How To Answer The ENG235 AIoT ECA (PowerGrid SG)

Quick answer: The ENG235 End-of-Course Assessment rewards one clear AIoT use case that you carry through all eight questions, backed by a layered architecture, the 8Vs, an IT versus OT security comparison, an ethics section and a RoI method that goes beyond profit. Pick the use case first, then let every answer point back to it.

If you are staring at the PowerGrid SG brief and wondering where to start, this guide gives you a mark map, a working order and the mistakes that cost marks. The wording of the task is on our ENG235 ECA question page.

Stuck on the engineering side? Our electrical and electronic engineering tutors can talk through your design choices.

What the brief asks

PowerGrid SG is a fictional smart grid and utilities programme run by Singapore’s grid operator and national water agency. It has old SCADA, separate electricity and water metering and siloed sensors. Pressures include data-centre and EV charging peaks, intermittent solar, water loss, ageing underground assets, meter privacy, pricing fairness and cyber risk. You are acting as a consultant who must recommend how AIoT can help, within local law.

The assessment is worth 70% of the course and 100 marks, in APA style. That is a big document, so plan the marks before you write.

Mark map for Q1 to Q8

Matching word share to marks is a simple rule, and many students ignore it. Q3 and Q5 each carry as much as Q1, so they need real depth.

Step 1: Choose one use case and keep it

The case lists several objectives: demand forecasting, fault and leak detection, DER and EV integration, faster dispatch and ethical data use. Do not try to answer all of them as separate projects. Choose one that touches the most of them.

AI load forecasting feeding demand response is a good candidate. It uses meter and sensor data, needs edge and cloud, raises privacy questions, creates security exposure and has measurable savings. Predictive maintenance or leak detection also works if you can show the same chain.

Then reuse it. In Q3, draw the architecture for that use case. In Q4, name the data sources and models it needs. In Q5, attack it. In Q8, price it. Markers can see that the report holds together.

Step 2: Answer each question in a working order

Q1 and Q2: problem, value, fit

State the operational problem in one paragraph, show the IoT data and AI model that address it, then split value into public and commercial. For Q2, tie your choice to the energy reset pillar of the Singapore Green Plan 2030, which covers cleaner energy, solar and EV charging (Singapore Green Plan, n.d.). Keep national claims modest and sourced.

Q3: architecture

Use layers: sensors and devices, communications, edge or fog, cloud, applications. For each, say what it does for your use case. Then add a short note on latency (what must run locally), scalability and reliability (what happens when a link fails). A labelled diagram helps, with a table as backup.

Q4: data and the 8Vs

The 8Vs are volume, velocity, variety, veracity, value, variability, validity and visualisation. Do not just define them. Give a grid example for each in a line: volume is millions of meter readings; velocity is streaming telemetry; variety is meters, weather and logs; veracity is a drifting sensor; and so on. Then walk through ingestion, storage, processing and analytics, and name the models.

Need more help with this part? Our big data analytics coursework page shows what we cover.

Q5: IT versus OT

The comparison is usually where marks are won or lost. IT puts confidentiality first; OT puts safety and availability first. OT assets last decades, are patched rarely, and cannot simply be shut down after an incident. Then trace one attack path from an IT phishing email through a remote vendor connection to a control command. Cite recognised guidance such as NIST SP 800-82 Rev. 3 (Stouffer et al., 2023) and Singapore’s OT cybersecurity masterplan (Cyber Security Agency of Singapore, 2019).

Q6: ethics

Cover three things and give a control for each. Granular meter data can reveal daily routines (surveillance), so minimise and aggregate it, in line with PDPA duties (Personal Data Protection Commission, n.d.). Models and prices can burden some households more than others (bias), so test by group. Decisions need owners, logs and human override (accountability). Federated learning is one way to keep raw data local (McMahan et al., 2017).

For a worked example of how these answers sit together, see our ENG235 PowerGrid SG sample extract, for reference only.

Q7: business model

Think in business model canvas blocks, but keep it short: value proposition, customer segments, revenue or funding, key resources, cost structure. The twist is suitability. A national grid is critical infrastructure, so pure pay-per-use or venture-style models rarely fit. Discuss regulated tariffs, public funding and service fees, and why reliability outweighs growth.

Q8: RoI

Set out a method: baseline, costs, benefits, time horizon, discount rate, sensitivity. Include financial benefits (avoided network upgrades, lower balancing costs) and wider ones (fewer outage minutes, emissions avoided). If you show numbers, label them illustrative. Add non-financial indicators such as forecast error and peak reduction.

Checklist before you submit

• One use case named in the first page and appearing in Q3, Q4, Q5 and Q8.
• Each question answered under its own heading, with the marks in mind.
• An architecture figure or table with latency, scale and reliability noted.
• An IT versus OT comparison, not just a threat list.
• Controls for each ethical risk, not only the risk.
• Assumptions in RoI labelled and tested.
• APA 7 in-text citations that match the reference list.

Common mistakes

• Listing technologies without a purpose. “Use 5G, blockchain and digital twins” earns little unless each solves a named problem.
• Treating the 8Vs as a glossary. Tie each V to your data.
• Generic cybersecurity. Talking about passwords and firewalls without OT constraints misses the point.
• RoI as one number. Show the method and the uncertainty.
• Stating Singapore facts from memory. Check agency pages before quoting policy or targets.

If the structure feels clear but the writing time does not, our SUSS assignment support and report writing guidance pages explain the options. You can see what past students say on our student reviews page.

FAQs

Do I need a diagram for the architecture question?

The brief asks for a high-level reference architecture, so a labelled diagram is the natural format. Add a short table or paragraph so the marker can see how each layer supports your use case.

Can I use more than one use case?

Q1 asks for one high-impact use case. You can mention others as later phases, but keep the analysis on one. Depth beats breadth here.

How do I explain the 8Vs without padding?

Give each V one sentence of definition and one sentence of grid example. A compact table does this well and saves words.

How much should I say about OT protocols?

Enough to show you understand why OT differs: long lifespans, limited patching, availability first, and protocols designed before security was a concern. You do not need packet-level detail.

Are RoI numbers allowed if the case gives no figures?

Yes, if you label them as assumptions or illustrative and explain how you would replace them with real data. Never present invented numbers as facts.

Which Singapore sources are safe to cite?

Use official pages: the Green Plan site, the Cyber Security Agency, and the Personal Data Protection Commission. Quote only what the page actually says.

Want to see who writes and reviews engineering content here? Read about Wei Zhao, who holds a Master’s degree in Electrical Engineering from the National University of Singapore, or browse the SUSS question archive. For network and wireless design points, see our telecommunication topic page, and learn more about us.

References

Cyber Security Agency of Singapore. (2019). Singapore’s Operational Technology Cybersecurity Masterplan 2019. https://www.csa.gov.sg/resources/publications/singapore-s-operational-technology-cybersecurity-masterplan-2019/

McMahan, B., Moore, E., Ramage, D., Hampson, S., & Aguera y Arcas, B. (2017). Communication-efficient learning of deep networks from decentralized data. Proceedings of Machine Learning Research, 54, 1273–1282. https://proceedings.mlr.press/v54/mcmahan17a.html

Personal Data Protection Commission. (n.d.). About us. https://pdpc.gov.sg/Who-We-Are/About-Us

Singapore Green Plan. (n.d.). Our vision. https://www.greenplan.gov.sg/vision/

Stouffer, K., Pease, M., Tang, C., Zimmerman, T., Pillitteri, V., Lightman, S., Hahn, A., Saravia, S., Sherule, A., & Thompson, M. (2023). Guide to operational technology (OT) security (NIST Special Publication 800-82 Rev. 3). National Institute of Standards and Technology. https://csrc.nist.gov/pubs/sp/800/82/r3/final