Blog

Based on the technical architecture of my digital home I’m currently running 42 micro-services. Each micro-service has a special role in our daily lives. This article provides an overview of the use-cases that these micro-services enable.

I have multiple goals with these use-cases:

• Comfort: automating daily routines, especially the ones that are highly repetitive and require extensive steps;
• Reducing energy cost: we have a fully electric house (with the exception of the second car) and consume a lot of energy. Prices are dynamic and with planning our consumption we can reduce our energy cost;
• Reducing grid load: we have a fully electric house (with the exception of the second car) and consume a lot of energy. We try to consume energy outside of grid congestion hours (between 16.00 and 21.00) and take the availability of solar energy into account.

It's important to understand I'm using a price index to select the best moment of the day to consume energy. I calculate the price index based on the EPEX energy prices and the expected amount of solar energy.

1. Heating optimization: when winter comes I calculate the expected number of hours the heatpump requires to keep our home comfortable. Based on the energy price index the heatpump is switched on. Additionally, when a lot of sun is expected heating is disabled and the sun heates up our home during the morning, sometimes reaching a comfortable 21C when it’s freezing outside;
2. Hot water optimization: hot water generation is planned based on the current water temperature and the energy price index;
3. Ventilation optimization: we have six CO2 sensors in our home. The ventilation system is controlled based on the input of the CO2 sensors. The ventilation slows down when we are not at home to save energy in the ventilation system itself as well as heating losses;
4. Car charging planning: car charging is planned based on the planned time to leave the next day, battery charge state and the energy price index;
5. Appliance planning: the dishwasher, washer and dryer programs are planned based on the time of the day the program is being started and the energy price index;
6. Security alarm system: the security alarm system is automatically armed when we leave. Camera snapshots are send to our mobile devices when the virtual perimeter alarms are triggered, motion detectors are triggered, or when persons approach the front door;
7. Light control: lights are automatically controlled based on the time of day, position of the sun (season), motion detectors, and state of the security alarm system;
8. Solar screen control: solar screens are automatically controlled based on the weather predictions and inside temperature;
9. Beer/drinks fridge control: our outside fridge only turns on on Friday, Saturday and Sunday when temperatures reach over 20C and sun is expected;
10. Radio automated start: this was an annoyance: the radio is automatically started when the Chromecast comes online. No more apps to manually control the radio. In the weekend another radio channel is selected to switch from working to weekend mode;
11. Energy monitoring: 50 energy monitoring sensors ranging from the consolidate building energy consumption, freezer, fridge, lighting, car, heatpump, solar panels, TV and IT equipment.

I still have a short “wish” list:

• Automatically turn off solar energy production when energy prices are negative;
• When financially sound introduce a battery in the environment to further reduce energy consumption from the grid;
• Keep an eye on traffic speed on the main road next to our house;

In general I’m in an optimization phase. Most use-cases are up and running and require regular tweaking and fine-tuning. Sometimes optimization means taking 10 minutes to change an algorithm and push the micro-service into production. Sometimes it means two weeks of (evening) work to fundamentally change the architecture of a use-cases. It’s a hobby project, it keeps me from the street and challenges my technical skills.

2024 is over. Welcome to 2025!

In October 2024 I analyzed the data YTD: Minimizing grid consumption (2). In the days after NYE I had another look at the data and noticed some irregularities. The algorithm that calculated the direct consumption percentage didn’t match with the values registered by my energy meters. I updated the algorithm and recalculated the data from 2024 to make it consistent with the values registered by the energy meters.

The good news is that things look better compared to analysis I did in October:

I closed the year with a direct consumption of 34.1%. This is a huge leap forward from the 19% in 2023! However, unfortunately I did not achieve my goal (40%).

First things first: where is the improvement coming from? As I mentioned in the previous blog, I spend my time on improving two things: car charging and heatpump planning.

Car charging

In total the car consumed roughly 6400kWh in 2024, of which 1730kWh were produced by solar energy.

At this moment I’m planning the charging of the car based on an ON/OFF mechanism. The car is charged when most solar energy is expected. Then it will charge at 11kW (full speed). What I can still improve is to take the amount of expected solar energy into account, and adjust the charging speed accordingly. This will lengthen the charge and spread it out over the day, and probably result in a higher direct consumption, especially in the summer months.

Heatpump planning

The planning of the generation of domestic hot water was already operational for a longer time. I improved some small things in the algorithm, for example, setting the maximum power consumption of the boiler at disinfection runs depending on the availability of solar energy.

In total the heatpump consumed 2650kWh in 2024, of which 739kWh where produced by solar energy.

In the last months of the year I made some minor adjustments to the planning of heating the house. As the isolation of our house is modern I can turn off the heatpump for a couple of hours without noticeable impact on the inside temperature. I’m reducing the set temperature during the night and turn off the heatpump during the six hours the EPEX energy prices are at their peak. As the heatpump is consuming a lot of energy during the winter period, when there isn’t a lot of solar energy anyway, I don’t think this will actually contribute to my main goal.

Conclusion

Unfortunately I wasn’t able to achieve my target, but I made a good step in the right direction. I made some small adjustments to the goals, hoping I can still achieve 50% direct consumption without using a battery system. As the “salderingsregeling” will end in 2027, it should still give me enough time to evaluate if there is a need for a battery system before 2027 starts.

Last year, I described my view on home energy consumption and my personal goals here: minimizing grid consumption.

Today, the summer of 2024 has passed and we’re moving into the “dark” season. It’s time to look ahead towards the end of the year, look back at my previous goals and adjust them where necessary.

Before we do that let’s have another look at my ambitious goals for the next years:

Status quo

At the end of 2023 roughly 20% of the energy we consumed was directly consumed from our solar panels. Today, in the beginning of October 2024, our direct consumption of the year 2024 (year-to-date) is 28.4%.

The start of the year seemed good. However, in the beginning of the year the solar panels deliver less energy compared to the summer months, so the direct consumption is higher in the “dark” months. In the end, I’m far away from the 40% I was hoping to achieve this year.

Changes

This was not totally unexpected, but still a bit of a disappointment. Especially considering the changes I’ve worked on during the year:

Introduction of a price index

I was already monitoring the epex-spot prices and predicting the production of solar energy. I used both data points to schedule energy consumption, but the algorithm was (too) complex to maintain and wasn’t always showing optimal results. I’ve created a price-index that combines the epex-price and expected solar production into a single value.

Energy planning using the price index

To use the price-index I needed to adjust the scheduling process. This was a difficult change because appliance specific micro services where doing their own planning. This meant I needed to change the “controllers” for the HomeConnect appliances, heatpump and car. I took this as an opportunity to create a separate scheduling micro service. The controllers now request the scheduling service for the best time to start a appliance specific program. The scheduler calculates the best start time based on the expect program duration, consumption and time to complete. The image below visualizes the result of the new scheduler process. Most programs are scheduled during the daily solar peak, when the price index is at it lowest point. However, when I get home from work at the end of the afternoon, and need to start driving early morning again, the scheduler uses the price index to chose the best time to charge the car during the night.

Unplanned car charging

When I don’t know when I need the car I just connect it to the socket. The controller then considers the actual solar production and redirects this energy to the car. To ensure a stable charging process I needed to build in some delays. The curve is not perfect, but I think it’s close enough.

There is some discussion on the impact of the power quality of “smart charging”. I need to understand this a bit better and probably need to adjust my algorithm.

What’s going on?

Why am I stuck at 28%? To answer this question I need to look at the two largest consumers: the car and the heatpump.

Car

Let’s start with the car: the direct consumption of the car is 31%. It will be very difficult to further improve this, as it’s depending on a combination of my work schedule and the weather. The situation has improved since the beginning of the year. I’m expecting the year end to be difficult as the solar production will go down in the last months of the year. I might be able to find small optimizations in the scheduler over the next months.

Heatpump

The heatpump is even more complex as it uses two energy sources (the compressor and the COP1 heater) and has three different goals: heating, cooling, and domestic hot water. For this analysis I have only looked at the difference between the compressor and the COP1 heater.

The digital twin of the heatpump was already operational at the end of 2023. It gives me full control of when domestic hot water, disinfection, heating and cooling are enabled.

I’ve started to work on a couple of things:

• Detailed planning of the generation of domestic hot water;
• Detailed planning of the disinfection process (to prevent bacteria);
• Reducing the operation hours of the heating and cooling process.

The domestic hot water is now only generated during the day, based on the price index. The scheduler does not wait until the hot water is “empty”, but starts generating hot water if it expects a sunny day to generate a buffer for the day after.

The disinfection is typically planned once a week. I needed to add a bit of slack and, depending on the predicted availability of solar energy, it now schedules the best time to run a disinfection run between 6-9 days after the last run. Additionally, the power of the disinfection is adjusted based on the maximum available solar energy. If a maximum of 3kW of solar energy is expected, the disinfection power is adjusted down from 9kW to 3kW, it’s lowest value. The disadvantage is that disinfection takes longer, but that has not proven to create new problems.

Last winter, the heatpump was running the entire night to heat the house. Then, on sunny days, the sun started heating up our living room through our glass “walls” resulting in an inside temperature of 25 degrees Celsius. The energy consumed by the heatpump in the night and morning was, more or less, wasted. The controller now looks at the predicted temperature and solar radiation of the upcoming day and only turns on the heatpump if a combination of a low temperature and low sun radiation is expected. Cooling is only enabled if the inside temperature reaches a configured threshold. This mechanism should prevent wasted energy, but is not optimized for generating heat a the best time of day.

This has resulted in 20% direct consumption, which is far from the target.

The largest problem is the heating in the winter. There is no solar energy produced so the direct consumption is low.

When I only look at the direct consumption over the last 90 days (7 Juli – 7 October) the situation looks a lot better.

Conclusion

I don’t think the 40% direct consumption target is realistic for 2024. I probably need another year to come to 40%.

Over the next months I need to start understanding in which situations the heatpump is consuming energy from the grid. Additionally, I will slowly start looking into the consumption and planning of the WTW (ventilation) unit, fridge, and freezer. At the same time I need to prevent the risk of food poisoning by (control) software failure.