building analytics guide for junior data analyst and scientist
This project is developing and written in weekend during my work experience in engineering startup.
Applying new technology in property is a challenge and misery. Low level worker don't trust you, high level manager want to do digital transformation in low cost for their achievement and have an instant success. They don't care what you did and how you did, only the result is what they looking for.
A typical example is: energy saving. During my research and development in the company, I design a more fair methodology for energy saving based on machine learning, top management level people still only looking for the electricity bills. Whatever how I tuned the AI model into 90+% accuracy, and finished back-test for the model validation. They keep asking why electricity bill can't show it out. As what i mentioned to them, AI model is helpful on providing multiple parameter correction, it is better than direct comparison or traiditional statistical modelling and is preferrable in ASHRAE guidelines/IPMVP for energy saving verification. Just an easy question, electricity is a summary of all you equipments consumption including the cooling load, outdoor temperature, performance changes. Under the global warming, the outdoor temperature shall be increased. Increasing outdoor temperature will reduce equipment efficiency (e.g. chiller efficiency affected by wet-bulb temperature). With increasing outdoor temperature, the cooling demand from building shall also be increased, more machine is required to open to provide equivalent cooling services. Linking different issues and causes, the major point that electricity bill cannot show saving comparing to last year is explainable. In a foreseeable future, only increment is occurs in electricity bills. Hope this point help junior data analyst/scientist to explain to top level management, advance the revolution of property industry.
Another thing is, AI is a portable database, it only able to do prediction when it is under its training history. We cannot provide accurate prediction without data, like forecasting the newly replaced equipment performance as each equipment have its own characteristics after deployment (water balancing, parallel or series connection etc.).
Engineering always is the hardest, valuable and professional industry. I believe that people who can master engineering can work well in another industry. As sensor data collection, network protocol, quality are the biggest challenge to PropTech industry. You have to know different, cross disciplinary knowledge if you want to be the giant in PropTech.
General Data Protocol:
- LoRa
- NB IoT
- Bluetooth
- WiFi
- UWB
- BACnet
- Modbus
- Lon
Common Data Quality:
- Incorrect data
- Duplicated data (High, Low level data)
- Insufficient documentation
- maintenance record
- data point mapping
- Sensor replacement schedule is long, so most of data maybe removed out
Common Data Analytics Issues:
- cross disciplinary knowledge
- water balancing
- hvac knowledge
- air balance
- IAQ
- computer science
- algorithm for optimization
- machine learning
- energy saving verification
Industrial Revolution 4.0 is not talking to make manufacturing leading again, it about improving the efficiency.
Seeing the entry requirement is high in PropTech, I would like to share all the knowledge that I countered in building data analysis and help junior data analyst/scientist to have a quick start in building data analytics.