4 simple steps to start your data-driven energy efficiency journey
Energy efficiency and data analytics. For years, people have been talking about how much value data doesn’t have. Big Data, AI, Data Science - all these buzzwords imply that there is value hidden in big data sets. And that’s true. But finding that value is easier said than done - especially for non-IT companies. But where should you get started? In this blog post, let’s take a closer look at how much we can learn from a single data point - the load profile. Using a real-world use case, we will show you how to understand an object’s behavior without even having to enter the object in question.
1. Get a first overview using monthly data
Let’s start with the basics. What is the load profile anyway? It is the historical record of the power demand of a plant - in terms of electrical energy. In Austria, energy network operators must provide load profiles for special consumers. A special consumer is any plant with an electricity demand of more than 100,000 kWh/y of electrical energy and 50 kW of the connected load. These criteria apply not only to large industrial plants, but also already to hotels, administrative buildings, and hospitals. This makes the use of this data source convenient. No internal employee has to search for specific data records within the plant from the outset.
What you see here is an example load profile. It’s the record of the power demand of our example factory for over a year.
In this first graph, we see a significant seasonal effect (higher power demands in the summer months). The effect leads to the assumption that cooling (i.e. in the form of HVAC) plays a huge role. Furthermore, there seems to be a significant change in the behavior of the system. The power demand before the summer months is smaller than that after summer. Interesting indeed. So let’s dig a bit deeper to find out what else is hidden in this load profile, shall we?
2. Look at the “average week” to understand usage
In another step, we want to understand what an average week looks like. To this end, we calculate the average power profile per day of the week and compare them. The result looks as follows.
On all days - including Saturday and Sunday - the power increases to reach the daily maximum value at 6:30 a.m.
The working days then have the identical course except for Friday, which ends a bit early. At 20:00 a larger plant (approx. 45 kW) is switched off, at 21:15 another consumer with 11 kW is switched off.
We also see this behavior on weekends. Weekends see much less work in general- so is it necessary for them to be switched on in the first place? If not, this would result in a saving of over 50,000 kWh per year.
This information helps experts to understand what happens within the building better and faster. Good visualizations and their interpretations help to get an overview of the situation in a very short time. And most importantly: they make complex matters comprehensible.
3. Plot an ordered load profile to understand baseload and get benchmarks
In addition, we would like to investigate the baseload behavior of our object. For this purpose, we use the ordered continuous load profile. Here, the hours are displayed in which a defined power demand was required.
We can see the baseload compared to the highest occurring peaks. This becomes especially interesting if we can compare more than one similar facility.
In this case, the baseload amounts to about 30% of the annual consumption. In other words: the activities during active periods of the year account for 70% of the electricity consumption. Using these metrics, we can compare these objects to similar ones. In this case, we are looking at an average Austrian hospital. Comparisons like this can, however, also lead to finding more useful solutions.
4. Use a heatmap to find outliers and find anomalies
Last, but not least, we want to introduce a particularly beautiful tool. Heatmaps can be used to understand a facility’s behavior in deep detail.
Continuing with our example, we can produce the following heatmap. Before analyzing it, we should try to understand the structure. On the horizontal axis of this heatmap, we see the hours of the day from 00:00 to 23:59. On the horizontal axis, starting from the top, we arrange all the days of the year starting from 01.01. all the way to 31.12. The colors on this heat map correspond to the power demands ranging from ~10kW (green) to ~90kW (red).
So- let’s have a closer look into what we can learn from this heat map.
- We see the seasonality again. The summer months look much different than the winter months.
- In the first half of the year, we see that the weekends need much less power than weekdays.
- There seems to be a significant spike at the beginning of the workday. It only occurs in the summer months, this has to do with cooling.
- Sometime in July, the working periods seem to have gotten longer.
On the basis of the electricity consumption analysis, we can find anomalies. Areas in which electricity is not used efficiently. We suspect that the building control system is not set properly and there is a great need for action here.
By talking to the facility manager in charge, we can now start to identify problems. Our visualizations guide us and provide a common understanding of the situation. As a result, the FM manager is in a position to set priorities, identify problems and create solutions.
Saving potentials of 10-20% are possible
In our case, we were able to work with the facility manager to find non-optimal settings in the control of the building services. The ventilation system was running 24/7, but the operation was only required between 7 a.m. and 5 p.m. and only from Monday to Friday. In addition, the fresh air temperature setpoint was set high in the winter. Changing this resulted in savings of over 15%.
Switching off the air conditioning on weekends, brought a further saving of over 6%. Also, the system was adjusted in such a way that it cycles less, which reduces the wear costs. A few smaller measures yielded another 2%.
In total, an electricity saving of over 23% was realized. This corresponds to approx. 75.000 kWh or approx. € 9,000 per year. In this case, no extra investment was made. Only the control system was adjusted. You don’t always find such high savings. From experience, we assume an average of 5 to 10%. Yet, the analysis offers further benefits that should not go unmentioned.
The combination of simplicity and power makes these analyses a great tool to get a first glance into the energetic behavior of a system. They are, however, just a first step to entering a data-driven cycle of continuous energy efficiency improvements.
For us, they give us an idea of where to look next. Which data to analyze next and what to look for. In this example, we would now start to check the control system to understand the seasonal changes better.
Another matter to look into is the large consumers that are switched off in the evenings. Their operation to both check whether they could be kept off on the weekends and improve their operation in the next step.
If these investigations are interesting to you, stay tuned. We will cover them in another blog post in the future.
Load profile analysis are coming to nista.io
We understand, that the visualizations presented here might not be built easily by everyone. This is why we decided to partner up with our friends at e7 and bring power profile analysis to nista.io - AI powered energy management . So if you are interested in understanding the heartbeat of your facility better, you should definitely get access!