Project

Vehicle Electrification

Goal: Projects the Smart-X {Cities, Buildings, Grids} group did in vehicle electrification

Updates
0 new
0
Recommendations
0 new
0
Followers
0 new
3
Reads
0 new
52

Project log

Roger N. Anderson
added a research item
With the increasing demand for oil and its decreasing supply, the cost of fossil fuels has increased considerably over past decades and the trend is expected to continue. Furthermore, the increasing awareness about the environmental impact of burning fossil fuels has generated a lot of interest in the search for cleaner and more sustainable sources of energy. The transportation sector is dependent on fossil fuels for over 90% of its energy requirements. Using electric vehicles (EVs) to replace large fleets of vehicles, for example those used for public transportation, delivery of goods, mail, and cargo, car rental, etc. in large urban areas, is being looked at as an easy way to reduce dependence on fossil fuels for transportation.
Roger N. Anderson
added a research item
Roger N. Anderson
added 5 research items
Agenda: Artie • Historic overview on what has been done in ConEd • Columbia's work predicts future failures • People would be using the tool with lots of metrics and dashboard • Dollar/Delta MTBT • Revenue/feeder • GUCCI story, light designer, lightening, 5 watts/sq ft, increase to 25 watts/sq ft, but the revenue increases much more/sq ft • Do the work starting from Phase 0 • Electric distribution • R&D managing power plant, the most difficult thing is something people do not know about • The No. 2 is the known things that may come back to have a problem • GE is still not in an integrated fashion • Luke Air Force Base uses solar to generate electricity and store them • DOD is sustainable • GE embed itself into the solutions for customers Roger • Introduction on the work we have done • Mission of scale and sustainability • We want to see the synergy • CCLS Presentation o Terrawatts not keeping up with demand o Peak oil o Curse of dimensionality o "Plenty of room at the bottom" Richard Feynman o Adaptive Stochastic Control Systems o Load & Source Tuning o A123 battery with 70% usage left o Lean management o AO and performance cyclicity
2. The GE ecomagination Challenge (the “Event”) is sponsored by General Electric Company (“GE”). Multiple entries are permitted. Each entry will be reviewed independently. Multiple individuals or entities may collaborate to submit a single entry. Participation is subject to all federal, state and local laws and regulations. Void where prohibited or restricted by law. You are responsible for checking applicable laws and regulations in your jurisdiction before participating in the Event to make sure that your participation is legal.
GE, FedEx Express, Consolidated Edison and Columbia University demonstrate the nation’s first electric delivery vehicle depot implementing smart charging technology connected to the distribution grid in an intelligent manner.
Roger N. Anderson
added a research item
Build a Columbia Machine Learning System for intelligent Load and Source Management of Electric Delivery Vehicle recharging facilities. We will build a local intelligence system at the EDV depot that is also linked to the utility electric distribution control center and the onsite energy management system.
Roger N. Anderson
added a research item
This paper will outline the results from a preliminary investigation into the power consumption and load profiles of a delivery vehicle depot operated by FedEx and located at 120 Leroy St, New York City. This study originated from the perceived need to control electric vehicle charging behaviors within the constraints of the already installed electric power distribution system inside the building. The data that is presented and interpreted within the EV charging context was captured by GE's Energy Services during a load study on the building's existing load panels and main feed. The load study was conducted to verify that the existing electrical distribution equipment was adequate to supply the planned electrical vehicle supply equipment (EVSE) installation.
Roger N. Anderson
added 3 research items
CCLS at Columbia won a global competition for Research Funding for new, innovative AI and ML supporting GE product development. In our case it is ML for recommending charging times for FEDEX trucks in lower Manhattan that will not disrupt the delivery schedules for Fedex packages 24 hr/365 days of the year.
Our Martingale Stochastic Controller for Control Center Operations embodiment's purpose is for remote sub-sea decisions in a real time control e-Field setting affecting the form and timing of gas, oil, and water production in the ultra deepwater based on reinforcement learning. The policies this Stochastic Controller generates are termed in finance theory “martingales.” That is, the decisions are always in-the-money with respect to both profitability and engineering efficiency. The controller achieves lower level control objectives at the same time that it is maximizes the expected value of discounted cash flows under reservoir, technical, and market uncertainties. The Martingale Stochastic Controller makes it possible to implement Real Options as well as regulatory, supervisory, and scheduling control, operational and capacity planning functions necessary to conduct remote technical and business operations. A key element to a reinforcement learning based controller is the ability to simulate the processes being controlled.
Fedex (David Herford) provided daily package volume to better train our Machine Learning model Need Online feeds in near real time EDV Charging load time series Transformer load time series Con Ed meters can be gotten real time from Raw pulse data via BMS-like SCADA system Weather forecast and observed feeds
Roger N. Anderson
added a research item
improve the Building Management Systems (BMS). One key aspect of such is the efficient utilization of the electricity, which entails developing sophisticated and accurate models for electric load forecasting. Load forecasting helps the building management make important decisions such as scheduling of large power consuming machines such as HVAC, the addition of capabilities for Electric Vehicle charging, and planning for new infrastructure development. Also, this method of lowering our carbon footprint, by efficient scheduling of processes, is particularly worthwhile because it actually reduces energy costs to the building, unlike many environmental initiatives that require large monetary investments. The building in our study is the Federal Express World Center at 130 Leroy Street. The building has multiple sorting facilities with huge power-drawing conveyor belts and exhaust fans. As such, its power consumption patterns are very different from a normal office building where HVAC is the dominant load. In the FedEx facility, other exogenous factors such as package volume play a crucial role.
Roger N. Anderson
added a research item
Fedex corp = fedex express (80,000 vehicles and 600 planes) + fedex ground + services (Fedex office = old Kinkos) + freight (old Watkins heavy trucks) Hubs Memphis, Paris, Guangau 1992 = cuppertino calif = lead acid 1997 = hybrids (EDF + Eden corp 2005 = EV planning with Li Ion 2007 = London, then Paris, then LA, and now NY
Roger N. Anderson
added a research item
The world is struggling to find methods to reduce green house gas emissions, decrease dependence on hydrocarbon energy sources, enhance aging urban infrastructures, and improve living conditions. Electrifying the transportation sector has been heralded as a significant part of the overall solution. In dense urban areas such as New York City, delivery vehicles will hopefully constitute a significant portion of the electric transportation future. The technology that will decide if this happens quickly will not be the Electric Delivery Vehicle (EDV) itself. Building the truck is the easy part since they do not carry large weights over long distances. It will be the reliable recharging infrastructure to assure that no matter the state of the electric grid, the deliveries will be made on time. And that will require a Smart Grid controller, with coordination capabilities between the utility and the fleet depot. In particular, the switch to electrified charging depots for these large fleets of delivery vehicles will cause localized stresses on already heavily loaded urban distribution grids. In 2011, Columbia University's Center for Computational Learning Systems will be implementing a Machine Learning System MLS controller for delivery fleet depots that will provide optimal recharging policies and actions to account for changing weather conditions and daily grid uncertainties in order to start the EDV transformation in New York City. The Columbia MLS controller will recommend recharging strategies to a large number of advanced recharge stations in the depot of one of the major package distributors in Manhattan. Also, storage batteries will be part of the recharging facilities to assure delivery reliability, and battery systems from the EDV's will be reused within the Universal Power System of the rechargers after their vehicular usefulness is over. These recharge stations will receive commands such as when to optimally start and stop charging both the EDV and the battery systems, while recording and then transmitting energy and other usage information to the Manhattan Electric Control Center of Con Edison. The utility can send signals to the fleet MLS controller to alter the current or future load profiles to ensure proper grid operations while also successfully meeting the vehicle fleet's energy demand constraints. Most importantly, the fleet MLS controller will be able to respond to electric load management directives to decrease or increase the current draw from the on-board vehicle inverter. The Columbia MLS controller will be connected not only to truck recharge docks, but also to an advanced supervisory control and data acquisition (SCADA) system from a major supplier, who will be capable of scaling the integrated recharge system to cities worldwide once we work out the logistics of keeping these EDV's on the road 24/7. This smart system will serve as the charging depot's central command center for vehicle recharging. The Columbia MLS controller uses sophisticated machine learning algorithms to set the tuning parameters contained in the SCADA's control algorithms. The MLS controller is being developed in partnership with the Castle Laboratory of Princeton University.
Roger N. Anderson
added a research item
How the General Electric that was once America’s biggest, the maker of power turbines, jet engines, MRI and other medical devices, the seller of insurance, became a shadow of its former self by 2018.
Roger N. Anderson
added 2 research items
A new generation of lithium-ion batteries, coupled with rising oil prices and the need to address climate change, has sparked a global race to electrify transportation.
Roger N. Anderson
added 2 research items
'Radical new charging technology' sparks UPS switch to all electric fleet in London. Smart Grid required to mix solar, battery, wind and power plant electricity, all managed locally at each UPS Depot in London.
Roger N. Anderson
added a research item
Roger N. Anderson
added a research item
How to plan electric Vehicle recharging at 24/7 Fedex Depot in Lower Manhattan, NYC.
Roger N. Anderson
added a research item
MY VITA as of June 1, 2020, with newest Patent Numbers issued for 5 new Continuation Patents.
Roger N. Anderson
added a research item
Update of Columbia Machine Learning System for Electric Load Prediction for EDV Charging at Leroy Street Depot in NYC
Roger N. Anderson
added a research item
By using initial energy usage data from a given manufacturing facility, as well as seasonal patterns and weather predictions, the machine learning forecasting system can predict the facility's future electricity consumption. The system ties into electric vehicle (EV) recharging infrastructure associated with the facility, optimizing EV recharging loads and schedules based on usage and weather information. By comparing actual usage with usage predictions, the forecasting system continues to learn from the manufacturing facility it monitors, becoming more accurate and saving additional money and energy the longer the forecasting system is used. A continuously learning energy forecasting system is scalable and saves costs over time The feedback loop built into the machine learning forecasting system scores the accuracy of its own predictions, minimizing errors and inefficiencies over time. The system thus gets better at helping manufacturing facilities and companies avoid peak consumption penalties. It allows companies to more efficiently use their capital assets, extending the lifetime of those assets and reducing expedatures. The system is scalable to hundreds of EVs associated with the manufacturing facility, and has been demonstrated in a package delivery facility that processes tens of thousands of packages per day.
Roger N. Anderson
added 5 research items
Federal Express 2 , GE Global Research 3 ! Introduction With the growing energy demand in our buildings, there has been a continuous need to improve the Building Management Systems (BMS). One key aspect of such is the efficient utilization of the electricity, which entails developing sophisticated and accurate models for electric load forecasting. Load forecasting helps the building management make important decisions such as scheduling of large power consuming machines such as HVAC, the addition of capabilities for Electric Vehicle charging, and planning for new infrastructure development. Also, this method of lowering our carbon footprint, by efficient scheduling of processes, is particularly worthwhile because it actually reduces energy costs to the building, unlike many environmental initiatives that require large monetary investments. The building in our study is the Federal Express World Center at 130 Leroy Street. The building has multiple sorting facilities with huge power-drawing conveyor belts and exhaust fans. As such, its power consumption patterns are very different from a normal office building where HVAC is the dominant load In the FedEx facility, other exogenous factors such as package volume play a crucial role. However, the actual and forecast real-timed data was not available till recently so we used proxies for it. Data Electric load data for the building has been available for the last 5 months sampled at a frequency of every 15 minutes. Additionally, a calendar of holidays, observed weather data (temperature and dew point-source: Central Park NOAA observation data!via the Weather Underground) and a day-ahead weather forecasts (source: NOAAs National Digital Forecast Database via the Weather Underground) were fed into several different Machine Learning (ML) models to predict day-ahead electric load for the building. Mean Average Percentage Error (MAPE) was used to measure the accuracy of our predictions and select a best performing ML model ML Models There are various ML models which can be deployed to forecast the load profile such as regression, time series, neural network, tree classification and Support Vector Regression (SVR). We tested each of these and selected a SVR model after back-testing each model with the actual data. Our best-performing SVR model uses 8 covariates. Since there is a cyclical component in the load profile covariates such as previous day load, previous week load, previous day average, previous week average, time-of-the-day and day-of-the-week were incorporated. Additionally to account for the HVAC, an index called Humidex (composite of temperature and dew point) was included as a covariate. In order to model package volume, we had to add a covariate with discrete sets of values for different kinds of holidays. In regard to the choice of kernels, we used Radial Basis Function in our SVR model to project the data into a higher dimensional feature space, which improved results substantially. A challenge that remained was to capture the unpredictable spikes caused by the operation of the large conveyor belts and exhaust fans (time and duration of occurrence and magnitude) in the load profile during the busy package loading and unloading hours, when the electric load goes up by more than 100 percent.
Roger N. Anderson
added 2 research items
Manufacturing facilities of all kinds have more complex electrical load profiles than office space or homes. The latter is generally sinusoidal with highest load in the day and lowest load at night. Manufacturing facilities are instead controlled by the work cycle, which often involves significant nighttime loads. All of these buildings are migrating to more and more of an electrical economy, in which electric vehicles are delivering their produce and powering the cars of their employees. We have invented a Machine Learning (ML) System that learns the patterns of these complex systems, as well as the simpler systems, and by adding predictions of upcoming weather and seasonal work pattern changes, forecasts building loads for both the manufacturing facilities and their EV recharge facilities so that optimal energy usage results. Peak consumption penalties are avoided, workers are happier and safer, and companies contribute less to environmental polution. In one enablement, the invention has been reduced to practice in a large package delivery facility that routinely processes 8000 to 10000 packages per day. Conversion to Electric Delivery Vehicles (EDVs) has commenced, and control of recharging times and speeds must not interfere with the conveyor belts and air quality equipment that keeps workers safe and deliveries on time. Peak loads occur in the morning, afternoon, and around midnight. Duration of these spiky loads depends on package volume, which is held constant at a continuous flow of poackages so that excessive load results in longer duration spikes in electricity consumption rather than in higher spikes in load. Our newly invented ML Forecasting system uses Support Vector Machine Regression (SVMR) to predict the day-ahead electric load of the facility using past histories of load for that day, hour, and weather prediction. It has a feedback loop that scores the statistical accuracy of its prediscions against the actual buildiong load, which is controlled by package volumes coming in and going out. The ML system learns from its errors, which are minimized over time. I can then convert this electrical load prediciton to a forecast of package volumes for the next day, week, month, and season. In another enablement, our Machine Learning Forecasting System is built into a commercial battery recharge optimization system so that tomorrows expected package load and weather forecasts are used to successfully optimize the time windows allocated for EDV fleet recharge and intensities of power to the batteries in each vehicle. Peak load spikes are avoided which draw penalties from the utility. In another enablement, the ML Forecasting System is used as a simulator, in that it can compute the scaling to hundreds of theoretical EDVs at this facility or other differently sized depot facilities to identify how electric load can be predicted and minimized, therefore requiring less new capital equipment from the utility since added supply may no longer be needed from the utility as the depot expands from 10% to 100% EDV's in the near future.
Roger N. Anderson
added 2 research items
Energy has been and will continue to be a significant global challenge. The United States consumes about 106.96 quadrillion BTU of energy per year, of which about 30% is used for transportation (EIA, 2007). Within the transportation sector, a great deal of energy is attributed to internal-combustion engines, and additionally contributes to greenhouse and noxious gas concentrations in the atmosphere that are detrimental for the climate, air quality and public health. Reforms to address these emissions have already begun, most significantly with increased interest in the electric vehicle. Although increased use of electric vehicles (EV) decreases carbon dioxide output, there is little infrastructure to support their wide-scale use. This thesis uses a thin vertical slice approach to analyze the viability of converting a garage to support intensive electric vehicle use.
Roger N. Anderson
added 2 research items
GE,Federal Express,Consolidated Edison and Columbia University demonstrate the nation’s first electric delivery vehicle depot implementing smart charging technology connected to the distribution grid in an intelligent manner.
Roger N. Anderson
added a research item
Electrification for commercial vehicle fleets presents opportunity to cut emissions, reduce fuel costs, and improve operational metrics. However, infrastructure limitations in urban areas often inhibit the ability to charge a significant number of electric vehicles, especially under one roof. This paper highlights a novel controls approach developed at GE Global Research in conjunction with Columbia University to fulfill the stated needs for intelligent charging of a commercial fleet of electric vehicles. This novel approach combines traditional control techniques with machine learning algorithms to adapt to customer behavior over time. The stated controls system is designed to regulate the charging rate of multiple electric vehicle supply equipment devices (EVSEs) to facilitate cost-optimal charging subject to past and predicted building load, vehicle energy requirements, and current conditions. In this embodiment, the system is primarily designed to mitigate electric demand charges that may otherwise occur due to charging at inopportune times. The system will be deployed at a New York City FedEx Express delivery depot in partnership with the local utility, Consolidated Edison Company of New York.
Albert Boulanger
added 4 research items
Electrification for commercial vehicle fleets presents opportunity to cut emissions, reduce fuel costs, and improve operational metrics. However, infrastructure limitations in urban areas often inhibit the ability to charge a significant number of electric vehicles, especially under one roof. This paper highlights a novel controls approach developed at GE Global Research in conjunction with Columbia University to fulfill the stated needs for intelligent charging of a commercial fleet of electric vehicles. This novel approach combines traditional control techniques with machine learning algorithms to adapt to customer behavior over time. The stated controls system is designed to regulate the charging rate of multiple electric vehicle supply equipment devices (EVSEs) to facilitate cost-optimal charging subject to past and predicted building load, vehicle energy requirements, and current conditions. In this embodiment, the system is primarily designed to mitigate electric demand charges that may otherwise occur due to charging at inopportune times. The system will be deployed at a New York City FedEx Express delivery depot in partnership with the local utility, Consolidated Edison Company of New York.
Vehicles, both personal and commercial, have become ubiquitous forms of transportation in the developed world. The auto industry is amidst a technological transformation in identifying alternative sources of energy to power vehicles due to two driving forces: environmental pollution prevention and depletion of fuel resources. This driver for developing "smarter" solutions to create a "smarter planet" is crucial to advancing the science behind electric vehicles (EVs). EVs have been in existence since the mid-19th century, and electric locomotion has been the commonplace in many other vehicle types such as trains. The focus of this chapter is to discuss the feasibility of EVs in smart cities. In particular, the chapter explores the types of EVs, advantages and challenges faced by EVs to penetrate the market, and to outline state-of-the-art research and technologies that are driving the creation of newer and better EVs for adoption in the smart cities of tomorrow.
Electric vehicles are not penetrating the market as quickly as expected. This is due to limited driving range, time required to recharge a battery, and lack of charging infrastructure in most metropolitan cities. We propose a charge sharing network in which we use inductive power transfer to wirelessly exchange charge between vehicles. In our network, vehicles that have excess charge to share, can sell charge to vehicles needing charge to reach their destination. In this paper, we describe a game theoretic approach to offering incentives for electric vehicles to participate in the charge sharing network. We utilize Nash Bargaining theory to show that participation in the network can yield profits for the seller driving to their destination and that we can increase the number of cars reaching their destination without needing to stop for recharging.
Albert Boulanger
added a project goal
Projects the Smart-X {Cities, Buildings, Grids} group did in vehicle electrification