Emerging Technologies and Applications for a Smart and Sustainable World

By Akhil Jabbar Meerja, Mamun Bin Ibne Reaz and Ana Maria Madureira

This reference distills information about emerging technologies and applications for smart city design and sustainable urban planning. Chapters present technology use-cases that have radical novelty and high scalability with a prominent impact on community living standards. These technologies prepare urban and rural dwellings for the transformation to the smart world.

Applications and techniques highlighted in the book use a combination of artificial intelligence and IoT technologies in areas like transportation, energy, healthcare, education, governance, and manufacturing, to name a few.

The book serves as a learning resource for smart city design and sustainable infrastructure planning. Scholars and professionals who are interested in understanding ways for transforming communities into smart communities can also benefit from the cases presented in the book.

• Language: English
• Publisher: Bentham Science Publishers
• Release date: Sep 12, 2022
• ISBN: 9789815036244

Book preview

IOT-based Smart Energy Management in Buildings of Smart Cities

K. Manimala¹, *

¹ Sivanthi Aditanar College of Engineering, Tiruchendur, Tamilnadu, India

Abstract

Buildings consume nearly one-third of global energy and are responsible for one-fourth of CO2 emissions, thereby playing a crucial role in polluting the earth. Cities are more vulnerable as there are more buildings and a huge population due to employment opportunities. Hence, there is a need for the transformation of cities into smart cities with viable environments by making buildings smart. Smart cities with energy-efficient buildings can improve the economy and reduce pollution effects, thereby improving the quality of city life. As human errors and carelessness jeopardise energy conservation and eco-friendly initiatives in traditional buildings, automatic internet of things (IOT) monitored building control, also known as a smart building, is a need of the hour if the world is to advance toward smart cities. The management of the cities should estimate their energy consumption in advance and plan strategies that will help in reducing the energy consumption of both commercial and residential buildings towards creating a pollution-free smart city. The IOT sensors produce continuous streaming data, which necessitates big data analysis to improve the performance of building in terms of energy consumption. Big data analysis based on machine learning techniques is currently being employed for such an automatic analysis and management of buildings based on IOT sensor data. This chapter focuses on bringing out the commercially available IOT sensors for collecting building data, their efficiencies, extracted features and the commonly used machine learning techniques, their strengths, and drawbacks and also identifies the research gap and work to be done for further improving big data analysis of smart energy management.

Keywords: Big data, Gateway, HVAC, IOT, Machine Learning, Neural network, Raspberry pi, Sensors, Smart city, Smart building.

---

* Corresponding author K.Manimala: Sivanthi Aditanar College of Engineerirng, Tiruchendur, Tuticorin Dist, Tamilnadu, India; E-mail: [e-mail]

OVERVIEW

Technology plays a major role in serving the requirements of people and fulfilling the objective of a smart city for ensuring prosperity [1]. The transformation can begin from the buildings where people reside to the commercial buildings, offices, and industries where they work to make a living. The energy usage in such

should be done in a smart way for the concept of a smart city to become a reality. Such smart usage will lead to energy consumption and reduction of pollution effects by reducing gas emissions. The advancements in technology like IOT, machine learning, and big data ensure energy saving by real-time monitoring and controlling of appliances based on the need of occupants. Almost all the countries have started realizing the impacts of global climate change and started initiatives for reducing it. Energy conservation is one such step, and this necessitates the installation of internet of things (IOT) sensors in buildings along with associated machine learning (ML) techniques for analyzing the massive data collected by the sensors using big data analytics. Hence, the buildings need to be incorporated with such sensor monitoring technology, and they must be inspected for compliance checking regularly. It is necessary that the buildings to be constructed in the future should be smart buildings, and the existing buildings need to be transformed into smart buildings to accomplish the dream of the smart city. The smart building (SB), which is an intelligent building that thinks for itself, will provide comfort for occupants as well as energy savings by predicting occupant behaviour in order to operate and control energy equipment only in rooms occupied by them. The machine learning techniques [2, 3] learn from the sensor data that are installed to learn the habits of occupants, and the heating, ventilation and air conditioning devices are tuned accordingly for optimal energy consumption by the controllers based on the knowledge obtained from the trained model. Even though the recent HVAC systems are designed with high efficiency, including artificial intelligence-based voice control, exploring smartness that learns and adapts itself to the environment is still in the budding stage [4]. The smart building concept aims at fulfilling the multi-objective of minimizing energy consumption while maximizing occupants' comfort in a building. This concept can be extended to any building, including residential to commercial and industrial buildings in a city, thereby supporting the concept of the smart city.

The smart building concept starts with the planting of various IOT sensors like CO2 sensor, sound sensor, temperature sensor, humidity sensor, the motion sensor in a building to read building parameters like CO2, sound, temperature, humidity, movements, etc [4]. For such learning, the training data should be formed, which includes discriminant features obtained from sensor data [3]. The features are obtained by applying feature extraction and feature selection techniques on the massive sensor data collected by IOT sensors. The preprocessing steps, namely data cleaning and data integration, should be taken before extracting features to remove irrelevant, noisy data and integrate data from different sources [5]. The prediction of machine learning is used by the controller to adjust the actuators of HVAC devices.

The main component of this smart building module is the forecasting module of a machine learning network that learns sensor data and predicts the behavior of occupants. Only when the room is occupied by people, the HVAC devices are turned ON and the AC is tuned depending on the number of occupants. Heaters and air conditioners need sufficient time to heat or cool the room to a suitable temperature for the occupants, which is dependent on the ambient temperature and the number of occupants. Hence, it is necessary to know the occupancy of the building in advance to tune the devices accordingly, which necessitates learning from past sensor data and the subsequent big data analytics for knowledge discovery [2]. Several machine learning techniques like Artificial Neural Networks, Bayesian Network, Probabilistic neural network, Feed forward network, decision tree and Support Vector Machine are prescribed in the literature for smart building intelligence.

Even though lighting systems form a less portion of energy consumption, several studies have shown significant energy savings when the energy saved in all the buildings in a city due to lighting is taken into account [6]. Fig. (1) [3] shows the framework of a smart building using IOT & Big data Analytics.

Fig. (1))

Framework of SB Model.

This chapter concentrates on the following aspects of smart building:

To summarize the works done so far in the smart building area

To brief the inference obtained from the literature

To discuss the commercial IOT sensors used for collecting building data

To discuss the features and Machine learning techniques used by the researchers

To discuss the processors used for implementing IOT based big data analysis algorithm

To brief edge and fog computing concepts in SB

To discuss the connection of IOT devices and processors

To identify the research gap and suggestions for future researchers

STATE OF THE ART

Buildings are subjected to several changes like rearrangement of spaces for usage by the occupants and due to external factors like temperature, rain, etc. In fact, the weather inside the building changes due to other buildings surrounding it and weather changes [4]. The IOT technology has opened a new paradigm of monitoring the building parameters [5] and communicating the same over the internet for further analysis and action. In addition to that, the energy consumption pattern and room occupation behavior can be learned by the Machine learning technique that operates on sensor data so that they can predict the future movement of occupants. Several researchers have been working on this area to identify features to be extracted from sensor data and proper choice of ML technique. It is proposed to categorize the works into five areas namely lighting & HVAC loads control, luxury load control, detection of occupants and big data analytics.

Lighting & HVAC Control

Lighting contributes around 40% of total energy consumption in a building [6]. Several research works have focused on reducing energy consumption by controlling lighting loads based on the presence of occupants in a building. The implementation of an intelligent lighting system to provide necessary lighting for a building with low energy consumption was proposed by researchers in a previous study [6]. The literature presents another work to control the intensity of lighting based on Q-learning, which has considered the suggestion of occupant’s views in their control algorithm [7]. Few more works utilized signal processing tools for lighting control and to count the number of people occupying a particular room for smart control [8]. Controlling HVAC devices for reducing energy consumption form the major work of several researchers as these devices consume more energy compared to lighting loads. Kalman filter was proposed for fault detection of HVAC devices [9] followed by other signal processing techniques for energy saving [10] and also to improve the comfort level for occupants. Several wifi-based sensor nodes were used for monitoring HVAC data in the work suggested by a few authors [11] and the data were simultaneously uploaded to the cloud server to initiate the ML algorithm for controlling the HVAC actuators. The literature is also rich in works for HVAC control based on occupancy detection using cost-effective embedded processors [12, 13].

Luxury Load Control

Few works of literature report on the control of luxury loads like washing machines, electric ovens, dishwashers, etc., which are now part of life for residential consumers. Signal processing-based methodologies were used for forecasting the usage of these devices based on IOT monitored data and issued necessary control signals for their proper usage and energy saving [14]. Recurrent neural networks [15] and deep learning [16] methodologies were also suggested for the control of these devices.

Detection of Occupants

Several sensors like infrared (IR) sensors, motion sensors, and camera sensors are proposed for identifying the number of people occupying a particular room of a building in association with several machine learning rules. Algorithms using the OpenCV library are used for analyzing camera images to identify the headcount for occupant detection [17]. The application of the camera sensor is limited as it is intrusive in nature and affects the privacy of occupants in addition to requiring more storage space. Some applications have used the sound sensor to identify the occupancy of a room [18]. The sound sensor is used to cover a wide range compared to the motion sensor. The CO2 sensor is one more option for measuring the number of occupants. The level of carbon dioxide is used in such sensors for the head count. Tyndall et al. [19] have proposed a thermal imager for detecting the presence of occupants while Raykov et al. [20] used a passive IR sensor for the same applying behavior extraction. The sensors should be located at various locations in room for capturing accurate movements within the building. The signal should be sampled at an appropriate sampling rate for proper monitoring of events and 100 ms is normally chosen as the sampling rate of the sensor [21].

Big Data Analytics

The IOT sensors equipped in buildings not only support energy saving but also the identification of faulty equipment. Thus, they can also be named predictive maintenance systems and several works suggest the method for detecting faults in HVAC systems based on the sensor data [9]. Few works listed out the integration of signal processing step along with IOT sensor data for calculating energy usage of the building.

The adaptability of buildings to occupants' comfort is possible with the collection of building parameters using various IOT sensors, energy usage patterns, and human movements and feeding the data to the intelligent system to predict the future energy usage and human behavior based on past collected data. The data is an integration of human activities and other room conditions. The smart building energy management system makes suitable decisions in operating actuators like switches and controllers to modify the conditions of the room and provide comfort to the occupants and reduce energy consumption [21]. For this purpose, big data analytics is needed and this software is responsible for prediction, model building, and decision making for reducing energy consumption by examining massive IOT sensor data. Such a complex process involves designing efficient algorithms based on machine learning techniques for identifying the relations that exist in sensor data for model creation to predict human activities and control the actuators for energy saving [21]. Fig. (2) (see also [2]) shows the big data analytics process of Smart Building.

Fig. (2))

Big data Analytics of Smart Building.

IOT SENSORS

The smart building concept begins with data sensing using IOT sensors. There are various sensors available in the market at a cheap price. Table 1 represents the list of commercial sensors.

Table 1 Commercial Sensors.

The sensor units can be standalone and directly connected to the embedded microcontroller or integrated with a wi-fi module for sending the data wirelessly to the embedded microcontroller or the gateway or cloud server. It is necessary to ensure that the data collected by the sensors are properly uploaded to the processing unit. Fig. (3) shows sensors used for building data collection.

Fig. (3))

Sensors for building data collection.

DATA PRE-PROCESSING

The size of the data collected from the sensor is very significant for a Machine Learning algorithm to provide an accurate prediction. If the size is fewer means, the accuracy will be less and if the size is more, the computational complexity will increase [22, 23]. Hence, the size of the data set should be properly chosen so that the ML algorithm predicts correctly with less computation. Mostly 1 week to 1 year data was considered by the earlier researchers [24]. Data pre-processing encompasses data cleaning, data integration, data transformation and data reduction. Data cleaning removes irrelevant and noisy data [25] while data integration integrates all sensor data collected from various sensors. Data transformation transforms data from one domain to another domain suitable for ML processing. Data reduction is necessary to remove redundant and outlier data so that the accuracy of ML prediction can be improved. Principal component analysis [26] and Linear discriminant analysis are commonly used for data reduction.

FEATURES

After data pre-processing the significant features need to be identified to give as input to the ML algorithm. Researchers have suggested different feature sets and analyzed the performance of ML-based Big data analytics using the feature set used by them. Hence, there is so far no commonly agreed feature set for designing a smart building training model. The list of features proposed in the literature [27, 28] includes the total time of data collection, Room occupying time, Sound, CO2, Wide and narrow Field motion, Average occupants, Temperature, Humidity, Light intensity, Number of the door opening and closing, Solar intensity, Outdoor temperature and humidity, humidity, Previous Indoor Temperature, Air conditioner (AC) status, Heater status, Air conditioner Temperature, Air conditioner humidity etc. Some of the works have included feature selection techniques using optimization algorithms for selecting only the pertinent features from the feature set formed for improving the accuracy of Machine learning prediction.

IOT PROCESSORS

Raspberry Pi 4

Raspberry pi is a blessing for big data analytics in the form of a small card-sized computer. Real time data processing using IOT sensors could easily be performed on this processor as it is possible to build a cheaper portable cloud server on it [29]. This processor has the ability to process sensor data at a fast rate. This processor enables the processing of IOT sensor data obtained in remote locations which are deprived of any internet connectivity to access cloud computing. This processor provides efficient data processing similar to cloud computing in a portable mode. The features are it is a Cortex A72, 1.5 GHz and 64 bit processor with 2GB SDRAM. It has Bluetooth and wireless provisions with USB ports and Gigabit Ethernet. It has both a display port and Camera port along with stereo audio and composite video port. It has provision for loading operating system through SD card slot. It has two supply points of 5 volts from USB-C and GPIO header.

IOT 2020 Industrial Gateway

Siemens gateway is based on intel Quark Chip Simetic IoT 2020. It is designed for industrial IOT Solution which is an open platform for collecting both analog and digital data, processing based on machine learning algorithms and uploading data to the Cloud server. It also has provisions for receiving data from cloud servers The specifications are: it is based on intel