This paper presents the design of a Sensor Node for sensing environmental variables in Buildings with the objective to control HVAC (Heating, Ventilation and Air Conditioning) system using a BEMS (Building Energy Management Systems) to achieve energy savings. BEMS is a control system used to manage the energy consumption of a Building trying to achieve energy savings. The designed Sensor Node is capable of measure six different environmental variables: temperature, humidity, and carbon dioxide, using a Sensirion SCD30 sensor; particulate matter 2.5, and 10, using a Sensirion SPS30 sensor, and lighting using a BH1750FVI sensor. The control unit for the Sensor Node is designed using a node MCU ESP8266 microprocessor which is capable to receive the sensors signals. The microprocessor are connected to a XBEE module for communications purpose with a main control system. The main control system is developed in a Raspberry Pi. The objective of this design is control the set point of the HVAC equipment using the values of the sensed environmental variables. The main contribution of this work is to control wirelessly the operation of the HVAC equipment through the set point for a determinate space using the environmental sensor node measures obtained of the space to be controlled.
List of Conference Committee, Steering Committee, Conference General Chairs, Program Chairs, Technical Committee are available in this pdf.
Modern machine learning models consume massive amounts of energy. In a widely cited paper [9], the authors compare the estimated CO2 emissions from training common NLP models like BERT [5], GPT-2 [1], ELMO [8] and transformers [10]. Similarly, in [7], the authors compare other models like Meena [2] and GPT3 [4]. For example, GPT-3 training consumes around 550 metric tons of carbon. Together, data center use including machine learning model creation is projected to grow exponentially in the coming years [6]. As machine learning becomes used by more and more organizations for their business processes, it is imperative new paradigms for efficient model training and inferencing are developed. Our research project is motivated from a recent flight search which showed multiple flight options along with their estimated carbon emissions (Figure 1) . It is left upon the discretion of the traveller to pick a suitable flight based on CO2 emissions, comfort and convenience. We extrapolate the same for training a batch of ML jobs and create a scheduler which can appropriately allocate jobs to different datacenters at different times. In addition, based on the data characteristics and ML task (classification, regression etc.), we can also recommend the appropriate model to be used.
Soil nutrient mapping has been implemented in agriculture for the last 40 years. Ongoing economic pressures in agriculture to increase crop yields while sustaining farmer profitability demands in-depth knowledge of soil nutrients. Spatial and temporal variability of yield-limiting factors have been recognized for a long time, and, with such information, farmers have technologies to manage their fields site-specific. However, farmers still tend to manage their fields uniformly because it takes cost, labor, and money to assess and monitor soil health conditions by traditional lab-based methods. We have used a rapidly emerging hyperspectral sensing technology to assess plant-available nutrient content in soil, especially nitrogen, for the use of optimizing in-season nitrogen management. The new and emerging technology would empower farmers with data to help them make better decisions to grow more profitable crops, protect the environment, while growing more nutritious food
The United States Department of Energy estimates that global energy consumption will double in the next two decades which will lead to extensive utilization of fossil fuels resulting in economic and environmental challenges. Although biofuels are a great alternative, the current generations cannot sustain the demand without threatening food supplies. Third-generation biofuels have lipids that can be converted into biodiesel, jet fuel, and gasoline. Despite these advantages, they are not produced commercially as the lipid yield of the natural strains are roadblocks in their large-scale use.This research aimed to increase TAG production in the microalgae Isochrysis galbana by inoculating it with Bacillus megaterium for more efficient biofuel yield. The study uses a bioreactor created using PVC pipes, 8-port sprinkler, CO2 tank and bottles to grow the microalgae. Secchi stick was used to measure optical density before extracting dried biomass with a french press. After conducting an ANOVA test to compare the samples, the p-value was lesser than 0.01. Later the post-hoc test was run and the groups were found to be significantly different especially for the 4x concentration versus 1x concentration. Results showed that through the co-cultivation of this algae and bacteria, TAG yields are increased. To support the findings of the experiment, bulk RNA sequencing data for Isochrysis galbana and Bacillus megaterium was compared to identify the pathways that were affected in the microalgae. The increase in biofuel productivity is crucial for the bioremediation of wastewater, the bioremoval of CO2 and for improving the economics.
The mass generation of green energy through the pathway of Renewable Energy Sources (RES), is one of the major decisive components contributing to the concept of decarbonization. The energy sector assessment reveals that out of per kWh energy produced, renewable energy emits between 9 and 1000 gCO2 equivalent on a life-cycle basis. Contemplating the literature and the latest reports released by world-wide energy department(s), this study presents a comprehensive concept of carbon footprint, its causes and effects including major impact on biodiversity and human life. The study discusses the substantial steps taken by developing countries like India in leveraging RES for the production of energy and its progressive impact on ecology. The discussion would be inclusive of major change and transformation which has brought significant drop of 16% in the greenhouse gas (GHG) emissions by installation of 104.9 GW of RES excluding Hydro. The appropriate steps taken forward therefore align with the Sustainable Development Goals (SDGs) of the United Nation (UN) in achieving net zero emission (NZE).
Many of us have defective electrical devices in storerooms or drawers. Recycling electrical and electronic waste (E-waste) allows for reusing the original materials rather than mining new ones, especially if they are considered Rare Earth Elements. It is critical to separate and sort E-waste before recycling, which can be economically viable for products containing valuable metals and materials. Identification of the device and all related information will greatly support pre-dismantling. This work proposes a pipeline and automated system to identify the device using several techniques: artificial intelligence for device classification into device type, brand and model, multi-sensors for identifying internal components like Infrared or X-ray, a web crawler for creating image datasets using keywords from search engines, and a web scraper for retrieving device specifications, like dimensions, battery type, visual data, available colours, etc.
Unmanned aerial vehicle (UAV) applications can be powerful tools in horticultural research. However, they have not yet been widely explored in the literature. One significant area of interest for the horticultural community is identifying invasive plant species that, if left unchecked, can hinder the growth and health of native plant species. To address this issue, we assembled a novel data set of invasive and native plant species for seven southern states in the United States and developed a plant classification technique using pre-trained convolution neural networks and transfer learning. We explored extracting features from our data set using several state-of-the-art deep convolution neural network models, including InceptionV 3, MobileNetV 2, ResNetV 2, VGG16, and Xception. We then used the extracted features to classify the plant species using a convolutional neural network with cross-validation. Our experiments demonstrated the potential of our proposed method for achieving performance with a 94% accuracy using the MobileNetV2-DCNN model with data augmentation, hyper-parameter optimization, and the softmax classification technique. The advantage of our approach is the learning process is automated and highly accurate. The data set to train the plant species classifier will be available on request.
In a study published by French power management [1], a zero-carbon regime will be achieved by reducing energy consumption by 40% and multiplying the energy production from renewable sources by 4. Increased integration of renewable sources will lead to an increased vulnerability of the power grids thereby leading to increased occurrences of local and global blackouts which can have disastrous social and economic consequence [7]. The dynamic nature of wind and solar energy means that the grid can either be under-utilized or over-utilized, the later can lead to outages and thermal damage of grid components. In order to stabilize the grid, there has been a push to controllable loads such as battery storage [8], vehicleto-grid [8] and hydrogen production [9]. In this paper, we propose the use of ML training jobs as a controllable load to stabilize grid. There has been prior work in using compute-centers as controllable load for efficient grid operations [10], [4] where the demand response requirements are met by taking advantage of variable power consumption in compute-centers. However, our main idea focuses on using a specific type of compute, machine learning (ML), to act as controllable loads. As illustrated in Figure 1, there has been a rapid growth in the amount of compute required by training tasks of machine learning. For example, the demand for ML training has increase by 150% per year at Facebook [6]. We propose that the increasing popularity as well as specific properties of ML, especially deep learning training, makes it an ideal candidate as controllable loads.
The vehicle routing problem (VRP) is a combinatorial optimization problem that involves finding optimal routes traveled by a fleet of vehicles to serve a set of customers. Ever since it was introduced, the literature on VRP has expanded exponentially and it has become quite disjointed and disparate. Given the huge number of the VRP's field of application and disciplined, examining the VRP literature has become hard. The aim of this work is to trace the routes of the evolution of VRP variants over the years. We aim to present a systematic literature review based on 285 papers, using a bibliometric analysis, and a classification framework. This work would enhance a systematic identification of gaps in the literature and as a result, lead to future research agenda and highlight scopes of improvements in several VRP areas.
Ensuring healthy lifespans and promoting equity in health outcomes is a critical component of the United Nations Sustainable Development Goals, and progress has largely been measured using national life expectancy averages. While national averages reveal broad trends over time, they also largely obscure differences and inequalities at the subnational level. In studies of sustainability and public health, divergence refers to the increasing difference between two values over time. A major topic of interest is how subnational units, such as counties, deviate from the national average over time. In this work, we examined how three factors critical to evaluating the health of communities and quality of life, namely life expectancy, education, and income, have deviated in the United States at the county level in comparison to the national average. We found that while deviations in income and education have remained largely constant, the life expectancies of some counties have increased significantly over the national average, while others have stalled or even decreased in comparison. There has been a systematic divergence in life expectancies in US counties over a 34-year period leading to increasing disparities between counties. Gains in life expectancy were primarily concentrated in metropolitan areas while rural counties lagged, increasing urban-rural disparities over time. Identifying the upward or downward trends in life expectancy at the county level and understanding the root causes and risk factors driving those trends is critical to formulating customized Precision Public Health interventions and policies to improve health outcomes and create sustainable communities.
Network communication is crucial in the Energy Grid of Things (EGoT). Without a network connection, the energy grid becomes just a power grid where the energy resources are available to the customer uni-directionally. A mechanism to analyze and optimize the energy usage of the grid can only happen through a medium, a communications network, that enables information exchange between the grid participants and the service provider. Security implementers of EGoT network communication take extraordinary measures to ensure the safety of the energy grid, a critical infrastructure, as well as the safety and privacy of the grid participants. With the dynamic nature of network communication of the EGoT, the information provided by the customer or the service provider can be falsified by a malicious attacker. Therefore, a trust model is necessary to monitor any abnormal activities. This paper describes a distributed trust model system that meets the need of the EGoT. This paper describes methods for evaluating and improving the distributed trust model using standard hypothesis testing metrics such as true positive, false positive, true negative, false negative, equal error rate, and F1 score. Example calculations are shown based on generated sample data.
"Employee turnover intention" is linked to significant direct and indirect costs for the organization. Despite the increased interest in ethics-related topics, research has not comprehensively studied turnover intention from an ethical perspective. The purpose of this paper is to outline research conducted on the impact of ethical leadership on employee intention to leave the organization. By conducting a systematic literature review, 36 articles which study the relationship between ethical leadership and turnover intention were selected and analyzed. Findings mostly confirm the negative indirect relationship between ethical leadership and turnover intentions. Data are presented based on the coverage of the articles, the methods used, the geographical scope and the unit of analysis. The study focuses on commonly found mediators (such as ethical climate, organizational commitment, supervisor trust, job satisfaction, work meaningfulness, job-related stress) and moderators (such as stress, core self-evaluation, perceived organizational support, culture), and any potential negative impacts of ethical leadership on work outcomes. Future research suggestions are also listed. Changes brought about by COVID-19 have rendered an honest, fair and transparent leader more necessary than ever. The review extends our understanding of how ethical leadership influences turnover intentions and provides insight into potential future challenges which need to be addressed more thoroughly. The distinct contribution of this paper lies in the examination of the role of the ethical leadership style in mitigating negative employee outcomes.
Dry and semi-arid regions of the world witness a decrease in freshwater supply due to urbanization and population growth. Effective water utilization and management is the nexus at the consumer end for overcoming water scarcity. Limited research has yet concentrated on the end-user water study and its ideal solution, suggesting the necessity for research encompassing the end user s viewpoint for effective water management. The proposed study adopts a novel approach to determine the optimum daily per capita demand by implementing the Genetic algorithm (GA), an Artificial Intelligence (AI) tool. The multi-linear regression (MLR) model is applied to data from the socioeconomic household survey to perform the optimization. A household survey in the Jaipur Municipal Corporation region was conducted as a case study to validate the proposed study. Instead of the 133.66 LPCD specified by the regulating body, the executed optimization yields values from 106.62 to 131.52. The proposed methodology provides an alternate user perspective to minimize water wastage, which is further implemented with additional constraints.
In recent years, the introduction of electric vehicles (EVs) has been progressing worldwide in order to achieve carbon neutrality by 2050. EVs are a distributed energy resource, such as photovoltaics (PVs) and household battery energy storage systems, attracting attention as a means of ensuring power quality, enhancing resilience, and achieving energy independence. Alternatively, in the power distribution system, power flow is becoming more complicated due to the penetration of PVs and EVs. Thus, it requires reviewing the configuration of power network facilities and changing their operation methods. Therefore, in this study, the power flow is smoothed by optimizing the charge/discharge schedule of the transit electric buses.While many studies currently focus on private EVs, the study focuses on transit electric buses with a high certainty in operation. It constructs a bus model faithful to the actual operation. The reverse power flow (RPF) and normal power flow (NPF) were smoothed by charging/discharging using the proposed method, and the peak cut rate was 67.1% for RPF and 18.1% for NPF. Additionally, the proposed method utilizes the available battery capacity more effectively than the comparison methods and contributes to NPF and RPF smoothing.
Remote terminal unit (RTU) is the core device of the monitor system in a utility company. A dedicated team of engineers and technicians need to work together to ensure their smooth operation with minimal interruptions. A predictive model can provide a list of RTUs with highest likelihood to fault and will help to prioritize the efforts of maintenance and repair. This paper introduces multiple machine learning and deep learning methods to produce such results based on historical RTU activity data. It shares experience in such practices as experimental design, sampling, feature generation, feature selection and model evaluation.
Class 2 transformers are small line-frequency transformers that are widely used for control systems that require 24 VAC signaling, including residential and commercial HVAC systems, industrial control systems, doorbells, and much more. In this work, we sampled and tested seven Class 2 transformers, each across different operating conditions, in order to characterize their efficiencies and note their shortcomings. We also provide possible improvements and solutions. We see on average a peak efficiency of 84.43% with 5.37 W of power loss when operated at 75% (30 VA output power) of their rated power, a 1.84% efficiency drop from the temperature rise that occurs at steady state when operated with full load, 2.8 W of no-load loss at 120 VAC input, and a no-load loss contribution of over 50% when operating at less than 75% load power. With these values, there is a clear goal to strive for in order to improve or create an alternative to these Class 2 transformers.
This study reviews how IoT as an emerging and disruptive technology may be utilized to address the challenges of sustainable development in transportation system in developing countries. We conducted a qualitative sequential exploratory mixed method to take a deep dive into few leading smart transportation startups in Iran. We offer insights on how these firms acquired the know-how, how they engaged in the process of innovation to develop solutions that address unique challenges of developing countries. We framed the analysis on adoption of four technologies, Internet of Things, Artificial Intelligence, Cloud Computing, and Digital Twin for sustainable transportation.
This paper focuses on short-term load forecasting for the day ahead using an Ensemble learning-based Random Forest method. The study uses real-time hourly load data and meteorological data from Bengaluru city, Karnataka, India, to predict the load. The inputs considered for the load forecasting are load profile data, dry bulb temperature, dew point temperature, and humidity data for 31 days from January 1, 2021, to January 31, 2021. The results obtained from the Random Forest model are compared with those obtained from the Ensemble learning-based Bootstrap Aggregation model to evaluate the effectiveness of the proposed method. The study uses statistical parameters such as Maximum Absolute Percent Error (MAPE), Maximum Absolute Error (MAE), and Root Mean Square Error (RMSE) to analyze the predicted load. The findings indicate that the proposed Random Forest model yields better results, with a Mean Absolute Percentage Error (MAPE) of 2.75% compared to the other Ensemble learning-based Bootstrap Aggregation method.
Accurately assessing and measuring population vulnerability across geographic scales is critical for health risk management of sustainable communities. Currently, the most widely used metric for vulnerability analysis in the United States is the Centers for Disease Control and Prevention's Social Vulnerability Index (SVI), a metric which has proven its utility in a number of circumstances. However, at times, its structure limits its use in international contexts and longitudinal studies. We propose the use of the Community Human Development Index (CHDI), a scalable and predictive tool, to address these gaps. Based on the setting, the CHDI can work either as a complementary metric alongside the SVI or as an alternative vulnerability metric. In our comprehensive study of over 70,000 census tracts in the United States, we found that CHDI correlates well with SVI at the neighborhood scale, while also capturing uniquely different information about the vulnerability of populations. We also found that the CHDI predicts community health outcomes as well or better than the SVI, and can be used as a proxy to measure changes in overall health risks to communities. As a Precision Public Health metric for population health vulnerability analysis and as a performance indicator of community health, CHDI can be used to provide an analytical basis for policy decision making. CHDI can be utilized to promote health equity across a broad range of communities through more accurate health risk assessments leading to strategic, targeted, and timely interventions for better health outcomes.
