Project

Intelligent Building Management

Goal: Research, develop and commercialize ML-Based Intelligent Building Management systems

Updates
0 new
2
Recommendations
0 new
0
Followers
0 new
2
Reads
0 new
107

Project log

Roger N. Anderson
added a research item
Book Explaining how Buildings need a Brain, which is Di-BOSS.
Roger N. Anderson
added a research item
Demonstration of DiBOSS Digital Building Operating System.
Roger N. Anderson
added a research item
The global technology company Selex ES of Finmeccanica is bringing to market Di-BOSS™, a new generation of digital building operating systems that will be unveiled at IBCon 2013 (Intelligent Building Conference) at the Orange County Convention Center in Orlando, June12-13. The new system, a joint development of Selex ES, Rudin Management, one of the largest privately held property management companies in New York City, and Columbia University! through! its School of Engineering and Applied Science, will be showcased in Booth #6769 where representatives from all three partners will be on hand to demonstrate the Di-BOSS™ system. "Di-BOSS™ is an entirely new kind of digital building operating system solution, a design that could only have resulted from this remarkable, collaborative effort," says Fabrizio Giulianini, Chief Executive Officer of Selex ES. "Di-BOSS™ represents an additional significant step in the development of Selex ES smart systems portfolio that spans integrated networked infrastructures for large urban environments through to specific break through products with a direct impact on citizens' lives, allowing for a new, smarter, more responsible and collaborative approach to energy management in sky scrapers." One primary feature of Di-BOSS™ is its ability to track occupancy on a large scale. "The technology to link the building management system with occupancy to control energy use is a cutting edge capability," says Columbia University energy researcher and professor, Dr. Roger Anderson. "The Di-BOSS™ system's continuous feedback loops give building managers reliable data to make decisions that significantly improve operating efficiency and better serve the people in the building." Di-BOSS™ Rudin Management piloted the Di-BOSS™ system in one of its largest New York City properties. Rudin employees provided critical user feedback that influenced the system's user interface design, report formats, and analytical capabilities. Di-Boss was developed "from the engine room out", organically with the guidance of experienced building engineers. The live testing quantified the scope of energy savings obtainable with the Di-BOSS™ as well as the improvements to security and comfort. "The Di-BOSS system's practical ease of use and ability to connect of all the building's systems are critical features that appealed greatly to our building managers and engineers. The Di-BOSS™ system had an immediate positive impact on our energy bill," says John Gilbert, Executive Vice President and Chief Operating Officer at Rudin Management. "We were able to use its real time forecasts to customize the next day's start-up and ramp-down schedules based on the weather and predicted occupancy. We also analyzed our tenants' specific energy consumption trends and worked together to generate some cost savings for them as well. That's not only money saved but energy resources conserved." The Di-BOSS™ system is "smart" and learns by continuously analyzing occupancy trends and other data points, such as weather and energy readings from multiple systems. Di-BOSS™ Press Office
Albert Boulanger
added a research item
The US EIA estimated in 2017 about 39\% of total U.S. energy consumption was by the residential and commercial sectors. Therefore, Intelligent Building Management (IBM) solutions that minimize consumption while maintaining tenant comfort are an important component in addressing climate change. A forecasting capability for accurate prediction of indoor temperatures in a planning horizon of 24 hours is essential to IBM. It should predict the indoor temperature in both short-term (e.g. 15 minutes) and long-term (e.g. 24 hours) periods accurately including weekends, major holidays, and minor holidays. Other requirements include the ability to predict the maximum and the minimum indoor temperatures precisely and provide the confidence for each prediction. To achieve these requirements, we propose a novel adjoint neural network architecture for time series prediction that uses an ancillary neural network to capture weekend and holiday information. We studied four long short-term memory (LSTM) based time series prediction networks within this architecture. We observed that the ancillary neural network helps to improve the prediction accuracy, the maximum and the minimum temperature prediction and model reliability for all networks tested.
Roger N. Anderson
added a research item
Tuning of Building Management Systems with Turner Construction Optimization of large micro-grids like the new Manhattanville campus with Columbia and military bases with DOD
Roger N. Anderson
added a research item
Roger N. Anderson
added 5 research items
Large and Complex facilities will have multiple Load and Source Management requirements with Abundant Alternative Energy Sources and Many Large Storage Devices.
Janet The plan for tomorrow is to first meet in our conference room to go through introductions, receive a background/better understanding on how ASU, Columbia, etc can support Luke's energy efficient program, have us explain our energy efficient program/efforts/outlook, and pinpoint areas of interest/potential.
Selling a "Smart Property Management System", a "Total Plant”, that sits at top of the “Pyramid”, and optimally manages whatever intelligent and legacy systems a Property Management Company has at each building in its portfolio. The SPMS orchestrates the energy, safety and environmental performance of each building in a coordinated, optimal way across the portfolio….all day, every day in real time. The “Secret Sauce” is that SPMS is designed by the owner/operator rather than just by the vendor who sells it.
Roger N. Anderson
added 3 research items
MODERN OFFICE TOWERS dominate the skylines of the world's great cities, and occupants of these structures consume up to 40 percent of the total energy usage of the cities in which they work and live. There is an emerging need to more efficiently manage the energy use of this existing office inventory, as well as new space under construction across the world. Luckily, a revolution in energy use is under way for the urban or vertical city that combines on-site generation, efficiency and new storage options. Beginning with electricity usage, the smart grid is being extended from utilities into the customers' premises and end systems. Users are now empowered to integrate the data embedded and collected by the machines that run these facilities. Total property integration will empower management to operate more efficiently. With innovative information and data management analytics, we are now able to process the terrabytes of data generated by building management systems as well as the other subsystems that perform valuable functions in the operation of high-rise buildings into a complex computer-aided lean management system that we call total property optimization. This optimization revolution is being driven not just by advanced software, but also by new and innovative hardware such as smart elevator systems, the single most critical transportation system of the vertical city. There are new building sensors that track real-time occupancy on every floor, variable frequency drives for HVAC fans and motors, LED lighting networks where the ballast has been
Determine: Goal: • How much to preheat / precool at the beginning of the day • Minimization of overall steam costs by trading off preheating / precooling costs against demand penalty costs
We have developed a Total Property Optimization (TPO) System for owners who manage a portfolio of large energy consuming buildings and facilities such as microgrids (University campuses and Military Base Facilities), commercial and residential skyscrapers (Rudin Management in Manhattan), Electric Vehicle fleets (Fedex Express, UPS, Hertz, Avis) and large management and construction companies (Turner Construction, Lend Lease). The prototype TPO is being deployed as part of a DOE Smart Grid Demonstration Project at Con Edison in New York City (U.S. Patent 7,395,252 and others pending). In order to optimize energy use, efficiency and impact on the environment, the TPO must contain modules for 1) data acquisition of the energy consumption and environmental footprint of each individual building within the portfolio, 2) machine learning models such as Support Vector Machines and CCLS invented ML Models such as Martingale Boosting, U.S. Patent 8036996) to continually learn from the past performance of each building in the facility and its response to exogenous forces such as weather and climate, 3) an interactive interface with the human-in- the-loop management of the facilities to convey distinct action recommendations for all buildings within the portfolio, 4) a cost benefit verifier, and 5) a feedback loop so that the machine learning models can maintain the facility on the efficient frontier of energy economy and environmental stewardship.
Roger N. Anderson
added 3 research items
It is the belief that sufficient warning, on the order of 30 seconds +/-, may be provided in the likelihood of a major regional system outage that would interrupt power and result in a major blackout. The measurements that indicate this impending grid stability crisis include frequency, voltage and phase relationships relative to voltage and current. This is how transmission utilities already monitor the transmission system using protective devices and algorithms. Commercial customers or end-users, if the signal were available, could use similar warning signals to improve the safety of people and protect and safeguard their equipment and systems. There may also be an opportunity to create new capabilities beyond the transmission system events using local measurements to predict local problems.
We clearly see the impact of occupancy on the space temperatures. 9 AM arrival and afternoon departures of 345 Park result in increases in temperature (each person gives off the equivalent of a 100 watt light bulb. We are driving to remove the afternoon hump by backing off on the chillers and fan speeds. Manual operation has shown this to be effective in the summer of 2012. With forecasts of space temperature and occupancy, we can anticipate and make the system work automatically.
No solution provider delivering visibility and advanced analytics of integrated Skyscraper sub-systems with tangible operational results as of 2012
Roger N. Anderson
added 11 research items
The degree to which comfort conditions can be achieved depends not only on internal ambient temperatures, but also on the temperature of the internal masses. The thermal mass can have a great effect on the energy performance of the building, and the HVAC plant and controls should be designed to respond to a specific building's needs. In this project, we used the operational data of 345 Park Ave. Our goal of this project is to model the thermal response of the building according to the operational data. This would be the first step to for the optimal scheduling of the HVAC (or any subsystem that support it like chilled water) running time, with a goal for optimizing energy savings while maintaining required-by-lease-terms comfort levels.
Machine Learning system then prioritizes this ranking to provide a color coded rating of the likelihood of which feeders and components are the most at risk and susceptible to actually failing witin the next 14 days . … Just near Rudin Buldings for now… Susceptibility inputs are: Live NetRMS and FRA Live Load PocketWeight Live PQNode Events Static Composition Attributes Jeopardy database Hipot Index database Outage history database Susceptibility outputs are: Feeder rankings Section rankings Xfmr rankings LPW variances A List of the most at risk Feeders is forwarded to the Feeder Hotlist program to evaluate their cumulative Load Pocket Weight problems along their length. The Xfmr Load Transfer Variance from the Xfmrs that are on the feeders that are out-of-service to their nearby pick-up Xfmrs is evaluated and tracked over Time using: Load Transfer Variance Feeder Hotlist Ranking PQNode Events and Outages
Roger N. Anderson
added 10 research items
Value added: Lower the usage of energy to maintain lease requirements for space temperatures and accurately identify how many building classes are needed. Specifying Set Points can yield efficiencies and energy savings in buildings that manually adjust now. • Make Set Point recommendations for Supply Air Temperatures (SAT), fan-by-fan in zonal HVAC buildings. • Other Set Point recommendations, if possible, such as hot and cold water temperature, Variable Frequency Drive (VFD) settings. • Better optimization of Space Temperatures using "Black Box" model (another Machine Learning (ML) based system identification method). Timing: Expect availability during Fall 2013. Value added: Identification and quantification of Cost/Benefit CBA: • Baseline building performance metrics for quantifying energy and environmental savings before Di-BOSS v1 installation. • Continuously Monitor and Analyze TPO v1 and Di-Boss v1 performance in ALL 16 buildings at Rudin Management and also new customers. • Make any necessary TPO v1.x software modifications to incorporate discoveries made in installation and analysis of all16 Rudin buildings and new coustomers. Timing: Expect availability during Fall 2013.
Somewhat correlated with Humidex What does this mean? Model likely needs more covariates to predict during very hot or very cold days What can be done? Filtering training data based on forecasted weather might help Covariates do not fully capture decision-making process Other ideas?
Roger N. Anderson
added 3 research items
Di-BOSS is the next-generation Digital Building Operating System Solution that optimizes energy consumption, saving money and increasing security without sacrificing comfort. Di-BOSS continuously commissions data from multiple operations systems, floor-level occupancy, and ambient and forecast weather conditions. Data streams are integrated into a portable dashboard control panel, providing complete visibility.
With today's energy prices and the focus on conservation and sustainability, intelligent building operators need systems that provide reliable, real-time data that forecast effective system-wide decisions. At 345 Park Avenue, a two million SF skyscraper built in the 1960's and 560 Lexington, a 330,000 SF building with more modern control systems, we have developed the prototype for the building operating system required to simultaneously optimize all subsystems in large office buildings. These two buildings demonstrate a new operating system that analyzes continuous streams of data, responds with proactive operational recommendations, and learns the personality of each building, utilizing a Smart System that learns and improves with time. Goals: Operations Energy Efficiency Sustainability Tenant Experience Financial Optimization Challenges: • Prove that Energy Efficiency gains from manual control can be replicated by automated building operating system • Develop different classes of office buildings so that the installation and use are easy and quick • Training people in optimal building management • Communications integrated seamlessly with Intelligent Building Operations • Valid in International skyscrapers in different climate zones Successes: • We will demonstrate intelligent problem identification and solution recommendations that allow for continuous, real-time re-commissioning • Machine Learning provides a reliable forecast 24 hours in advance to recommend operating schedules that optimize tenant comfort and safety while minimizing energy consumption • Machine Learning also provides a now-cast of the next 2 hours to guide operators in set-points, preheating and cooling, and start-up and ramp-down timing • Tenants are engaged in the energy savings and environmental footprint. « « Sponsored By:
DiBOSS is a platform neutral “cockpit” that provides situational awareness by integrating data from existing independent building systems (HVAC, BMS, Occupancy, Elevators, Fire and Security) to proactively correct, adjust, and re-commission these systems for total performance optimization, efficiency and reduced carbon footprint.
Roger N. Anderson
added 8 research items
Temperatures peaked in the 90’s all week Humidity peaked in the 60’s all week Each day at Noon, ConEd invoked a mandatory Demand/Response load shedding event Di-BOSS, using its patented Nowcast within theTPO forecasting tool, steered within the tenant comfort horizon each day 345 Park Avenue (1.8 million sq ft) shed an avg of 200KW in the morning and a further 500KW beginning at noon each day, saving 22,400KWH for the week while maintaining comfort and safety of all tenants, for a ~10% electricity reduction in electric load for the week. 560 Lexington (300,000 sq ft) shed 250KW after noon each day, saving 7520KWH for the week while maintaining comfort and safety of all tenants, for a ~20% electricity reduction in electric load for the week
The Now-Cast module of the Total Property Optimizer (TPO) is also a human-in-the-loop system, which uses advanced analytics to provide building operators with the ability to steer the building to the most efficient energy comfort level floor-by-floor. The Now-Cast module of TPO uses a Support Vector Machine learning system to learn the thermodynamic response of each floor to HVAC set point changes, and uses supply air and return air temperatures along with real-time monitoring of space temperatures on each floor to steer the floor using the TPO Horizon Indicator. The Now-Cast space temperature trajectory suite of machine learning sits atop a primary layer of 24-hour predictions, and gives insights into and makes predictions about the effects of the current setting of the buildings temperature values and the values of the building operator's control levers on ambient space temperature. Utilizing both historical and predicted data, it uses a blend of relevant covariates to guide the building operators in ensuring their decisions will not break tenant lease requirements. This, in short, takes any guesswork out of the building's operation. Each run of the suite provides temperature predictions for 2 hours, resulting in 8 predictions (at 15 minute resolution) per floor. The Now-Cast predicts a 2-hour-ahead a space temperature trajectory using the TPO machine learning suite that relies on 3 forms of input data: 1.) Real-time space temperature values: The real-time BMS data feed provides a view of the current temperature of the air in critical parts of the building. Basic thermodynamic modeling allows the Now-Cast to identify correlative relationships between the various air and water temperature HVAC settings and the ambient space temperatures. 2.) Supply air and return air temperatures measured by the BMS and set points specific fans feeding central air and heat to the appropriate floors: The real-time BMS data also provides a view of the current set point values for a variety of the engineering team's control systems. These are often in the form of thermostat set point values: 3.) The forecast temperatures from the TPO Adding the 24-hour forecast values to this past history, TPO is able to provide an abstract covariate set into the Now-Cast predictions. Intrinsically, the Now-Cast includes short term covariates like current occupancy, outside weather and tenant response to holidays. The nature of the Now-Cast as time-series data allows for robust covariate
A Support Vector Machine (SVM) model was selected for day-ahead forecasting of space temperature, electric and steam loads, and occupancy, after back-test experimentation with actual building and weather data. SVM is a classification algorithm, which uses the concept of maximizing a dividing hyper-plane as the methodology to learn the functional mapping between the input covariates and output forecasts. Since there is a cyclical component in the building load profiles, covariates such as previous day space temperature, electrical, steam and occupancy loads, previous week load, previous day average, previous week average, time-of-the-day and day-of-the-week are incorporated in the learning model. Additionally to account for the weather variant HVAC load, an index called Humidex (composite of temperature and wet bulb dew point) is included as a covariate. Each of the four forecasting components for electric (building and tenant), steam, space temperature and occupancy, are designed to perform two functions: prediction and learning of parameter optimization variables, with the optimization function performed daily and the prediction function performed hourly. That is, the parameter-optimized SVM prediction function is used hourly by TPO to calculate 24 different day-ahead SVM model forecasts for space temperatures of key floors, electrical and steam consumption for the building, electrical usage for tenants with smart meters, and building occupancy corresponding to each hour of each new day. For the daily parameter optimization, a grid search with exponential distance between the grids is used to find the optimal values of the parameters in the SVM model. In view of varying time series input data, a customized cross-validation algorithm was also implemented. TPO partitions the training data into two sets: 1) all available data but the latest week which is used to train the SVM model and 2) the latest week's data that were left out of the training which is used to validate the prediction. Minimizing the day-today variance of the error metric Mean Absolute Percentage Error (MAPE) is used as the performance score for parameter optimization. These MAPE values are then averaged using exponentially decaying weights with the most recent week receiving the highest weight. The set corresponding to the minimum average MAPE are selected as the optimal parameters for each hour's forecasting model run for that day. Retraining occurs a little after midnight each day. For the hourly forecasts, the diagram in Figure 1 (see below) represents the basic internal processes performed for each respective component addressing one of the four forecasting functions for electric (building and tenant), steam, space temperature and occupancy. For example, the EPRED component for electric load prediction involves: (1) a process to "clean" the input data from the building systems and sensors; this data is processed to obtain covariates (variables exhibiting correlated variation), (2) the covariates are then applied by a task specific TPO SVM performing analytical machine learning,
Roger N. Anderson
added 6 research items
Previously, various machine learning models such as Bayesian Additive Regression Trees (BART), Support Vector Regression (SVR), Artificial Neural Networks (ANN) and time series methods such as variants of SARIMA were used in our group to perform load forecast on electric load data. Amongst the tested models, Support Vector Regression (SVR) stood out as the best performer. But the day-to-day variance of error metric Mean Absolute Percentage Error (MAPE) is large and it seems that the model can be further improved. One of the contributors to the variance of error is change of seasons. A model that incorporates concept drift and uses different models during different seasons is hypothesized to outperform the current model.
Energy efficiency in buildings is one of the most important topics facing energy sustainability and urbanization today. Accounting for how weather interacts with buildings gives insight that can improve building energy performance. The Center for Computational Learning Systems (CCLS) has developed Total Property Optimizer (TPO), a suite of machine learning based decision and optimization algorithms for energy usage optimization of buildings. TPO is incorporated in SELEX Elsag's (a subsidiary of Finmeccanica) Di-BOSS system that takes in multiple sources of building data including the building's Building Management System (BMS). TPO uses weather data-both historical observations and forecasts-as part of its optimization algorithms. As Di-BOSS expands from its current two locations in midtown Manhattan to the Financial District and eventually to other cities, it is important to understand the climatic differences that affect each building's thermal profile. On the micro-scale, the buildings are affected by forcings unique to urban environments: Urban Heat Island, wind effects from street canyons, shading from surrounding urban topography. In this study, I attempt to identify patterns in the urban climatic conditions surrounding the subject buildings. I then attempt to improve upon the building optimization techniques by refining the weather inputs based on observed patterns.
The CCLS group is interested in controlling a number of systems within each of the buildings: steam heating, air conditioning, thermostat set points, CO2 set points, ventilation, reheating with emergency heating coils, blinds, variable frequency circulation fans, et cetera. While most of these systems do have small levels of interaction (e.g. CO2 levels are raised by injecting outside air, which may require heating or cooling), they are generally modeled as independent at the system level. Control problems are building dependent, due to the di↵erent systems and building layouts. We more closely examine control problems for three systems: steam preheating in the winter season, thermostat set points, and building start up/ramp down sequences.
Roger N. Anderson
added 9 research items
This statement of Work (SOW) defines the activities and deliverables required for the design, engineering development, interfacing to SELEX Elsag “Smart Property Management System” (SMPS) and test of the Total Property Optimization (TPO) sub􏰁system developed and being developed by Columbia University. TPO is expected to advise the building operators how to achieve certain level of KPI (refer to § 5) for each monitored building and to plan in advance certain assets of the equipment in order to face out potentially dangerous situation that might originate from blackout, storms and the like. TPO is a complete software package that is fed by SPMS and that writes data back into SPMS. Moreover TPO will show advises and recommendations via its own specific Human Machine Interface.
The global technology company Selex ES, a subsidiary of Finmeccanica announced that Rudin Management, one of the largest privately held property management companies in NYC, will install the Di-BOSS™ digital building operating system with Columbia’s TPO included in its remaining 14 commercial office towers following successful commissioning of system at 345 Park Avenue and 560 Lexington Avenue. The Di-BOSS operating system is a joint development of Selex ES, Rudin Management, and Columbia University’s Center for Computational Learning Systems.
Roger N. Anderson
added 2 research items
Commercial and residential buildings are designed for tenant comfort and safety, with energy-efficient equipment managed by Building Management Systems (BMS). BMSs can integrate a number of components to assist building operators with maintenance and operation. BMS can be used to retrieve building energy-related data, such as data reading from electric sub-meters and space temperature sensors. Such systems can be operated efficiently to reduce costs of energy consumption while maintaining quality of comfort for tenants. For example, in the case of a commercial building, air condition systems can be regulated during business hours by lowering the speed of electric fans using variable frequency drives (VFDs) to reduce energy costs. However, BMS do not guarantee tenant comfort and reliable building operation because they do not integrate with lighting systems, tenant owned supplemental air conditioning and heating systems, elevator management systems, power systems, fire systems, security systems and the like. Accordingly, there is a need for improved techniques for optimizing the system-of-systems in larger buildings in order to simultaneously optimize comfort, safety, energy efficiency and equipment reliability in building operations and management.
Efficiency and safety will drive the design of buildings of the future – true vertical cities endowed with distributed intelligence, which will ensure, at the same time, cost savings, increased comfort and compliance with high environmental standards.
Roger N. Anderson
added a research item
A win-win value creation for both the grid and the customer will be demonstrated through our C2SOS business risk management system that predicts through simulation the reliability exposure of the electric grid and interacts with customer's load management systems in their buildings to recommend appropriate actions to reduce operational risks to both the utility and the customer. These Monte Carlo methods based simulations predict the risk of electric supply not meeting load through forecasted weather impacts on load and supply, responsiveness from distributed resource, and our existing machine learning based predictions of equipment heath. The simulator uses an interdependency approach to the transmission, distribution and customers to determine predicted and real time exposure to electric grid unavailability. If risks of providing electric service rise at specific locations on the grid as a result of say transmission constraints, or distribution constraints, changes in price signals are recommended for management of energy flow on the grid and information will be communicated to affected customers on the potential impact to reliability and suggest a change in their usage of energy for their internal operations. This will better enable customers to prepare internally for potential emergency situations that could impact their operations. In addition, C2SOS will suggest changes in the field, like changing positions
Roger N. Anderson
added a research item
The global technology company Selex ES of Finmeccanica is bringing to market Di-BOSS™, a new generation of digital building operating systems that will be unveiled at IBCon 2013 (Intelligent Building Conference) at the Orange County Convention Center in Orlando, June12-13. The new system, a joint development of Selex ES, Rudin Management, one of the largest privately held property management companies in New York City, and Columbia University through its School of Engineering and Applied Science, will be showcased in Booth #6769 where representatives from all three partners will be on hand to demonstrate the Di-BOSS™ system. "Di-BOSS™ is an entirely new kind of digital building operating system solution, a design that could only have resulted from this remarkable, collaborative effort," says Fabrizio Giulianini, Chief Executive Officer of Selex ES. "Di-BOSS™ represents an additional significant step in the development of Selex ES smart systems portfolio that spans integrated networked infrastructures for large urban environments through to specific break through products with a direct impact on citizens' lives, allowing for a new, smarter, more responsible and collaborative approach to energy management in sky scrapers." One primary feature of Di-BOSS™ is its ability to track occupancy on a large scale. "The technology to link the building management system with occupancy to control energy use is a cutting edge capability," says Columbia University energy researcher and professor, Dr. Roger Anderson. "The Di-BOSS™ system's continuous feedback loops give building managers reliable data to make decisions that significantly improve operating efficiency and better serve the people in the building." Di-BOSS™ Rudin Management piloted the Di-BOSS™ system in one of its largest New York City properties. Rudin employees provided critical user feedback that influenced the system's user interface design, report formats, and analytical capabilities. Di-Boss was developed "from the engine room out", organically with the guidance of experienced building engineers. The live testing quantified the scope of energy savings obtainable with the Di-BOSS™ as well as the improvements to security and comfort. "The Di-BOSS system's practical ease of use and ability to connect of all the building's systems are critical features that appealed greatly to our building managers and engineers. The Di-BOSS™ system had an immediate positive impact on our energy bill," says John Gilbert, Executive Vice President and Chief Operating Officer at Rudin Management. "We were able to use its real time forecasts to customize the next day's start-up and ramp-down schedules based on the weather and predicted occupancy. We also analyzed our tenants' specific energy consumption trends and worked together to generate some cost savings for them as well. That's not only money saved but energy resources conserved." The Di-BOSS™ system is "smart" and learns by continuously analyzing occupancy trends and other data points, such as weather and energy readings from multiple systems. Di-BOSS™ Press Office
Roger N. Anderson
added 2 research items
Power grid/marketplace overview, how energy users impact that system, various energy efficiency methods & technologies including status of alternate energy systems and finally opinions/projections/prognostications as of 2004.
Net energy saving: 30%+ over New York State Energy Efficiency Building Code: 50% extra insulation on roof and walls low-emissivity glass; high shading coefficient daylighting used as much as possible, in both public areas and 15 ft perimeter gas-fired absorption chiller/boilers sequentially-sized chillers to match loads separate HVAC units for each floor central building management system all fans, pumps, and motors have VSDs all offices have occupancy sensors high efficiency lighting: < 1 W/sq.ft. two 200-kW gas-fired fuel cells (4%) photovoltaic cells on South and East exposures (1%)
Roger N. Anderson
added a research item
OVERVIEW Forty percent of the energy in Manhattan is used in high-rise ofce buildings and, of that energy, forty percent goes towards controlling the heating, ventilation, and air conditioning (HVAC) systems controlling in these buildings. The costs associated with this energy are high, particularly for heating buildings with steam in the winter. This technology describes a module which helps forecast any necessary control adjustments that will keep the building at the desired temperature, and recommends preheating settings to reduce steam usage during peak hours. This technology not only increases energy efciency, but also reduces energy costs. Temperature trajectory module offers suggestions for HVAC temperature settings that improve energy efciency and lower steam costs This technology is another step towards more efcient, cost-effective energy usage. The technology is a temperature trajectory module that uses advanced analytics to make suggestions to the building manager who controls the oor-by-oor adjustments that modify the temperature. It uses the ambient space temperature, the current temperatures on each oor, and the various HVAC settings to make its recommendations for the upcoming two hours. Similar algorithms are used to make 24-hour predictions, which can determine whether to preheat the building with steam before peak hours render the steam a hundred times more expensive. This technology has the potential to provide great savings in energy and cost to Manhattan. A prototype module has been developed and tested in a tenant space over one winter season, reducing energy consumption by 7%, resulting in a savings of at least $75,000.
Roger N. Anderson
added 3 research items
Columbia University’s Center for Computational Learning Systems (CCLS) is working with Finmeccanica/Selex and Rudin Management to develop an integrated intelligent building portfolio management and operation support system called SPMS. CCLS’s contribution includes the research, development, and deployment of a SPMS subsystem called the Total Property Optimization (TPO) system. TPO is an effective machine learning technology that analytically monitors and learns critical building environmental behaviors and makes recommendations to optimize performance by building operations to maintain correct environment levels with economic benefits. TPO efficiently learns via analysis of massive data collected from an expressive range of environmental sources and sensors distributed throughout one or more buildings.
Selex ES, Inc. is reporting on the economic impact realized by our company from the period of July 1, 2013 through June 30, 2014, related to the following project(s) conducted with the support of the Center for Advanced Information Management. Project Title: The economic impacts include: Property Optimization for Di-Boss Roger Anderson
Recommender in TPO II works by learning from how the building is operated. This has shortcomings. It learns from past operators ‘ decisions. Are those decisions the best in terms of building operation objectives? The algorithms we use now to produce recommendations do not factor the uncertainty of the inputs and forecasts, e.g. weather (temp, humidity, sunlight), occupancy, energy prices (if market driven), variance in building operation and down equipment to produce robust decisions Ideally we would like to make optimized robust decisions based on KPIs (these are part of the objective function) that work within the range of possible building responses during the course of the day. ASC v3.0 represents a family of algorithms that optimize decisions under the face of external conditions (i.e. weather, real time pricing) not under our control (exogenous variables) that have some uncertainty. ASC v3.0 produces robust recommendations that yield even more more energy savingsthan v2.0.
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 2 research items
Machine learning software for optimization of building energy usage This technology is a machine learning algorithm which uses historical and real-time operational building data to predict building energy demands.
Forty percent of the energy in Manhattan NY is used in high-rise office buildings and, of that energy, forty percent goes towards controlling the heating, ventilation, and air conditioning (HVAC) systems controlling energy consumption in these buildings. The costs associated with this energy are high, particularly for heating buildings with steam in the winter. This technology describes a module which helps forecast any necessary control adjustments that will keep the building at the desired temperature, and recommends preheating settings to reduce steam usage during peak hours. This technology not only increases energy efficiency, but also reduces energy costs. The technology is a temperature trajectory module that uses advanced Machine Learning analytics to make suggestions to the building management system that controls the door-by-door adjustments that modify the space temperature. Similar algorithms are used to make 24-hour predictions, which can determine whether to preheat the building with steam before peak hours render the steam a hundred times more expensive. This technology has the potential to provide great savings in energy and cost to urban cities.
Roger N. Anderson
added 4 research items
New York City's steam system supplies 27 billion pounds a year to heat, cool, and power Manhattan buildings. Many commercial buildings use steam to meet their space temperature requirements. Contractually, landlords must maintain a space temperature within a specific range during the workday. As a result, peak demand for steam in New York occurs during the colder months of the year. The provider of these steam services, Con Edison, charges an additional on-peak fee for steam demanded between the hours of 6 and 11 in the morning from December to March, as the workday begins. This on-peak-fee is equal to $1,629 times the maximum rate of steam, measured in million pounds per hour (Mlb\hr), demanded during on-peak hours within a billing cycle between December and March. This rate is the most expensive within Con Edison's billing structure, by far. To minimize this charge during building start-up (see page 11, line 12 of the TPO patent application 070050.4640 NY02 755047, and Claims 1,2,9,10) building managers attempt to heat the building using steam before 6 am. By storing energy generated by steam before 6 am, they "preheat" the building using an effective Hydro-Battery. That is they pump heated water into the riser circulation system of the building before 6am and return the hot water to the HVAC system after 6am at little additional cost. Specifically, if the maximum rate of steam demanded before 6 am is greater than the maximum rate demanded between 6 and 11 am, the building has been "preheated", and the cost of that off-peak-fee steam is 100 times cheaper. An example of preheat during building start-up is shown below in figure 1. These graphs come from building data supplied by 345 Park Avenue, a building that has shared its real time data with Columbia University. Typically, building managers at 345 Park preheat each day during the on-peak winter heating months. The two dates in each plot have similar weather based on the their heat indexes. In each plot, a spike occurs each day before 7 am. To store heat, building managers turn on water pumps to fill the vertical riser pipes with hot water. This sudden increase in demand for heat results in a steam demand spike. To release the stored heat, building managers turn on the HVAC fans that then circulate air heated by the hot water throughout the building. On preheat days, steam demand spikes occur before 6 am. One can observe that preheat does not always result in a greater daily steam consumption or spike in steam usage. Given these temperature and peak demand penalties, building managers do not yet know the optimal method to reduce steam demand and steam costs during these on
Roger N. Anderson
added a research item
Machine learning software for optimization of building energy usage This technology is a machine learning algorithm which uses historical and real-time operational building data to predict building energy demands. ## Unmet Need: Prevention of large-scale, excess energy consumption Current methods to implement energy efficiency in commercial and residential buildings can be designed for tenant comfort, energy efficiency and system reliability in mind with the use of energy-efficient materials and building management systems (BMSs). However, these BMSs do not always guarantee reliable building operation because they do not measure or provide visibility and data analytics of space temperatures or occupancy variations sufficiently, leading to buildings often consuming energy at levels that exceed design specifications. As a result, there is a current need to implement systems capable of optimizing building management and improving energy efficiency. ## The Technology: Machine learning algorithm for monitoring buildings and optimizing energy use This technology consists of a machine learning algorithm which incorporates both past and real-time operational building data, such as humidity and temperature, to predict building energy demands. Historical temperature and humidity readings are used to train the algorithm, which then makes future predictions and recommends operation actions. Using an automated online evaluator, predictions are compared to real-time data to further refine the algorithm, detect anomalies, and ensure reliability. This technology allows for optimization of overall building energy use and can be applied in both commercial and residential buildings to reduce energy costs. This technology has been validated in a multi-tenant Manhattan office building for both spring and winter seasons.
Roger N. Anderson
added 14 research items
Overview • Our software, Di-BOSS™: Digital Building Operating System Solution, delivers recommendations that enable commercial building managers to make proactive, energy-efficient decisions. • Di-BOSS continuously commissions data from multiple operating systems, floor-level occupancy, and weather conditions. • We rely on support vector regression and extremely randomized trees to model the building's thermodynamics and predict energy consumption and occupancy. • On top of our models, we use dynamic programming to optimize operational decisions, such heating decisions. • Actionable data and analyses are presented in a real time control panel for building managers. Data • Our data includes sensor measurements every 15 minutes • Data includes air and water temperatures, HVAC set points, occupancy, heat exchanger and fan indicators, outdoor conditions, weather forecasts, as well as energy demand. • We provide a graphical interface to visualize these various time series. 24 Hour Prediction Models • We provide the following predictions on a 24 hours in advance: • Occupancy • Energy usage, both steam and electric • Space temperature for regions on each floor • Start-up: when to start turn on fans • Ramp-down: when to turn down the building • For space temperature, we provide building managers with a more granular two hour forecast for each sensor. • Additionally, we provide recommendations for the optimal time to start the heat exchanger during winter months. • For each one hour interval in a day, we predict occupancy, space temperature, and energy usage using support vector regression [Harris et al., 1997] • We retrain this model every hour, preforming 5-fold cross validation to set our parameters. • Our features include on data from the previous 60 days such as: • Steam demand on days with similar weather • Humidex, a measure of heat and humidity • Building temperature • Day of the week • Percent of tenants open for business • Time of day • Regular business hour indicators • Historical observations of the target variable • We display evaluation metrics of these models using RMSE as well as upper and lower bounds on our forecasts NowCast • Using methods similar to our 24-hour predictions, we provide a more granular 2-hour forecast for temperatures different regions • Notable features include: • Supply air temperature: temperature entering the floor • Set points: thermostat settings • Return air temperature: air temperature exiting the floor • Results from our 24-hour predictions Start-up/Ramp-down • Framing start-up and ramp-down as classification problems, we define our actions as the decision boundary between day and night mode. • We train an SVM using the past 14 days of data, limiting our search space to the historical range of start-up and ramp-down, respectively. • With historical data and historical predictions in our feature set, we recommend a decision time. • Recommendations are evaluated by building managers for feasibility based on their domain knowledge, which are used to find an acceptance rate. Preheat Recommendation • Between December and March, Con-Edison charges an additional demand charge each billing period for steam usage, proportional to the maximum rate of steam demanded on weekdays between 6 and 11 a.m. • This fee is the largest fee in Con-Edison's billing structure • In the past, building managers at 345 Park have attempted to lower their steam costs by "preheating:" turning on the heat exchangers to before 6 a.m. to start storing heat earlier in the day • Because of the cost structure, preheating everyday is unnecessary • We frame preheat recommendation as reinforcement learning problem in which our actions are our preheat times and our states are the outside weather and the steam use factors that constitute Con-Edison's billing structure: total usage, maximum usage, and maximum morning usage • We the expected steam demand for each portion of the billing structure using Extremely Randomized Trees [Geurts, 2006] • With dynamic programming, we find the optimal policy for a given cycle • Our model outperforms both historical usage patterns as well a policy of consistently preheating Initial Use • Our software has been in production in two skyscrapers since 11/2012. • We recently added two additional buildings this fall • Based on our predictions, we have saved $63k per month. • Our preheat model is expected to save an additional $40k per month. Figure 2: Space temperature plots for the west side of the 2nd to 8th floors of 345 Park. Our forecast (red) expects space temperature to remain within the contractual range (yellow/green band). Actual pace temperature (dark blue) stays within our 68% confidence interval (orange) and is followed by the NowCast (pink) for the following two hours. We display our recommendation for start-up (red square), preheat (pink triangle), and ramp-down (blue circle). Finally, we provide building managers with supply air temperature (light blue), temperature of air entering the zone, and their thermostat set points (gray). Supply temperatures are above the typical range (green band). Figure 2: Occupancy (left) and steam (right) predictions for the week of 11/18/2013. Historical data is in blue, while predictions and confidence intervals are in red and orange. Our steam plots also include our recommendations for start-up (blue circle), ramp-down (red square), and preheat (pink triangle).
STAGE OF DEVELOPMENT Proposal PENDING DISCLOSURES None Listed 0. INTERNAL ABSTRACT Building power management and control systems typically rely solely on the conditions and usage configuration of the local building to inform how power is consumed. This forgoes the potential to optimize power usage by considering conditions and factors external to the building; these conditions may include energy information from utilities (i.e., the smart grid), weather information from meteorological services, and announcements from government agencies. Building Aggregation Is Smart (BAIS) is a technology for combining internal power management information with relevant external data to enable optimal power control decisions that reduce energy costs. The technology's two-tiered architecture comprises hardware modules for providing local power usage information to property managers and modules for aggregating information from multiple properties to obtain greater energy cost reductions. BAIS makes use of an additional technology called Secure Internet Protocol-Based Smart Adapter (SIPSA) that enables secure bidirectional communication between a legacy power management system and the smart grid over the Internet. SIPSA can be used to both retrieve external information required to optimize power control decisions and to transmit local power usage data to other parts of the BAIS architecture. 1. MARKETING TITLE Energy cost reduction by aggregation of local power usage data 2. MARKETING ABSTRACT Typical building power management and control systems rely solely upon a building's internal power usage configuration and conditions. These systems forgo the potential for optimizing power usage by not considering available data from external sources such as utilities, weather servicesss, and governmental agencies, to make power control decisions. This technology describes software architecture for the aggregation and sharing of local power usage and a hardware/software adapter for
Overview • Our software, Di-BOSS™: Digital Building Operating System Solution, delivers recommendations that enable commercial building managers to make proactive, energy-efficient decisions. • Di-BOSS continuously commissions data from multiple operating systems, floor-level occupancy, and weather conditions. • We rely on support vector regression and extremely randomized trees to model the building's thermodynamics and predict energy consumption and occupancy. • On top of our models, we use dynamic programming to optimize operational decisions, such heating decisions. • Actionable data and analyses are presented in a real time control panel for building managers. Data • Our data includes sensor measurements every 15 minutes • Data includes air and water temperatures, HVAC set points, occupancy, heat exchanger and fan indicators, outdoor conditions, weather forecasts, as well as energy demand. • We provide a graphical interface to visualize these various time series. 24 Hour Prediction Models • We provide the following predictions on a 24 hours in advance: • Occupancy • Energy usage, both steam and electric • Space temperature for regions on each floor • Start-up: when to start turn on fans • Ramp-down: when to turn down the building • For space temperature, we provide building managers with a more granular two hour forecast for each sensor. • Additionally, we provide recommendations for the optimal time to start the heat exchanger during winter months. • For each one hour interval in a day, we predict occupancy, space temperature, and energy usage using support vector regression [Harris et al., 1997] • We retrain this model every hour, preforming 5-fold cross validation to set our parameters. • Our features include on data from the previous 60 days such as: • Steam demand on days with similar weather • Humidex, a measure of heat and humidity • Building temperature • Day of the week • Percent of tenants open for business • Time of day • Regular business hour indicators • Historical observations of the target variable • We display evaluation metrics of these models using RMSE as well as upper and lower bounds on our forecasts NowCast • Using methods similar to our 24-hour predictions, we provide a more granular 2-hour forecast for temperatures different regions • Notable features include: • Supply air temperature: temperature entering the floor • Set points: thermostat settings • Return air temperature: air temperature exiting the floor • Results from our 24-hour predictions Start-up/Ramp-down • Framing start-up and ramp-down as classification problems, we define our actions as the decision boundary between day and night mode. • We train an SVM using the past 14 days of data, limiting our search space to the historical range of start-up and ramp-down, respectively. • With historical data and historical predictions in our feature set, we recommend a decision time. • Recommendations are evaluated by building managers for feasibility based on their domain knowledge, which are used to find an acceptance rate. Preheat Recommendation • Between December and March, Con-Edison charges an additional demand charge each billing period for steam usage, proportional to the maximum rate of steam demanded on weekdays between 6 and 11 a.m. • This fee is the largest fee in Con-Edison's billing structure • In the past, building managers at 345 Park have attempted to lower their steam costs by "preheating:" turning on the heat exchangers to before 6 a.m. to start storing heat earlier in the day • Because of the cost structure, preheating everyday is unnecessary • We frame preheat recommendation as reinforcement learning problem in which our actions are our preheat times and our states are the outside weather and the steam use factors that constitute Con-Edison's billing structure: total usage, maximum usage, and maximum morning usage • We the expected steam demand for each portion of the billing structure using Extremely Randomized Trees [Geurts, 2006] • With dynamic programming, we find the optimal policy for a given cycle • Our model outperforms both historical usage patterns as well a policy of consistently preheating Initial Use • Our software has been in production in two skyscrapers since 11/2012. • We recently added two additional buildings this fall • Based on our predictions, we have saved $63k per month. • Our preheat model is expected to save an additional $40k per month. Figure 2: Space temperature plots for the west side of the 2nd to 8th floors of 345 Park. Our forecast (red) expects space temperature to remain within the contractual range (yellow/green band). Actual pace temperature (dark blue) stays within our 68% confidence interval (orange) and is followed by the NowCast (pink) for the following two hours. We display our recommendation for start-up (red square), preheat (pink triangle), and ramp-down (blue circle). Finally, we provide building managers with supply air temperature (light blue), temperature of air entering the zone, and their thermostat set points (gray). Supply temperatures are above the typical range (green band). Figure 2: Occupancy (left) and steam (right) predictions for the week of 11/18/2013. Historical data is in blue, while predictions and confidence intervals are in red and orange. Our steam plots also include our recommendations for start-up (blue circle), ramp-down (red square), and preheat (pink triangle).
Roger N. Anderson
added 8 research items
TPO Communications (TPOCOM) is the set of TPO components that enable messaging and interactions with the System Integration Facility (SIF) ; TPOCOM is a pair of independent subsystems referred to as the Sender and the Receiver. TPOCOM Sender reads TPO predictions generated by TPO analytics and sends/pushes them to the SIF Server. In turn, the SIF Server receives building sensor data from the respective properties, formats this data into a common SIF format, and then pushes the data back to the TPOCOM Receiver. Provided below in Figure 1 are the TPO process flow diagrams for the TPOCOM Sender and Receiver Modules. These diagrams display the internal functions performed supporting Send and Receive between TPO and the SIF server. Novel to TPOCOM are that the TPOCOM Sender is managed by the TPOCOM Manager, which schedules the Task Runner component to run periodically; the Task Runner performs the following ordered functions with the other respective TPO components: 1. Task Runner gets new/updated recommendations/predictions from the TPO database in which the TPO analytic processes have stored their most recent computed results (data points). 2. RowPoint conversion converts the data fetched from TPO database into 600 to 800 SIF format data points per hour to be used in reporting recommendations/predictions by TPO analytics. Task Runner interworks with the RowPoint Converter by which prediction data points are formatted and assembled; these data points represent degrees of confidence reported in a graph for the various temperature, energy, and occupancy visualizations predicted/recommended by TPO. 3. Task Runner then requests TPOCOM to push/send the data to the SIF Server. 4. Also, the Task Runner maintains a heartbeat handshake protocol with the SIF Server once a minute to keep the connection between SIF and TPOCOM alive. TPOCOM Sender uses a set of unique XML libraries to marshal TPO data into SIF data Format (see illustration below). Also novel to TPOCOM, the TPOCOM Receiver responds to Web Service callback events registered to act on receiving data from SIF. The TPOCOM Receiver calls the MSG Dispatcher to begin processing the incoming sensor point data received from the SIF; this processing entails: 1. Parsing incoming SIF points. 2. De-multiplexing the points based on their identifiers. 3. Converting them to TPO database format. 4. Using the Connection Manager to connect to appropriate TPO databases. 5. Writing the converted points to the TPO databases. The TPOCOM Receiver makes use of a set of unique XML libraries to parse and process the incoming data from the SIF.
et al. High-rise office buildings are by contract, obligated to provide services to the tenant companies that they house. The operation of these skyscrapers is of great importance, and is a fundamental requirement for the success of tenant companies and ultimately of the economies they occupy. The buildings occupy a very important business space, affecting the decisions of local government and power-grid companies, as well as providing necessary services of housing, security, power, and infrastructure to their tenants. Though the buildings provide a wide array of services to these tenants, one of the most operationally pertinent questions manifests in determining the operational hours of the building, distilled to startup (building turn-on) and ramp down (building ramp down from full operations to turn-off) times. Choosing an intelligent startup time for the building is very important: start too late, and operators will have to spike energy use and will struggle to meet contractual obligations of temperature settings within certain hours; start too soon, and needless energy and money will be wasted in powering a monstrous building needlessly. Similarly, choosing an intelligent ramp-down time presents unique and pertinent challenges. As our country moves into the big-data age, the opportunity to optimize the efficiency of these buildings and connect them seamlessly into an intelligent grid should be pursued with full vigor. The startup and ramp down recommendation generators of the Total Property Optimization (TPO) system both use advanced analytics to correlate 24-hour predictions and trends with recorded or calculated startup and ramp down times. Whereas deriving optimal startup and ramp down times requires access to a wide array of data, including values of HVAC control levers and current space temperatures, meaningful startup and ramp down recommendations must be made up to 24 hours ahead of time. The inability to meaningfully predict 24-hours expectations for all data in the building dataset makes this a fundamentally difficult question. This is where these TPO modules step in, allowing the mapping between an input space of 24-hour predictions and output "times" to be discovered. Each run of the startup and ramp down recommendation generators accepts as input temperature and electricity consumption data from the past 14 days, along with the predictions of these for tomorrow, and outputs the expected date and time of the next predicted startup and ramp down events, updated for 24 hours ahead. Problem: The building control system Total Property Optimizer (TPO) follows an operational philosophy of maintaining a robust and expressive human-in-the-loop system to utilize engineer/operator expertise in arriving at optimal building operation strategies.
Roger N. Anderson
added 13 research items
Introduction Ever since Nicholas Negroponte theorized about "smart and ready " real estate in his insightful and groundbreaking book BEING DIGITAL in 1995, technologists and real estate executives have been pursuing and researching what this concept actually means and what steps building owners need to take to make their properties intelligent and efficient. In fact, the question "What makes a building smart?" is still being debated today. Also in 1995, we at Rudin Management began the redevelopment of 55 Broad Street in New York City, considered by many as the grandfather of all smart buildings. For the tenants, we brought fiber to the desktop, supplied multiple choices of broadband carriers, provided power cleansed from the vagaries of the electrical grid, and added carrier neutral fiber to the risers. We also created a "hearth" and the "Digital Sandbox" where people could come together and share ideas. This year we celebrate the 20 th anniversary of 55 Broad's rebirth as a technology center and its incredible impact as the catalyst for the redevelopment and resurgence of lower Manhattan. This year, we also proudly unveil our latest tool that allows buildings to be even smarter and retain the intelligence they have acquired by accessing a new operating system that uses machine learning and data analytics to remember and learn from past experiences and then creates prescriptive and predictive pathways that guide our building operators toward optimal performance in real time. We have taken lessons learned from the development of 55 Broad Street, 3 Times Square, and 32 Avenue of the Americas, and introduced learned memory into the science of building operations. The name of this new operating system is DiBOSS, which stands for Digital Building Operating System. This new platform performs several important tasks. First, it integrates all building subsystems into a single cockpit. Second, machine learning algorithms empower our buildings to remember, learn, and using past building performance and weather forecasts, predict and automate daily building operations and occupancy levels. The result is enhanced energy efficiency, 24/7/365 re-commissioning, reduced hot and cold calls, and a new tenant retention tool contained within DiBOSS's "Tenant Fractal". In sum, DiBOSS collects, remembers, learns, expresses and shares the important data that most built environments dump every day. DiBOSS also allows building subsystems to speak to
Roger N. Anderson
added a research item
CALM Energy is a Power Partner to building owners, where we install our Energy Watchdog Controller Product with energy generation and storage, saving buildings 20-40% on energy costs while improving resiliency." We enable commercial buildings to: • Lower energy usage through adaptive control of set points and schedules for HVAC and lighting • Lower electric bill demand charges through automated peak-shaving and peak-shifting • Reduce building operations and maintenance costs through monitoring and notifications • Balance building electric demand with renewables such as solar and wind • Improve resiliency from electricity outages We charge a fixed monthly subscription for using our low cost, simple to install, on-site controller solution to monitor and control existing HVAC and lighting in commercial buildings, which is fully compatible with legacy systems. We sell this product both directly and through channel partners in the building management space, providing the transaction and learning based control expertise. We provide engineering, design, construction, and financing services for energy audits, generation and storage installation, and energy efficiency, all in concert with our Internet of Things (IoT) Controller technology. Market Opportunity: • Commercial Buildings of less than 100,000 square feet represent 98% of commercial buildings and 65% of the commercial floor space. More than 90% of these buildings do not have a building management system to control energy costs. • Many building integrators provide energy monitoring, a few provide control, but few have the capability to automatically and continuously reduce energy costs, providing a vast untapped opportunity to sell our IoT based building micro-grid solution with generation and storage. • Most medium-sized commercial building management cannot spend the time away from their core business to vigilantly manage energy usage, making this automated product enormously value to this market segment. Intellectual Property: Through millions of dollars in software research and development with Utilities, the U.S. Department of Energy, NYSERDA and other funding agencies, we have developed a deep software technology based capability to propel our patented business model. Our Energy Watchdog Product leverages our Smart grid patents and pending patents on energy efficiency, electric grid markets, and demand response. Business Model: CALM Energy's business model of enabling channel partners to cost effectively sell our product offering to provide automatic and continuous reductions in energy costs, while improving resiliency, in combination with their existing products (e.g. HVAC), provides an opportunity for exponential business growth. Our focus on highly efficient and simplified workflow software based IoT deployment that includes integrated full building energy management further differentiates us from typical IoT equipment component or building monitoring and automation providers. We forecast revenues of $5 million by end of year two (2020) and $34 Million by end of year four, in a $30 Billion market that's growing at close to 10%/yr, with an investment of $2 Million from a strategic investor.
Roger N. Anderson
added a research item
Finmeccanica Planet Inspired in the Di-Boss Partnership for Intelligent Buildings
Roger N. Anderson
added 2 research items
This Video was made by Intel to showcase their Internet of Things (IoT) contribution the Columbia, Rudin, Finnmecnica Di-Boss Smart Buildings product. See related Article: Anderson, R. N., Rudin Management, Selex, Columbia Engineering, Di-BOSS, Digital Building Operating System, 345 Park Avenue, NY, 2015.
Roger N. Anderson
added 3 research items
The disclosed Patent subject matter relates to techniques for improving the efficiency and reliability of the operation of buildings and/or collections of buildings held by a property owner by using Machine Learning and Artificial Intelligence.
OVERVIEW This technology is a building management system that collects and uses historical and real-time data to provide reliable predictions of energy consumption and accurate recommendations for efcient energy management. Unmet Need: Integrative building management system Building management systems are used to monitor and control energy demand from mechanical and electrical systems. With the increasing prevalence of smart buildings, there is a growing need for more sophisticated building management systems that can integrate data from multiple systems and sources, including energy consumption by various building operational systems, tenant occupancy patterns, and even external conditions such as weather. Improved techniques for building energy management can increase the energy efciency, resiliency and reliability of building operations and management while ensuring tenant comfort. The Technology: Predictive algorithm for managing building energy supply and demand This technology is a building management system that integrates building data from internal and external systems to predict energy supply and demand. A machine learning predictive model is used to generate energy demand forecasts and automated analysis that can guide optimization of building operations to improve tenant comfort while improving energy efciency. The system provides actionable recommendations that can help to reduce energy waste and increase cost efciency for building operations. Applications: Multi-unit residential buildings Commercial ofce buildings Energy consumption prediction Energy waste reduction Advantages: Integrates data from a wide range of sources and systems for more accurate prediction
Roger N. Anderson
added 3 research items
A high percentage of newly-constructed commercial office buildings experience energy consumption that exceeds specifications and system failures after being put into use. This problem is even worse for older buildings. We present a new approach, ‘predictive building energy optimization’, which uses machine learning (ML) and automated online evaluation of historical and real-time building data to improve efficiency and reliability of building operations without requiring large amounts of additional capital investment. Our ML approach uses a predictive model to generate accurate energy demand forecasts and automated analyses that can guide optimization of building operations. In parallel, an automated online evaluation system monitors efficiency at multiple stages in the system workflow and provides building operators with continuous feedback. We implemented a prototype of this application in a large commercial building in Manhattan. Our predictive machine learning model applies Support Vector Regression (SVR) to the building's historical energy use and temperature and wet-bulb humidity data from the building's interior and exterior in order to model performance for each day. This predictive model closely approximates actual energy usage values, with some seasonal and occupant-specific variability, and the dependence of the data on day-of-the-week makes the model easily applicable to different types of buildings with minimal adjustment. In parallel, an automated online evaluator monitors the building's internal and external conditions, control actions and the results of those actions. Intelligent real-time data quality analysis components quickly detect anomalies and automatically transmit feedback to building management, who can then take necessary preventive or corrective actions. Our experiments show that this evaluator is responsive and effective in further ensuring reliable and energy-efficient operation of building systems.
Roger N. Anderson
added 4 research items
As our society gains a better understanding of how humans have negatively impacted the environment, research related to reducing carbon emissions and overall energy consumption has become increasingly important. One of the simplest ways to reduce energy usage is by making current buildings less wasteful. By improving energy efficiency, this method of lowering our carbon footprint is particularly worthwhile because it actually reduces energy costs of operating the building, unlike many environmental initiatives that require large monetary investments. In order to improve the efficiency of the heating and air conditioning (HVAC) system of a Manhattan skyscraper, 345 Park Avenue, a predictive computer model was designed to forecast the amount of energy the building will consume. This model uses support vector machine (SVM), a method that builds a regression based purely on historical data of the building, requiring no knowledge of its size, heating and cooling methods, or any other physical properties. This pure dependence on historical data makes the model very easily applicable to different types of buildings with few model adjustments. The SVM model was built to predict a week of future energy usage based on past energy, temperature, and dew point temperature data. The predictive model closely approximated the actual values of energy usage for the spring and less closely for the winter. The prediction may be improved with additional historical data to help the model account for seasonal variability. This model is useful for creating a close approximation of future energy usage and predicting ways to diminish waste.
Approximate dynamic programming (ADP) driven adaptive stochastic control (ASC) for the Smart Grid holds the promise of providing the autonomous intelligence required to elevate the electric grid to efficiency and self-healing capabilities more comparable to the internet. To that end, we demonstrate the load and source control necessary to optimize management of distributed generation and storage within the Smart Grid.
Roger N. Anderson
added 9 research items
Boosting algorithms are provided for accelerated machine learning in the presence of misclassification noise. In an exemplary embodiment, a machine learning method having multiple learning stages is provided. Each learning stage may include partitioning examples into bins, choosing a base clas sifier for each bin, and assigning an example to a bin by counting the number of positive predictions previously made by the base classifier associated with the bin. (PDF) Systems and methods for martingale boosting in machine learning. Available from: https://www.researchgate.net/publication/259648838_Systems_and_methods_for_martingale_boosting_in_machine_learning [accessed May 06 2021].
Albert Boulanger
added an update
A technical brief on our IP “Predictive building management system for improved energy efficiency in smart buildings”  is now online and is available for licensing . http://innovation.columbia.edu/technologies/CU12084-a
 
Albert Boulanger
added a project goal
Research, develop and commercialize ML-Based Intelligent Building Management systems