AI & IoT in the development of smart cities
Download as PPTX, PDF3 likes3,161 views
Smart cities utilize information and communication technologies to improve economic and social well-being while reducing environmental impact. Internet of things (IoT) technologies allow cities to become smart through applications like smart grids, waste management, traffic management, and load forecasting. Artificial intelligence and deep learning techniques can help with load forecasting and optimizing these smart city applications through analyzing real-time sensor data from areas like energy use, transportation, and infrastructure monitoring. The document provides examples of how cities like Barcelona, London, and Singapore are successfully implementing IoT and AI strategies to address challenges from urbanization and improve services.
AI & IoT in the development of smart cities
- 1. AI & IOT IN THE DEVELOPMENT OF SMART CITIES RAUNAK MUNDADA THURSDAY, APRIL 14, 2016
- 2. OUTLINE • What is a smart city? What is IoT? • Need for smart cities • Where does IoT come into picture? • Applications that make a city “smart” • Smart Grid • Waste and water management • Traffic management • Load forecasting: Details • What is load forecasting? • Artificial neural networks and deep learning • Conclusion
- 3. SMART CITY & IOT (INTERNET OF EVERYTHING)
- 4. SMART CITY • Urban development vision to integrate multiple information and communication technology (ICT) to create sustainable economic development and high quality of life in a secure fashion • Major Areas – • Smart Energy Management • Smart Healthcare • Smart Services • Smart Logistics • Smart Waste Management
- 5. TODAY’S SMART CITIES Barcelona Amsterdam India Singapore London San Francisco
- 6. NEED FOR SMART CITY • Urbanization – world’s urban population expected to double by 2050 • Necessity to develop smart and sustainable cities to cope population growth • Environmental challenges – Technology for efficient use of energy • 70% CO2 emissions derive from cities • 80% of global energy production consumed by cities • Economic growth – biggest contributor to a country’s GDP • By 2025, 600 biggest cities projected to accounted for 60% of global
- 7. SMART CITY KEY COMPONENTS SMART CITY Technology Factors - Wireless infrastructure; IoT; Network equipment; Computing infrastructure; Human Factors – Education; Creativity; a smart community Government Factors– Support of government and policy makers
- 9. INTERNET OF THINGS(IOT) • Network of physical objects connected to the internet • Collect and exchange data Smart wearables Smart Sports Smart Health Smart Homes
- 10. APPLICATIONS THAT MAKE A CITY “SMART”
- 11. SMART GRID • Goal • Electrical grid including operational measurement systems which relays information back to a central management system • IoT • Digital sensors monitoring the transmission networks (power meters, voltage sensors, fault detectors) • AI • Optimization of power transmission; promote conservation through demand-based pricing • Predictive maintenance of networks and load forecasting • Predict cascading effect of outages in a grid network • United Kingdom – OpenADR (Open Automated Demand Response) standard reduced peak usage in commercial buildings by 45% • Amsterdam – Smart street lighting allows municipalities to control brightness of street lights based on traffic and pedestrian movement
- 12. WATER MANAGEMENT • Goal • Conserve water (as simple as that) • IoT • Sensors to monitor water supply, water levels and sewerages • Real-time climate monitoring • AI • Predict/Optimize usage of water • Optimized rain-water harvesting • Predictive maintenance – avoid disruptions • Distributed sensor networks to plan for flooding; learning from the network data • Abu Dhabi - reduced annual maintenance plan by 40% using IBM’s solution • Barcelona (Spain) - The irrigation system in Parc del Centre de Poblenou, transmits real time data to gardening crews about the level of water required for the plants
- 13. WASTE MANAGEMENT • Goal • Optimize trash collection and monitoring • IoT • Sensors monitoring trash collection • Sensors identifying objects at disposal grounds • AI • Optimize trash collection by monitoring waste levels at public bins • Segregate recyclable materials from the non-recyclable ones • Philadelphia – ‘Big Belly’s’ trash can allowed the city to bring down their trash collection frequency from 7 per week to 5 per week and avoid overflowing trash cans
- 14. PUBLIC CARE • Goal • Efficient and quick health diagnosis • Avoid epidemics • IoT • Environmental monitoring • Various sensors to collect patient’s real-time data • AI • Provide diagnosis support by predicting possible medical outcomes • Remote diagnosis of patients • Based on GIS data, environmental data and patient data, predict possible outbreaks of epidemics • Samsung’s Smart healthcare solution – Mobile health camps with state-of- the-art in-vitro diagnostic (IVD) devices for testing for metabolism, blood cells, hormone and cardiovascular conditions at pilgrimage places like the Mecca
- 15. SAFE CITY • Goal • Ensure safety of citizens and prevent crime/attacks • IoT • Surveillance cameras • Monitoring citizens movements • AI • Predict crime locations and type of crimes • Alert citizens remotely • Optimize the presence of police personnel • Researchers from MIT trained a computer to learn from millions of images to identify possible crime locations • San Francisco – BART police using historical data to predict and prevent crimes
- 16. SMART TRAFFIC MANAGEMENT • Goal • Make life better on the road • IoT • Sensors capturing traffic movement, street light sensors, sensors at tolling booths, sensors installed in public transport systems • AI • Smart parking – finds a parking spot for you • Direct traffic through alternate routes; open-up reserved lanes; Automated traffic lights • Optimize public transport; Reduce fuel consumption in public transports • Helsinki (Finland) - fuel consumption has dropped 5 percent, customer satisfaction has increased by 7 percent, driver performance has improved, and mechanical maintenance has become proactive by analyzing data from sensors installed in their public buses • Barcelona (Spain) – Designed new bus network based on data analysis of most common traffic flows • London – Underground subway uses IoT for predictive maintenance, monitor on-going activity across the system and for infrastructure planning (https://youtu.be/NYpdNGl1hco)
- 17. LOAD FORECASTING
- 18. KEY FACTS AND TERMS • Electrical energy cannot be stored as it should be generated as soon as it is demanded • Accurate forecasts lead to • Savings in operating and maintenance costs • Increased reliability in power supply and delivery system • Intelligent decision-making for future development • React correctly and quickly to fluctuations in the supply of electricity from renewable energy sources • Three types of forecasts • Short term – 1 day to 1 week ahead prediction; used to schedule generation and transmission of electricity • Medium term – 1 week to 1 year ahead prediction; near term unit commitment decisions • Long term – Predictions beyond 1 year; develop power supply and delivery system
- 19. SHORT TERM ELECTRIC LOAD FORECASTING (STELF) • Aim is to predict future electricity demands based on historical data • Electricity pattern affected by • Time (day of the week, holiday, time of the day) • Environmental factors (temperature; wind and heat in case of renewable energy; big source of randomness) • Social factors (individual usage pattern; big source of randomness) • Economical factors • IoT devices • A network of smart meters • Real-time voltage monitoring • Sensors monitoring the environment
- 20. TECHNIQUES Traditionally - Time series models like ARMA Regression techniques Modern Techniques - Artificial Neural Networks Support Vector Regression Genetic Algorithms Deep Learning Essentially, all models are wrong but some are useful. - George E.P.Box
- 21. ANN – ARTIFICIAL NEURAL NETWORK • Based on the central nervous system • Output from ANN is - • FeedForward Backpropogation ANN (Supervised) • Input data to be learned and desired output for each data sample • Optimize weights so that the error is minimized • Restricted Boltzmann machines (Unsupervised) • Nonlinear feature learner based on a probabilistic model • Tries to maximize the likelihood of the data using a graphical model (bipartite graph)
- 23. DEEP LEARNING AND LOAD FORECASTING Input • Electrical load factors • Social Factors • Environmental Factors Deep Learning Architecture • Stack Autoencoders • Stacked RBM’s Regression Model • Linear Regression • Support Vector Regression • Bayesian Regression Output • Load Forecast (based on the required time period)
- 24. CONCLUSION • In case of a smart city, IoT makes the AI possible • Successful implementation could lead to an improved and sustainable life • Ability to solve a lot of problems faced by the world today • AI perspective – Neural networks and deep learning architecture seem to be one of the best ways to move forward • Amount of data collected is huge and varied • Deep Learning architecture brings us closer to the true goal of AI
- 25. RESOURCES • Ensemble deep learning for regression and time series - http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7015739&tag=1 • The urban internet of things - http://datasmart.ash.harvard.edu/news/article/the-urban-internet-of-things-727 • London underground IoT - http://www.fastcompany.com/3030367/the-london-underground-has-its-own-internet-of-things • Barcelona Smart City - http://smartcity.bcn.cat/en • What is a Smart City - http://bit.ly/1UZZBMA • Green Capacity Smart City - http://greencapacity.ru/information/smart-cities • Conceptualizing smart cities with dimensions of technology, people and institutions – http://inta- aivn.org/images/cc/Urbanism/background%20documents/dgo_2011_smartcity.pdf • OpenADR - https://en.wikipedia.org/wiki/Open_Automated_Demand_Response • Samsung Village - http://www.samsungvillage.com/blog/2015/04/06/samsung-suggests-smart-healthcare-solutions-to-people-in-the- middle-east • Waste Management “Big Belly” - http://bigbelly.com/places/cities/ • MIT Safe neighborhood - http://io9.gizmodo.com/this-ai-can-tell-which-neighborhood-is-safe-better-than-1639451437 • IBM Case study for water management - http://public.dhe.ibm.com/common/ssi/ecm/ti/en/tic14208usen/TIC14208USEN.PDF • Singapore Case Study - http://bit.ly/1SFDdSl
- 26. THANK YOU!
Editor's Notes
- #5: What is a Smart City - http://bit.ly/1UZZBMA
- #7: http://greencapacity.ru/information/smart-cities - statistics
- #8: Reference - http://inta-aivn.org/images/cc/Urbanism/background%20documents/dgo_2011_smartcity.pdf
- #12: OpenADR - https://en.wikipedia.org/wiki/Open_Automated_Demand_Response http://energy.gov/oe/services/technology-development/smart-grid “Smart grid” generally refers to a class of technology people are using to bring utility electricity delivery systems into the 21st century, using computer-based remote control and automation. Historically, workers sent out to collect data for electricity requirements, broken equipment, read meters etc. A key feature of the smart grid is automation technology that lets the utility adjust and control each individual device or millions of devices from a central location.
- #13: Singapore Case Study - http://bit.ly/1SFDdSl IBM Case study - http://ibm.co/25QiSDy
- #14: https://youtu.be/8e8Be9rq_C8 http://bigbelly.com/places/cities/
- #15: http://www.samsungvillage.com/blog/2015/04/06/samsung-suggests-smart-healthcare-solutions-to-people-in-the-middle-east/
- #16: MIT Research - http://io9.gizmodo.com/this-ai-can-tell-which-neighborhood-is-safe-better-than-1639451437
- #17: Barcelona - http://smartcity.bcn.cat/en London - http://www.fastcompany.com/3030367/the-london-underground-has-its-own-internet-of-things
- #22: Ensemble deep learning for regression and time series - http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7015739&tag=1 Geoffrey Hinton – RBN for deep learning
- #23: Deep Learning could be used for feature learning