examines the novel use of multi-label classification (MLC) for predicting occupancy of rooms based on data from motion sensors. Stating the occupancy prediction problem as an MLC problem enables the use of existing MLC algorithms and provides a solid foundation for evaluating the performance of the predictive models. Our implemented algorithms are benchmarked against an existing occupancy prediction technique (PreHeat) on a dataset from two commercial buildings. The results show that PreHeat and Support Vector Machine (SVM) outperforms other algorithms for rooms with high occupancy frequency. Other machine learning algorithms outperform PreHeat and SVM for rooms with low occupancy frequency. In total, SVM provides a more robust performance than other algorithms with a significantly higher count of highest prediction accuracy for observed scenarios. Our experimental results also highlight that prediction performance for commercial buildings depends more on occupancy frequency than occupancy rate, and the occupancy state before the prediction horizon. By presenting more accurate algorithms for occupancy prediction, we hope to foster the development of more energy-efficient HVAC scheduling systems to reduce overall energy consumption.

Categorization framework and survey of occupancy sensing systems

Abstract A large share of the energy usage in buildings is driven by occupancy behavior. To minimize this usage, it is important to gather accurate information about occupants’ behavior and to improve sensing systems for gathering such information. However, as research on occupancy sensing systems goes beyond basic methods with an increasing diversification, there is a clear need to enable adequate comparison of these systems and their properties. The systems which differ in methods and properties also lack a categorization framework for classifying different options. This article proposes a categorization framework constructed from analyzing and comparing existing sensing systems to address these needs. The classification framework is being constructed from a literature survey of 51 papers and articles presenting 46 different occupancy sensing systems. It is intended that this framework can enable developers to better benchmark and evaluate sensing system, enable organizations to identify trade-offs for adopting sensing systems and aid researchers in scoping out future research in the area.

PAD: protecting anonymity in publishing building related datasets

The diffusion of low-cost sensor network technologies in smart buildings has enabled the collection of massive amounts of data regarding indoor environments, energy use and occupants, which, in turn, creates opportunities for knowledge- and information-based building management. Driven by benefits mutual to occupants, building managers, and research communities, there is a demand for data publication to foster more sophisticated and robust models and algorithms. Data in the original form, however, contains sensitive information about occupants behavioral patterns, and publishing such data will violate individuals privacy. The current practice on publishing building-related datasets relies primarily on policies for dictating which types of data can be published and agreements on the use of published data. This approach alone provides insufficient protection as it does not prevent privacy breaches from occurring in the first place. In this paper, we present PAD, which to our knowledge is the first system that provides a technological solution for publishing building related datasets in a privacy-preserving manner while maintaining high data quality. PAD is able to offer a strong anonymity guarantee by perturbing data records. The unique feature of PAD is that it offers an interface to incorporate dataset users into the loop of data publication and customizes the perturbation such that useful information in the dataset can be better retained. We study the efficacy of PAD using occupancy and plug load data collected in real buildings. The experiments demonstrate that PAD can achieve high resilience to privacy threats without introducing any significant data fidelity penalties.

Poster Abstract: Improving Occupancy Presence Prediction via Multi-Label Classification

An opportunity exits to reduce energy waste by improving the scheduling of heating, ventilation, and air conditioning (HVAC) based on occupancy. However, to enable this potential requires more accurate methods for predicting occupancy. This poster abstract motivates the use of multi-label classification (MLC) for predicting whether or not rooms are occupied based on data from motion sensors. Our implemented algorithms are benchmarked on a dataset from a commercial building in comparison to existing work. By presenting more accurate algorithms for occupancy prediction, we hope to foster the development of more energy-efficient HVAC scheduling systems to reduce overall energy consumption.

PROMT: predicting occupancy presence in multiple resolution with time-shift agnostic classification

Improving methods for predicting occupant presence in commercial buildings is crucial for optimizing energy consumption. Also it is crucial for providing amiable indoor environmental conditions. To enable these improvements, we require a more accurate and flexible framework for predicting occupancy. The promt framework proposed in this paper is an accurate and flexible framework for predicting occupancy presence in multiple resolution with time-shift agnostic classification. promt assumes that no single prediction algorithm, model, or static model parameter can guarantee high fidelity occupancy prediction for varying occupancy requirements and for every kind of rooms. Given this assumption, the promt framework facilitates the deployment of several prediction algorithms and it performs an hyper-parameter optimization procedure on all deployed algorithms to obtain the optimal model for obtaining occupancy prediction in covered room. promt was benchmarked with datasets from two building cases by comparing the F-score of the prediction results obtained from all deployed algorithms. The results document that promt outperforms the performance of any single prediction algorithm by a maximum difference in F-score of 2.3% and a minimum difference in F-score of 0.58%. As a case study we demonstrate the use of promt for scheduling demand response events in a commercial building.

Real-time Occupancy Correction Method for 3D Stereovision Counting Cameras

In this poster, we present an occupancy count correction method - PreCount that corrects the count errors of camera sensing technologies in real-time. PreCount utilizes supervised machine learning approach to learn error patterns from previous corrections alongside some contextual factors that are responsible for the propagation of these errors. In our evaluation, we compare PreCount with state-of-art methods using the normalized root mean squared error metric (NRMSE) with datasets from four building cases. The obtained evaluation results using ground truth data indicates that PreCount can achieve an error reduction of 68% when compared to raw counts and state-of-art methods.

Room-level occupant counts, airflow and CO2 data from an office building

The area of occupant sensing is lacking public datasets to baseline and foster data-driven research. This abstract describes a dataset covering room-level occupant counts, in-room ventilation airflow and CO2 data from an office building. This dataset can among others be used for developing and evaluating data-driven algorithms for occupant sensing and building analytics.

Privacy-preserving building-related data publication using PAD

The massive data collected from buildings provide opportunities for data- and information-based building management. Furthermore, to benefit from collective efforts in research communities, there arises a need for methods to share building-related data in a privacy-preserving manner while being able to ensure the utility of published datasets. In this demo abstract, we present PAD, an open-sourced data publication system that offers k-anonymity guarantee. The novelty of this system is to incorporate data recipients feedbacks into the publication process in order to improve data utility. We demonstrate the interface of PAD and highlight how participants (as data publishers) can generate sanitized datasets using this interface. Also, we demonstrate how participants (as data users) can provide feedback to PAD for improving data quality.