Energy Storage and Optimising Usage
Over the last couple of years, I like many people, have begun to take more of an active interest in renewable energy and my own energy consumption. Having watched countless Fully Charged YouTube videos on renewable energy and EVs, I decided to look into getting solar panels and battery storage. After much deliberation, I decided to go with a UK company, Moixa who have their own 4.8 kWh smart battery and GridShare cloud-based software to control when the battery charges. Ever since it was installed, I have been regularly checking up on solar generation and battery charge level.
Having a battery installed alongside solar panels allows me to store any excess solar power not used at the time of generation, so I can use it later in the evening when the solar panels are not generating any energy.
The other thing the battery allows me to do is to make use of the time of use (TOU) electricity tariffs. These electricity tariffs are an evolution of the economy 7 tariffs that provide cheap off-peak electricity during the night. The latest incarnations have 3 or more prices for different time periods during the day. The most extreme is the Octopus Agile tariff which varies its price every 30 minutes which can be seen below for the 6th March 2020. Note how the price increases between 16:00 and 19:00, coinciding when people get back from school/work and make dinner.
These new tariffs reflect the variation in the wholesale electricity price market, from which energy companies buy their electricity. The variation in electricity price is due to two factors; variation in demand and variation in power generation. To keep things simple, most of the energy companies just charge one rate to their customers. As more of us get smart meters that are capable of recording energy use every 30 minutes, we can move on to these more flexible TOUs.
By being on a TOU tariff, customers will be incentivized to avoid using energy when it is expensive i.e. when supply does not meet demand. Getting customers to adjust their behaviour through TOU tariffs is one way in which the UK can rely on more variable renewable energy sources such as wind and solar.
By having a battery, I can take advantage of a TOU tariff only using energy from the battery during the period when the electricity price is highest. This relies on having the battery charged during those periods, even if its a cloudy day and the solar panels do not generate any energy.
Deciding when to charge the battery is an interesting challenge. When making the decision, I need to take into account:
my typical energy consumption
electricity prices for the next day
likely energy generation from my solar panels
Once I have considered all of these factors, I want to make the most optimal decision, i.e. the decision that reduces the amount I pay for my energy. As a data scientist, it's natural to turn to my machine learning toolbox; optimisation is after all at the heart of machine learning methods. Plus optimisation done with machine learning can be powerfully applied to this and many other practical problems.
So how I can make a machine learning solution to make optimal decisions for reducing my bills?
In follow-on posts I will cover:
Solar PV forecasts (both existing and how to go about generating my own)
How to model and predict my energy consumption and comparing to TOU tariffs
How to optimise battery charging, given TOU tariffs, Solar PV forecasts, and energy consumption using Bayesian Decision theory.