In this blog, I will show you three awesome ways to visualize electricity demand data for a substation, which supplies approx. 13000 consumers. The data consist of 4 years of measurements having a 15-min resolution. There were two types of measurements analyzed: transformer measurements and DG measurements aggregated at the level of a substation.
Daily Demand Profile Analysis
There has been an increasing trend of integrating various renewable distributed generation (DG) in the network, especially photo-voltaic plants (PVs). Therefore, due to the random behavior of DGs, it is crucial to separately analyze demand measurements, DGs and actual demand of consumers.
What do we mean with “actual demand”?
Actual demand in our case is a demand of consumers in the absence of a DGs and is calculated by subtracting DGs from substation measurements. This enables analyzing consumer behavior separately from DGs, which are highly dependent on solar irradiance (due to high integration of PVs in our case).
How to visually compare measured demand, DGs and actual demand?
In the graph below we compared it in three different ways. The first column refers to measured substation demand, middle column to DGs aggregated at the level of a substation and the right column to actual substation demand.
The first row on a graph shows daily load profiles. The second row shows histograms. By comparing the left and the right column we can see that DGs in general impact the middle range of the demand and have a little influence on peak values. This is an important fact since a network has to be constructed to withstand peak annual demand. The last row represents box plots for every 30-min interval. The bottom and the top of the box represent first and third quartiles, and the band inside the box is the median. Load profiles show, that there are greater load fluctuations during the middle of the day, due to high PV generation.
Annual Peak Demand Comparison
When analyzing demand data daily and annually peaks are usually important. One way to gain better insights and compare daily peaks between years is plotting swarm plot on top of box plots. Here every dot represents the daily peak, whereas circles are peak demand during working days and crosses are peak demand during weekends.
Why are these plots important?
These graphs are important because they show the density of daily peaks together with the statistical description. For example, on the graphs from 2013 to 2015, it can be seen that the annual peak loads are almost the same, but the density around peak values and median values are lower in 2014. This is due to the fact that the annual peak in 2014 occurred on December 30 and 31 before New Year’s Eve when there were very low temperatures and people are very active.
System operators usually observe only annual peak demand and make major conclusions about annual peak demand growth and consequently network reinforcements. By gaining better insights using these kinds of fancy plots, people are able to look at the data from different perspectives.
Using Heatmaps to Visualize Time Series Data
A very useful and easy way to visualize time series data is by using heatmaps. For example, if you have one year of data your heatmap could represent a calendar, where there are days in a month on the x-axis, months on the y-axis and square colors represent values that are of your interest. When observing electrical demand we are usually interested in daily peaks, therefore these are shown on the graph below.
It can be seen that daily peaks are higher in the winter time (especially in January & December) throughout the working days, due to heating and a significant share of industrial consumers, which mainly operate throughout the working days. Additionally, weekly patterns can be clearly seen (5 working days + weekend).
All the analyses were programmed in Python using Pandas, matplotlib and Seaborn libraries.
If you find this useful, please share this blog on LinkedIn and connect with me or join my group AI in Smart Grids on LinkedIn and connect with others working in this field.