tech

AI-based smart energy system developed

2 Comments
By Shinichi Kato, Nikkei BP CleanTech Institute

Hazama Ando Corp has developed a new smart energy system using an AI (artificial intelligence)-based energy management system (EMS).

The system, "AHSES (Adjusting to Human Smart Energy System)," was developed with help from MirrorLife Corp, Saitech Inc, Aval Nagasaki Corp and FirstLily, which deal with technologies related to the system. The new technology is expected to become one of the technologies that realize "ZEB (Net Zero Energy Building)."

The system enables to make full use of electricity generated by a solar power plant, whose output fluctuates depending on weather conditions. It efficiently and effectively supplies electricity to a building while enabling to reduce the capacity of a storage battery to be introduced.

In addition to a solar power generation system and a lithium-ion (Li-ion) rechargeable battery, it consists of (1) software that predicts power demand and makes an optimal operation plan, (2) a display that visualizes energy management, etc.

Based on data on the use of the building and weather, it predicts power demand by using machine learning and mathematical methods and supplies power from the solar power generation system and the Li-ion battery at optimal timings. The solar power generation system and the battery are connected by using direct-current electricity to smoothly control charge/discharge.

Because the control can be conducted by minutes, the system can support "dynamic pricing" (a method to adjust demand by changing price in accordance with supply-demand balance), which power companies plan to introduce in the future.

In a time of emergency, as a stand-alone power source, it can supply power to important facilities such as servers. Also, it is possible to flexibly expand/downsize the system in accordance with the scale of the building.

The development of the system started in April 2013. After testing each machine/device, it was introduced to Hazama Ando Technical Research Institute, and the entire system started to be tested. As a result, Hazama Ando confirmed the peak cut adjustment effect for electric power load and the function as an emergency power source with the prediction of power demand and the optimal operation plan.

The company also plans to add a co-generation system, aiming to optimally supply electricity and heat. And it intends to apply the new system to a smart grid that helps exchange electricity between buildings.

© Japan Today

©2019 GPlusMedia Inc.

2 Comments
Login to comment

This will definitely perform worse than expected. I see a lot of PREDICTION and OPTIMAL being thrown here, which are red flags for anyone who knows about this stuff. Co-generation, dynamic pricing, WOW! Sign me up!

Unfortunately, you know... making such bold claims is only going to lead to a huge let-down when the system cocks up. And it will. I think the way that the wind is blowing, this firm should back way off the claims and emphasize some real benefits to users. They might be miniscule or nonexistent. They might require a huge initial investment with benefits that are risky because they rely on perfect operating conditions, perfect prediction, and perfect optimization.

Ugh. We are not there yet, people, and don't let Elon Musk tell you different. Prediction itself is still imperfect. You might MIGHT predict the weather an hour in advance, but predicting cloud cover over my buiilding an hour in advance is usually impossible. Optmization is a unicorn. If you are using data updated hourly, your decisions for the next hour might put you at a disadvantage for the hour after that. And the more precise your models are, the more expensive they are. Is the system cost-beneficial? Can you rely on the system vendor to tell you that honestly?

I recommend SIMPLE systems for end users and progressively more complicated systems for larger scale units. The most sophisticated systems should be used for huge smart grids, not any smaller than communities of 1000 or more building units. Such large numbers can bring benefits from prediction and optimization, the the variance-risk of those predictions will be smaller.

Pretty sure I could write an AI algorithm for my own home. It would be better than ANY program someone else could make, and it would still suck pretty badly. But for a community of 5000 people? It might not only be possible, but worthwhile to make a decent program.

-1 ( +0 / -1 )

I think speedracer does not understand the reliability of big data and the ability to predict conditions over large areas. We do not care about his building, we care about the area being served as a whole.

1 ( +1 / -0 )

Login to leave a comment

Facebook users

Use your Facebook account to login or register with JapanToday. By doing so, you will also receive an email inviting you to receive our news alerts.

Facebook Connect

Login with your JapanToday account

User registration

Articles, Offers & Useful Resources

A mix of what's trending on our other sites