12
1 Energy AI Solutions RENEWABLE ASSET MANAGEMENT Minimize O&M cost and generation loss with unmanned monitoring

RENEWABLE ASSET - Amazon Web Services

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

1

Energy AI Solutions

RENEWABLE ASSET MANAGEMENT

Minimize O&M cost and generation loss

with unmanned monitoring

Solar’s hidden secret: the data

• A 1GW fossil or coal-based power plant on average will generate

approximately 10,000 data streams.

• A same sized wind farm might produce 51,000, 5 times as many.

• In case of solar farm, it makes 436,000, 44 times as many.

Large fossil plants and wind farms are typically built around a

relatively small number of very large assets, while solar farms are built

around a large number of very small assets.

e.g.,

• Wind farm has 88 turbines with a capacity of 141 MW.

• By contrast, Solar project generates 62MW, or less than half as

much, but features 144,000 solar panels.

• Not only will each panel generate information about power

production, temperature and other parameters, but also inverters,

trackers, junction boxes will produce continual streams about their

current state or possible problems.Renewable Energy World

By Michael Kanellos and Steve Hanawalt | 11.27.19

Solar Photovoltaic, which generates an overwhelming amount of data than other sources, is relatively effective in reducing

labor costs associated with operation and maintenance. Other power sources often need to make important judgments with

a small number of data, so human intervention is essential.

2

BACKGROUND

In case of solar PV, each decision made with a large amount of data would not have a critical effect,

which makes it suitable for utilizing Artificial Intelligence technology.

3

The State of Digital O&M for the Solar MarketPrice pressure in solar O&M is driving the adoption of digital solutions

How much solar capacity can one O & M technician manage? As Is 20 MW/person To Be 40~60 MW/person : 2 Times More Efficient

By using AI based on Big Data, labor productivity in O&M can be more than double by minimizing human intervention in

judgment pursuing unmanned monitoring. Reliable data collection and automated management minimizes work time on site

as well as unplanned field work by proactive maintenance

Data integration

Time-Averaged

dataAll Data

Value dataonly

Real-timedata

BACKGROUND

Wood Mackenzie

The increasing role of digital in the solar PV O&M space / 10.10.19

Our Solution : i-DERMS

4

“ Artificial Intelligence Powered

Integrated Distributed Energy Resources

Management System “

Data

Acquisition

AI

Analysis

Subscription

Actionable

Insight

Increase Yield

SE

RV

ICE

PR

OC

ES

S

Full CompatibilityHighly Accurate

Diagnostics

Human Interface Immediate Alerts

Frictionless Process Scalability

FE

AT

UR

ES

Based on the forecasting, alarming, optimization algorithms, we build a customized AI monitoring solution that

learns customers' O&M processes to minimize human intervention and error, improves O&M efficiency, and makes more yield.

We pursue that AI system monitors PV generation instead of human and the O&M manager relies on our alarming system only.

Time

Accuracy

MACHINE LEARNING & DEEP LEARNING

Learn and Customize

INTRODUCTION

Encored’s AI algorithms

5

A. Generation Forecasting Algorithms

We developed algorithms of various forecasting models. Furthermore we have outstanding deep learning algorithm to

select the best algorithm for a certain PV site repeatedly to achieve better accuracy of forecasting.

CORE TECHNOLOGIES

Training Model Selection

Prediction

Data

+

Store the best model

SUPPORTED MODELS

• statsmodels

• Ordinary Least Square method

• Quantile regression method

• Generalized Additive Model

• Holt-winters

• sklearn

• Support Vector Machine algorithm

• Kernel Ridge

• Deep learning

• Long Short-Term Model

• ETC

• Prophet

• Multivariate Adaptive Regression Splines

Encored’s AI algorithms

6

B. Alarming Algorithms

• Run fault diagnosis and generation forecasting algorithms by

inverter or string, or group of separate sensing device installed

(by each data unit)

• Self-learning abnormality (Fault) diagnosis on individual

inverter or string using customized AI algorithm based on

judging process of each customer.

• Each monitoring unit, i.e. inverter or string has its own

diagnosis respectively instead of applying a certain amount for

all set of monitoring unit.

Set Individual inverters or string-based abnormality criteria automatically, Accurate and quick alerts.

Anomaly decision algorithm based on the distribution of its deviation

Runs AI-Algorithms

by eachData set

Self-learnsFault

Diagnosis Threshold

AlertsActionItem

CORE TECHNOLOGIES

• Anomaly Detection using error in actual measurement and

prediction of Solar PV generation

• Notify about 99% error range after checking the

distribution of forecast error

• Anomaly notification is delivered according to the

time unit, daily unit setting

Period C : ABNORMAL

7

C. Optimization Algorithm Load - Solar PV - ESS Data Combined Algorithm Maximizing profit and life of battery

Encored’s AI algorithms

Machine learning-based algorithm that derives schedule 24 hours a day before the day starts

which minimizes the operation cost of individual site or ESS equipment,

away from the repetitive control according to the existing simple time unit scheduling

8-10% MORE PROFIT in average than using existing EMS

CORE TECHNOLOGIES

OUR PUBLICATION Kim, et al. (2019). Practical Operation Strategies for Energy Storage System under Uncertainty, Energies, 12(6), p.1098.

OPTIMIZATION ENGINE

• Goal is to minimize the value function that is the sum of

Sum of battery wear cost

Sum of electricity price

With a peak demand constraints

Problem formulation Solution

• Calculation of value function via state transition

By using the Bellman’s equation (backward induction)

Improving a penalty term to control the peak

8

Key References

Korea Hydro & Nuclear Power (PV 27MW, ESS 6MWh)

Korea East-West Power (PV 41MW, ESS 170MWh)

State of Hawaii (PV 500KW, ESS 750kWh)

The ESS charge / discharge optimization algorithm has successfully

completed a pilot project. Encored has applied to new and existing

renewable facilities and plans to further build an integrated control

system in November 2019.

The project, Development of Renewable Facilities Integrated

Management System, is now in progress using Encored’s energy

platform and AI algorithms for optimization.

AI-based Micro-grid operation now in progress, with Hawaiian island

specific data, which is process for getting ready on commercialization.

CUSTOMERS

We help utilities & governmental bodies resolve problems derived from the increment of distributed energy resources and

the necessity of demand control. i-DERMS is a solution which is very useful to manage multiple resources.

Phase 1: Jan. 2018 ~ Mar. 2019 / Phase 2: Nov. 2019 ~ Jun. 2020

Oct. 2019 ~ Aug. 2020 (10 months)

Nov. 2018 ~ Apr. 2021 (30 months)

9

PROJECT. Korea East-West Power

Key Performance

CASE STUDY

PROJECT. Korea Hydro & Nuclear Power

• Needs to have AI powered operation to maximize benefits

• Simulation result by i-DERMS optimizing algorithm is

to have more profit than traditional operation.

• Set up i-DERMS to operate 17 of PV+ESS sites

totaling 41 MW PV and 170 MWh ESS capacity.

• Needs to have AI powered monitoring solution

to manage 15 sites or more with 2 persons.

• Plan to increase its solar plant capacity

from 28MW in 2019 to 5.4GW in 2030.

• Target to minimize increment of O&M resources

with more PV capacity applying more AI algorithm

pursuing “Unmanned Monitoring”

Minimize the cost on O&M human resources Maximize benefit from Optimizing the Load-PV-ESS

(Unit : $1K)

Category Year0 Year+10 Year+20 Remarks

TraditionalOperation

1,569.78 7,726.91 10,422.77 Site A(ESS Only)

ESS 15MWhPCS 3MW

EnerTalkAI-Algorithm

1,836.74 8,845.77 11,690.51

Additional Profit

266.96(17%)

1118.86(14.5%)

1,267.74 (12.2%)

Category Year0 Year+10 Year+20 Remarks

TraditionalOperation

1,569.78 7,726.91 10,422.77 Site B(ESS Only)

ESS 15MWhPCS 3MW

EnerTalkAI-Algorithm

1,828.26 8,809.72 11,648.39

Additional Profit

258.48 (16.5%)

1,082.81(14%)

1,225.62(11.8%)

Existing Solutions

2 persons, 15 sites

28 MW 5.4 GW

yr2019 yr2030

N

Unmanned Monitoring

“”

Save O&M Manpower into Half Amount Using i-DERMS

10

Encored, Inc.

COMPANY OVERVIEW

We are US based company doing “Energy AI” located in San Jose, California and was established in

2013. Encored is a startup funded by George Soros and SoftBank, leading next-generation

technology. We are offering innovative solutions in the energy industry by using AI and Big-Data

algorithms. Our mission is to develop a purpose-driven connection from people to energy data by

creating a network between DERs and consumers.

Hyoseop LeeCSO

Bell Lab

Univ. of Wyoming

Ph. D. Seoul National Univ.

John ChoeCEO & Founder, Encored

President, LS IS

IEC-ACTAD International Expert

Busan Univ.

Jin LeeCMO

Intel

Ph.D. Stanford Univ.

Seoul National Univ.

KyoungIl ShinCTO

Choirock Contents Factory

Nexon

Seoul National Univ.

Investors

Frank HowleyVice President

Head of University-Industry

Foundation, UC Santa Cruz

Seyong LeeVice President

Hyosung

Younsei Univ.

Seonjeong LeeSenior Data Scientist

National Institute for Math Science

Ph.D. Seoul National Univ.

Our Team

Quantum Strategic Partners

11

I United States I3031 Tisch Way, 110 PlazaWest

San Jose, CA United States 95128

I South Korea I8F KTS Bldg. 215 Bongeunsa-ro

Gangnam-gu, Seoul, Korea 06109

I Japan I

27F, Shiodome Sumitomo Bldg. 1-9-2

Higashi-Shimbashi, Minato-ku, Tokyo, Japan

Encored, Inc.

Office

Business Area

AI-based

DER

Management

Energy

Data

Service

Residential

Demand

Response

Smart

Home &

Family Care

H.Q.

Microgrid

COMPANY OVERVIEW

For more information about

Encored’s Solar PV Monitoring and Management Solutions

Please send us email [email protected]

Encored, Inc.

3031 Tisch Way, 110 PlazaWest

San Jose, CA

United States, 95128

Encored Technologies, Inc.

8F KTS Bldg. 215 Bongeunsa-ro

Gangnam-gu, Seoul

South Korea, 06109

www.encoredtech.com

12