DeepMind, an AlphaGo development company that saves 1% electricity in 28, uses AI to open a power-saving mode for Britain.

DeepMind (the company that developed AlphaGo), a British artificial intelligence company acquired by Google's parent company Alphabet in 214, is currently negotiating cooperation with National Grid, which is in charge of power transmission and distribution in the UK, hoping to help reduce electricity consumption in the UK through neural network and machine learning technology without increasing other infrastructure.

? ▲ ? Developing AI technology of AlphaGo can not only play Go, but also be used for more livelihood-related purposes: saving electricity? Although it is still in a very early stage of discussion, about this cooperation project, DeepMind's co-founder and CEO (who is also the main game designer of Bullfrog's classic game "Theme Park") think that as long as it is optimized through AI, it can save 1% of electricity for the UK every year. In 214, the UK will produce about 33 TWh of electricity, and the cost will be tens of billions of pounds, so if it can really save 1 billion pounds every year. ? Electricity in Britain is a model of separation between power plants and power grids, that is, power plants and power distribution networks are under the responsibility of different companies, and the task of balancing power demand and supply still falls on the State Grid, which is responsible for power distribution. The power demand can actually be predicted, such as judging by standard human behavior patterns. For example, when people are awake, the power demand must be higher than when they are asleep, or when the weather changes, it gets cold or hot, we can predict the rise and fall of power demand first. However, when it comes to energy supply, it is relatively unstable and unpredictable. In particular, Britain relies more on renewable energy (green energy) such as wind and solar energy. Last Christmas, the proportion of green electricity supply in Britain exceeded 4% in a single day, which became an important milestone of renewable energy. However, the lack of power supply stability is also an old problem of renewable energy. How to maintain the imbalance between supply and demand is a big difficulty for the British national grid.

▲ ? If we want to replace nuclear energy and thermal power generation with renewable energy, we must face the problem of unstable energy supply. ? DeepMind believes that as long as learning technology through predictive machines can help the power system reduce the impact caused by external environmental shocks, for example, predicting the peak of power supply and electricity consumption through machine learning can optimize the use of renewable energy, that is, through their neural network-like technology, we can analyze all the millions of possible factors that may affect the power supply efficiency and find out the reasons and solutions that were not found by the British National Grid before. ? In fact, last year, DeepMind (after throwing out AlphaGo to sweep the world of Go) has already made a similar analysis for Google's data center, which needs a lot of refrigeration equipment, resulting in a huge power consumption, which can obviously save 15% of the power consumption of the data center. It is also possible to try it on the power grid of the whole country.

▲ ? Google data center needs a lot of refrigeration equipment to make the server run normally. If AI can save 15% of electricity, can it also have the same effect on the national power grid? ? Looking back at Taiwan Province, our annual power generation is about 2 billion kwh. If the cost per kwh is about 2 yuan, it will cost about 4 billion yuan a year. If 1% can be saved, it will save 4 billion Taiwan dollars. If possible, maybe Taipower should AlphaGo to DeepMind to help Taiwan Province's power network have a comprehensive physical examination, which may be better than always looking for the resources planning commission or the industrial research institute.