Across sectors, many companies are seeing total energy costs reaching unprecedently high values. There is no exception in the telecom business. In this challenge, a well-known company sought to reduce the annual energy costs of mobile network equipment by taking a data-driven approach to their energy-related decisions.
To provide a good and stable mobile network service for their customers, this company has over two thousand sites (or antennas) spread throughout the country, each with an independent tariff contract. Additionally, each site has batteries to secure a stable service during power loss events. These batteries can also reduce energy costs through charging and discharging periods selection.
Facing this scenario, the main goal of this project was to develop an optimization tool capable of reducing energy costs by simultaneously optimizing tariffs and battery usage while ensuring the site’s availability.
The challenge was targeted with a new approach to tariff planning, switching from a manual process, poorly leveraged by data insights, to a systematic, data-driven method based on each site’s consumption profile and available tariffs.
To contextualize how tariffs are applied, it’s essential to understand the fixed and variable terms. The fixed term must be paid monthly without exception, while the variable component depends on the site’s consumption. A tariff might define different hourly variable costs, establishing periods in which energy has a higher price and others where the power is cheaper. To illustrate, the figure below shows the consumption of a given site during a day and the different variable costs: the “Off-peak” hours, the cheapest ones, and “Extra-peak” hours, the most expensive ones.
As expected, the consumption is lower during the night and higher during the day – due to our routines. It is also interesting to verify that the variable costs tend to be higher during periods with high demand since it is easy to predict the hours when most people will require a connection. The difference between variable costs during the day allows for determining the best tariff for each site by selecting the one that reflects fewer overall costs according to the site’s consumption profile.
Battery usage is also a critical lever of the approach. A site’s battery usage policy can allow charging in the cheapest periods and discharge in the expensive ones, benefiting from the difference in the tariff. The developed approach leverages existing power loss history and consumption patterns to determine the minimum capacity that should be maintained to guarantee each site’s required level of service. Historical data was used to assess the criticality of each site, protecting the most crucial ones and taking advantage of the most robust ones. For each site, the safety threshold of the battery was determined to optimize this trade-off. An example of battery usage can be seen in the figure below.
In the above example, the battery of a particular site starts at 60% capacity. It fully charges during the less expensive periods until it is used to replace the power grid during the “Extra-peak” hours. The 60% threshold corresponds to the safety threshold defined by the approach considering the site’s criticality level for the mobile network grid.
By dynamically determining each site’s battery usage, the overall energy costs are reduced while increasing the networks’ protection against electrical power losses.
The new data-driven process substantially reduced the effort of determining the best tariff and battery usage conditions, decisions once made manually for each site. An optimization tool allows testing different scenarios, opening doors for “what-if” scenarios that help the planning team to evaluate their strategies quickly. Their operations are now more scalable and easier to endure when adding new tariffs or sites to their grid.
To test the approach, a comparison was made between the yearly expenditure with the current tariffs and battery usage policy and the annual costs of the new dynamically determined ones. The suggested recommendations directly reduced energy costs by up to 7 percentage points, significantly exploring the potential of batteries.
By: Luís Guimarães , Horácio Neri , Luís Bulhosa