Energy Talk Series II| Energy & Digitalization: Artificial Intelligence Application in Energy Sector
Uploaded by Prakarsa Jaringan Cerdas Indonesia | 19th January 2021 Script Writer : Miftahus Salam
Editor : Nisma Islami Maharani
In Energy Sector 4.0, implementation of Artificial Intelligence (AI), Big Data, Internet of Things and Cloud Computing is unavoidable. People needs clean, affordable, and uninterrupted energy with better service. To provide such requirements, Energy Sector needs to embrace:
1. The Transition to a Clean Energy Future
2. A More Digital and Distributed Grid
3. Individualized Customer Services
However, Energy Sector is facing myriad challenges, such as:
- Economic Downturn
- Reliability and Resilience
- Access and Affordability
- Human Resource
- Rising Energy Demand
- Cyber Security
- Forecast and Planning
- Extreme Weather
- Upgrade & Maintenance
- Electric Vehicle
Energy Sector 4.0 needs to shift its paradigm i.e.:
Product Driven to Solution Driven
Impersonal to Personalized
Reactive to Predictive
Human to Human-and-machine
Next, Energy Sector 4.0 needs to shift its technology to: Big Data, Cloud & Edge, Internet
of Things and Artificial Intelligence. AI is expected to enable Energy Company to:
Reduce Capital Cost
Lower Fuel Cost
Some examples (of myriad possibilities) of AI Applications in Energy Sector, e.g.:
- Forecast and planning - Non-Linear Control
- Equipment Inspection - Demand Load Prediction
- Predictive Maintenance - Weather Prediction
- Energy Management - Intelligent Thermostat
Some important examples of AI Application in Electric Utility, e.g:
Better Resource Management, to predict demand and reserve resources anticipation.
Digital Twin. Digital twin is simulation model of actual system in virtual system to achieve recommended decisions using what if scenarios. In digital twin, it can do many what-if-simulations to make prediction without sacrifice the equipment and capital.
Predictive Maintenance. Predict the failure and conduct maintenance in advanced. This prediction output is optimum time when the machine or its part need to be replaced or maintained.
Grid/Transmission System :
Power Transmission Inspection. The algorithm processes digital photographs captured by helicopter or drone, automatically identifying electricity poles, its components, and specific defects in such components (rust, broken protections, etc.).
Vegetation Risk Assessment. AI can help to predict the risk from vegetation in distribution. So it can minimize the risk.
Anomaly Detection. To cover all areas, we need machine to detect the anomaly to prevent failure.
Monitoring and Coordinating the charging of EV to keep the grid stable.
Improve forecast by evaluating systematically the large amount of data in electricity trading.
Improve precision measurement and efficiency
Managing Efficient reserve margin to prevent power black-out
Optimize utilization to make efficiency
Evaluating and analyzing the flood of data from the large number of decentralized power grids and keep the grid in balance.
Better means to stabilize the power grid by detecting anomalies in generation, consumption, or transmission in near real time, and then develop suitable solution.
Better control system can solve quality and congestion issues.
Usually, collecting data and building the infrastructure will be the hardest parts. Typical AI Application Development will include 5 inter-related activities:
Defining Targets and KPI’s
Optimize AI Algorithm (Deep Learning)
Distribution System :
In a Smart Grid system:
Evaluating, analyzing, and managing the data lake from AMI (Advanced Metering Infrastructure) that connected to various grid participant (consumers, producers, storage, facilities) connected to each other via the grid.
Predictive variable renewable energy resources
Estimate Demand pattern
Analyze load data characterization
Precision predictive system overload
Improve forecast for grid stability and supply security.
For End Customer:
Helps consumer to managing their cost by analyzing user preference (time, windows, prices, usage pattern).
Do’s & Don’ts in implementation of AI
Depend too much on traditional planning process
Wait for a superstar
Expect AI project to work the first time
Expect too much on machine learning engineer
AI was built by algorithm to process empirical data. This infrastructure called machine learning. Machine learning algorithms use computational methods to “learn” information directly from data without relying on a predetermined equation as a model. The algorithms adaptively improve their performance as the number of samples available for learning increases.
To develop machine learning, AI do deep learning. Deep learning is a machine learning technique. It teaches a computer to filter inputs through layers to learn how to predict and classify information. Observations can be in the form of images, text, or sound. The inspiration for deep learning is the way that the human brain filters information.
In the first step AI run the training phase. During this phase AI model is learning to control the portofolio. It explores different behavior trying to find the optimal control strategy. After it find the correct control strategy, its behavior becomes less erratic. After that phase, as the AI defines its strategy, the actual and expected signals almost perfectly overlap. AI is a hungry brain. AI need to be fed with SCADA, online sensors, offline database, and big data. On the 10th August, Indonesia Government will launch Indonesia National Strategy Framework for AI. This national strategy framework has 5 priority areas, i.e :
Education and Research
AI implementation in electricity/energy sector will include in education and research, and smart city areas. Artificial Intelligence has power to:
Learn from massive data of amount
Process multi-variable input
Execute predictions in near-realtime
Compute on the edge and cloud
Model complex non-linear process
From type of data, AI divided into two types. Unstructured data (picture, video, sound, social media or big data) and structured data (data in data base). In general application, AI was used to computer vision (video tracking, image classification, object recognition, image segmentation, and object detection) and natural language (search engine, text classification, language translation, and audio recognition).
In general, machine learning model include 8
Artificial Neural Network
Extreme Machine Learning
Support Vector Machine
Wavelet Neural Network
Adaptive Neuro Fuzzy Inference System
AI related with IoT, and it must be related with cyber sector. With the development of AI it will appear new challenges in cyber sector, i.e:
Human Resource Problem
Privacy & Data Protection
Other Emerging Technologies
Artificial Intelligence is an opportunity to improve productivity, efficiency, and competitiveness. To get maximize improvement needed well plan and strategy.
Visit us :