An open IoT platform for the management and analysis of energy data

Highlights

• IoT platform for the management of energy data in buildings.

• Includes several inner features to support data analytics in the energy domain.

• Based on the open IoT initiative FIWARE.

• Evaluated in a real pilot with comprising several buildings.

Abstract

Buildings are key players when looking at end-use energy demand. It is for this reason that during the last few years, the Internet of Things (IoT) has been considered as a tool that could bring great opportunities for energy reduction via the accurate monitoring and control of a large variety of energy-related agents in buildings. However, there is a lack of IoT platforms specifically oriented towards the proper processing, management and analysis of such large and diverse data. In this context, we put forward in this paper the IoT Energy Platform (IoTEP) which attempts to provide the first holistic solution for the management of IoT energy data. The platform we show here (that has been based on FIWARE) is suitable to include several functionalities and features that are key when dealing with energy quality insurance and support for data analytics. As part of this work, we have tested the platform IoTEP with a real use case that includes data and information from three buildings totalizing hundreds of sensors. The platform has exceed expectations proving robust, plastic and versatile for the application at hand.

Introduction

Several reports claim that residential and commercial buildings represent around 30%–40% of the overall energy consumption in Europe and in the United States [1], [2]. Because of this, buildings are known to be the largest end-use energy contributor followed by transport and industry, and therefore they are a clear target for potentially reducing global energy consumption substantially.

Despite being great consumers, there is some evidence that shows that public and private buildings have not fully exploited all opportunities available to increase their energy efficiency. On the contrary, they suffer from a rather substantial energy waste that is partly due to inefficient heating, cooling, lighting and other power system (equipment) [3], due to bad use of the systems (behavior) [4] and due to poor fabric efficiency [5]. Although the implementations of measurements to improve the first or the third category can be rather expensive, it has been seen that soft measurements that focus on the change of behavior of buildings’ users are cheap, but yet, can contribute greatly to the reduction of energy use [6].

In order to address the aforementioned inefficiencies due to lack of understanding on how the systems should be operated and other behavioral related aspects in the building sector, one could consider the use of Information and Communication Technologies (ICT) and, more specifically, of the Internet of Things (IoT). This new paradigm that also exists at the domestic level could be used as an instrument to make a realization of the so called Smart Building. In fact, it is foreseen that from 2 to 3 houses out of 10 will be equipped with up to 500 smart devices in the near future  [7].

The installation of smart meters and In Home Energy Displays to make households aware of their energy consumption is not new [8], [9]. The adoption of these devices seems to be an opportunity to exploit them for the reduction of energy use when looking at the available scientific literature (will be detailed later). However, one may also think that the technological effort to deploy such systems may be substantial and become a barrier to achieve this level of technification of the buildings. Nevertheless this technification seems to be happening naturally.

The large amounts of IoT data that will be coming from buildings in the near expected future will have to be analyzed to reveal insights that could help to obtain, expose and understand knowledge from buildings. In turn, this derived knowledge should be able to help to achieve meaningful energy saving strategies and interventions in the targeted buildings [10].

These wealth of information about energy use, offers a great opportunity according to some literature on energy feedback that suggests that intelligent feedback, (that with an extra larger of computation over simple observation) is an effective technique for the reduction of energy demands via behavioral change [11]. Only with a platform capable of making this possible, the implementation of this new paradigm will be successful.

In the IoT ecosystem, several platforms have emerged providing support from the sensorization stage to the stage of management and storage of the data in different forms [12]. In that sense, one of the most large-scale affords is the FIWARE platform, a key initiative of the Future Internet Public–Private Partnership (PPP) to create a well-aligned set of open enablers to receive, process, contextualize and publish IoT data from and for smart cities including from city-wide information to dwelling specific data.¹

Despite all the reasons exposed before, little efforts have been made so far in order to adapt such platforms to building energy management. This energy ecosystem comprises a set of particularities that should be targeted in a specific manner. After analyzing the few examples of studies that have tried to tackle this problem, one can see that it exists a pressing need to apply different data mining techniques in the building energy domain mainly focusing on consumption prediction and pattern discovery or failure tolerance [13]. Thus, IoT energy platforms should include functions for data analysis among their features.

Although giving insighting knowledge behind data is an instrumental aspect of the wealth produced by the IoT, existing platforms are still limited when it comes to integrate data processing and analytic techniques suitable for IoT ecosystems [14]. This is a fundamental limitation of the state of the art as it is key to ensure that the platform will work on the new paradigm of providing tailored, real-time energy feedback to people. This also includes features to support the easy extension of platforms to allocate new data mining techniques comprising common steps in the data mining process. Examples of such features are built-in data-cleaning mechanisms for data pre-processing and storage solutions that would facilitate the execution of online and offline data mining algorithms.

All the aforementioned limitations have motivated us to envision, design, develop and validate what we called the IoT Energy Platform (IoTEP). The key strength of IoTEP is that it is, to our knowledge, the first holistic solution to large scale building energy data management from IoT.

Unlike existing IoT platforms, IoTEP is mainly oriented to support and ease the analysis of large amounts of heterogeneous energy data. A simplified overview of the platform IoTEP is shown in Fig. 1 representing its key features.

To begin with, IoTEP has been designed to easily retrieve either the most up-to-date readings of each sensor within a building, or to retrieve the historic data from such sensors. By means of these two types of access, the platform facilitates the application of both online and offline data analyses over the collected data. As we will see on further sections, this functionality is implemented with two FIWARE storage components, the ORION context broker and COMET. For both enablers, a NGSI-based information model has been defined in order to homogenize all the measured energy-related data.

Secondly, a real-time data cleaning module has been designed as a built-in component of IoTEP. With this, sensor readings are filtered by discarding potential outliers before injecting them in the storage components. This ensures a more efficient use of the resources. For this feature, we have followed a Complex Event Processing (CEP) approach that allows the real-time processing of event streams.

In addition to the above mentioned features, the platform includes also a mechanism to detect volatility changes in the incoming energy data. This mechanism intends to perceive meaningful shifts in such data that might need to re-launch the data-mining services that run within the platform.

Finally, IoTEP features a novel mechanism to automatically identify high-level areas in a building with certain energy-related similarities by means of clustering techniques. The benefit of these virtual areas is twofold. Firstly, they provide alternative representations of the energy status of a building beyond its physical structure; and secondly, they can help in the performance of other data mining analyses by reducing redundancies and defining different granularity levels in the captured sensor data.

Summarizing, the platform presented in this paper intends to be the first stage towards the full adaptation of the IoT paradigm in the retrieval, management and, above all, analysis of energy data in buildings. Considering the need of developing tools that are able to provide personalized real time feedback to change behaviors, and with them, have the potential to reduce energy use, IoTEP is intended to become the stepping stone for the development of such tools.

The paper is structured as it follows: Section 2 provides an overview of the state of the art in this research area. Section 3 looks into the IoT energy platform, including its architecture and its functional modules. Section 4 provides an evaluation of some of the features of the platform; and Section 5 concludes the paper with some final remarks and conclusions.