Bde sc3 2nd_workshop_2016_10_04_p10_maja_skrjanc

35
Energy efficiency - big data challenges from case studies Jozef Stefan Institute Maja Skrjanc [email protected] BDE 2 sd Workshop for Energy, Brussels 4/10/2016 16/6/2015 Company Logo

Transcript of Bde sc3 2nd_workshop_2016_10_04_p10_maja_skrjanc

Energy efficiency - big data challenges fromcase studies

Jozef Stefan Institute Maja Skrjanc [email protected]

BDE 2sd Workshop for Energy, Brussels 4/10/201616/6/2015

CompanyLogo

4-oct.-16www.big-data-europe.eu

Big data in energy:o Going green, Cutting back, Energy preservation

Energy efficiency case studies (NRG4Cast, SUNSEED):o Districts, buildings, households (monitor, analyze, test,

predict, optimize)

o Measurements (consumption, grid)

Outline

Big Data & Energy

One of the hottest topics today is energy:o consumption, discovery and implementation

o renewable, reusable and affordable energy, both at an individual and business level

Energy saving – standard of living (e.g. 2000W society):o right energy-efficiency measures, districts can reduce energy use and costs,

and shrink buildings’ environmental footprint.

4-oct.-16www.big-data-europe.eu

Energy perservation Cutting-back:

o energy consumption - monitored and improved, companies can improve efficiency and reduce expenditures.

Going green:

o real-time and batch processing analytical tools evaluate: current green strategies and

assess if those strategies are actually working and other areas that they can change to green

o With increasing penetration of Distributed Energy Resources (DER) the smart grid needs more & deeper monitoring and control to maintain stable operation

4-oct.-16www.big-data-europe.eu

Analysis of Environmental domain

Common challenges:

o Different data sources (structural data, sensor measurements, annotations)

o Loads od data (history, on-line sensor measurments, various prediction models, various forecasts, etc)

Modern technology available:o Amount of data is too large to be stored: new evidence from the incoming data is

incorporated into the model without storing the data

Sustainable energy management system

10/4/2016

6

NRG4Cast project NRG4Cast - real-time management, analytics and forecasting software pipe-line

for energy distribution networks :

o using information from network devices, energy demand and consumption, environmental data and energy prices data.

generic framework able to control, manage, analyze and predict behavior in an extensible manner on other energy networks:

o gas distribution, heat water distribution and alternative energy distribution networks.

10/4/2016

7

Current and Expected impact

Economic/Socialo Energy consumption savings up to 20%

o Dynamic energy tariffs – new jobs

o Lower energy bills for consumers up to 10%

o Saving in operational and maintain costs up to 15%

Environmentalo Reduced CO2 emission up to 20%

o Saves on energy production up to 10%

4-oct.-16www.big-data-europe.eu

Three pillars of NRG4Cast

Monitoring & Prediction

of Consumption

and Production

Monitoring & Prediction

of Consumption

and Production

Prediction of electricityprices

Prediction of electricityprices Textual pipelineTextual pipeline

10/4/2016

9

Prediction of various impacts on the energy networks (accurate models)

Prediction of energy production of Renewable energy sources

Data fusion and requirements synergy

Integrated pilot

10/4/2016

10

NRG4Cast scenarios

Multimodal Stream Data Analytics

10/4/2016

12

Textual pipeline

10/4/2016

13

Architecture

10/4/2016

14

Achievements I NRG4CAST Ltd

Final NRG4Cast Prototype (6 diverse pilots, 1 integrated pilot) – validation on mass instalation

Analytics:o Prediction and stream modelling pipeline – semi-automatic

o Route Cause Analysis (RCA) module – novel approach to understand complex multi-level multi-sensor system

o Framework for energy managements systems - MSDA (Multimodal Stream Data Analytics). Hybrid approach by combining knowledge-driven and data-driven elements

10/4/2016

15

Achievements II Data Access and Integration (DAI) platform (cca 800 data streams):

o DAI platform has evolved into a completely new system, that provides reliable accessto the pilot data at all times and is able to re-stream this data to other componentsin the NRG4CAST platform

Textual pillar:o Although the practical value of achievements in the field of textual data analysis has

not been significant, the NRG4CAST project proposed an innovative way to handlefact extraction from the textual stream

Numerous SW testings (different components, different maturity levels)

Stream modeling pipeline - integration of many different heterogeneous data sources

10/4/2016

16

Challenges Technical Challenges:

o Data integration: Integration of real-time and static data - design the schema for the metadata database

Integration of real-time data coming from hundreds of sensors (time-aligmenent)

Variety of data interfaces for multimodal data

o Stream modeling pipeline - integration of many different heterogeneous data sources

o HW installation

o How to reach TLR7 level of SW maturity

o Numerous SW testings (different SW components, different maturity levels)

o Defining appropriate features for prediction models10/4/2016

17

Lessons learned

Domain knowledge is the key (also in solving tech challenges)

Input from business perspective necessary to push and drive product development:o market analysis,

o bussines plans

Cyclic technical development (one prototype each year) turned out to be winning combination

Intensive dissemination activities are necessary

10/4/2016

18

SUNSEED project

enable end-user to actively participate in dynamic market

to allow an operator to have complete control over the smart grid

SUNSEED main objectives

Establish practical, converged DSO-telecom, secure communications

network

Develop advanced measurement &control sensor node

WAMS

Use intelligent analytical and visualisation tools to manage

smart distribution grid resources

Large scale field trial~ 1000 nodes

New business models of converged DSO-telecom

infrastructure

SUNSEED project - Motivation

Changing nature of the Consumers (households or industry) -> Prosumerso energy generators from renewable sources (photovoltaics, wind, cogeneration)

o manageable loads

Utilities are „blind“ in LV distribution grido real-time monitoring is needed

Motivation (cont.)

Manage risks related with network operation o voltage violations, congestions, …

Increasing hosting capacity of additional DER into existing grid without additional reinforcements

Offering new services for customers

More efficient network operation o increasing network observability, controllability and management

SUNSEED Architecture

SUNSEED Architecture

Com. solutions

Data flow

Delegated security management

Monitoring & Analytics & Control

State estimation of distribution smart grids

Forecasting

Prediction of failures

Active Network Management

State estimation of dist. smart grids Key enabler of advanced services

WLS with Gauss-Newton iteration scheme

Linear Bayesian estimation

Short Term Load Forecasting

Load forecasts - on various nodes of DSO in the grid (end users, transf. stations), for various forecasting horizons (1h – 24h).

Data sources - load measurements, load estimations, weather status and forecasts, static data (working hours, holidays, …)

Short term wind gener. forecasting

propose an efficient SVM based multi-stage forecasting technique incorporating pattern matching for data pre-processing.

Fault Detection in Telco’s data Spatio-temporal model

• To detect and localize potential faults in telco and DSO network

Outcomes• Usual methods (plotting upload and

download speed matrix over time, analysing histograms, probability distributions) do not show enough structure

• Multidimensional scaling embeddings shows more structure

Challenges

Various communication protocols

HW development

HW elements are expensive, communication as well

Minimal set of measurement nodes at locations to maintain whole grid observability

Integration of different security levels

Huge potential – where to start with monetarization ? (various stakeholders)

4-oct.-16www.big-data-europe.eu

Business models

Utility & telecom operator CO OP business models for communication nets in distribution smart grids

Summary

Wide range of opportunities:o Environmental data, Behaviour data (grid, consumers), Social & Economy

o Knowledge discovery (monitor, understand, predict, optimize)

o Business models

Technical challenges:o Multimodal data integration, Data models

o Maturity of SW components, integration, support & maintenance

4-oct.-16www.big-data-europe.eu

Thank you!

[email protected]

https://sunseed-fp7.eu/

http://www.nrg4cast.org/

4-oct.-16www.big-data-europe.eu