When designing the underground transmission line to connect to a high-load consumer, such as a data center, there are numerous factors to consider, starting with circuit configuration. Assuming the new data center will connect to an existing transmission line, there are two connection options: a radial tap or a loop-in, loop-out (LILO).

A radial tap is simply a parallel branch off of the existing circuit, much like how a distribution line connects to a house. This requires a single corridor for the underground installation, running from the existing transmission circuit(s) to the data center. This generally means a much lower installation cost and faster construction. It also has much smaller land requirements, which can be advantageous in populated areas. However, radial taps have much less built-in contingency, exposing the data center to potential outages from upstream transmission faults, substation failures or line outages. This outage susceptibility can be costly when downtime at the data center can result in millions of dollars lost every hour.

A LILO connection splits the existing transmission line into two lines and adds two new pieces: the loop into the data center and the loop back out to the transmission line. While a substation will generally be required at the data center to step down the transmission-level voltage to a usable voltage level, in a LILO configuration the substation also acts as the separating point between the two newly created line segments.

A LILO configuration has much more inherent contingency than a radial tap. Transmission lines can deliver power bidirectionally. In the event of a failure on one of the lines connected to the data center, power can still be delivered through the other line connection. In the event of a failure at the data center itself, power can be routed around the data center connection at its substation. LILO configurations are more expensive, require more construction and design time, and take up more space. It should also be noted that for a LILO configuration, the newly installed cable segments must be capable of carrying the full through-load of the transmission line (as well as any emergency ratings) in addition to the new data center load.

In some unique cases, both configurations can be utilized in a phased construction schedule. Since radial taps are easier and quicker to install, they can act as a steppingstone to get the data center running. Then a LILO configuration can be built, potentially using the radial tap infrastructure as one side of the eventual loop configuration. 

Once the installation configuration is decided upon, the next step is seeing that the cable system is capable of handling the expected load of the data center. Ampacity software such as CYMCAP, ELEK and Cableizer makes it possible to calculate the allowable current on an underground cable system, given various input parameters. The calculations are performed to make sure the cables are not continuously heated beyond their steady-state-rated operating temperatures: 90°C for cross-linked polyethylene (XLPE) insulation cable and 105°C for ethylene propylene rubber (EPR) insulation cable.

While an in-depth analysis of underground cable system ampacity is beyond this paper’s scope, there are a few parameters of significance for data centers.

Load Factor

The load factor is the ratio of the average load to the maximum load over a given period of time. For a typical consumer load, the load factor is usually under 1.0 (typically 0.75-0.85), meaning the cable system is not always loaded to the maximum expected demand. This is because power requirements typically cycle between low and high points day by day. However, data centers typically do not cycle in this manner. Instead, they stay at or near constant load 24/7.

This will have very different implications, depending on whether the installation is to be a radial tap or LILO. For a radial tap connection, the data center would be the only load on that branch path of the transmission line. Because of this, the average load and maximum load are equivalent to the data center’s average and maximum, making for a load factor of 1.0.

For a LILO connection, the new cables will be carrying the entirety of the transmission line’s expected load, not just the data center’s. The existing transmission line most likely has a load factor lower than 1.0, as it is supplying a variety of load types. Connecting a 1.0 load factor load to a combined load factor of less than 1.0 will raise the overall load factor of the system, but it will still be less than 1.0.

As an example, assume an existing transmission line has a daily load profile of 120 MVA for 8 hours, 180 MVA for 8 hours, and 200 MVA for 8 hours. Calculating the load factor:  

Adding a new 20 MVA load for the data center, with the data center having a 1.0 load factor, every hour of the day will have an additional 20 MVA. The peak load has also increased by 20 MVA.

So, the overall load factor of the LILO cable system is approximately 0.85, significantly lower than the 1.0 load factor that would be utilized for a radial tap. This distinction is critical and must be accounted for during planning and cable sizing calculations, as many underground cable systems are thermally limited. Small changes in load factor can significantly impact allowable ampacity and emergency loading capability. Of course, in this example the cable sizes would be vastly different for 20 MVA versus 186.7 MVA, but it does help show the effect that adding a new load can have on the existing load factor.

Transient Preload

Directly correlated to the load factor is the transient preload, typically expressed as a percentage of the total full load ampacity. It is defined as the load on the cable system immediately prior to a short- or long-term emergency condition. In the case of a 1.0 load factor already having been determined to be applicable, the preload must be set at 100% to match. This is highly important for radial tap configurations, due to the radial taps typically operating at a 1.0 load factor. For LILO configurations, a lower preload may be acceptable, depending on the existing line’s load factor and expected total load change with the data center load added. Care should be taken to determine an acceptable preload, as incorrect assumptions can lead to overestimation of emergency ratings, which is particularly critical for radial configurations where no alternate supply path exists. The most conservative option is to assume a 100% preload.

Continuous Loading

The continuously loaded nature of a data center can introduce other challenges, especially in a radial tap configuration when the cables are nearly always at 100% of their maximum load. In an underground cable system, heat generated by the cables dissipates into the surrounding concrete, backfill, soil and, eventually, the air. Each material has a unique thermal resistivity dry-out curve, which shows the relationship between the moisture content of the material and its thermal resistivity. A higher thermal resistivity indicates the material is worse at transferring heat away from the cables, thus reducing the cable ampacity. As the moisture content decreases in a material, the thermal resistivity typically increases.

With the cable steadily at 100% load, there is no time for the soil to recover and cool down. This can lead to soil dry-out, increasing the thermal resistance and making heat dissipation even more difficult. Engineered backfills with specific thermal resistivities can help mitigate this. Geotechnical surveys should be conducted during planning and design to determine if the existing soil is viable or if an engineered backfill is required. Similarly, the cables being at 100% load means the mutual heating between the cables is never given a chance to decrease, causing long-term steady-state ampacity to be lower than it otherwise could be.

The ampacity challenges introduced by continuous loading are most significant in radial tap configurations. They can be mitigated by properly sizing the cables to account for loading conditions so there is a larger margin between the cable’s maximum rated temperature and its actual temperature. If there are plans to use the radial tap as a steppingstone to a LILO configuration, it might be possible to use a larger-than-needed cable for the radial tap connection so the same cables can later be used in the LILO connection. While the overall capacity of the cables will need to be much greater in the LILO configuration to account for the new loads being added to the existing data center load, the continuous loading challenges will be mitigated when transferring from the tap to the LILO.

Of course, if the cable size is already at or near the maximum size (around 6,000 kcmil), cable upsizing is no longer an option. Alternative solutions include increasing the number of parallel cables per phase or increasing the voltage class. If the plan is to transfer the cable system from a radial tap to a LILO and use the same cables, ampacity calculations should be performed on both the radial tap installation and the LILO, adjusting the parameters and required ratings as needed.

Contingency Scenarios

Another important consideration for radial tap configurations is performance under an N-1 contingency scenario. In this context, N-1 refers to the loss of one supply circuit while maintaining service to the data center. To support this, multiple independent radial circuits might be installed, each capable of supplying the load.

If one circuit is taken out of service, the remaining circuit or circuits could be required to carry a portion or the entirety of the data center load. This condition can be highly thermally limiting, particularly given the continuous high-load nature of data centers. As a result, N-1 operation must be carefully considered during ampacity modeling, including appropriate assumptions for load sharing, transient preload and emergency ratings. This approach requires multiple independent source circuits to be available for interconnection, which may not always be feasible, depending on the existing transmission system.

Designing the cable system with an N-1 contingency case in mind can improve reliability for radial configurations, but it might still provide less operational flexibility than a LILO configuration, in which alternate supply paths are inherently available.

The overall system parameters and cable technology also play important roles in underground connection design. First and foremost is voltage class. When a specific MVA requirement must be met, the voltage class of the system will determine the current required in each phase of the transmission line. A lower voltage class results in more current, and vice versa. Since underground cable conductor sizes max out around 6,000 kcmil, if the cable system still cannot meet the requirements, more cables per phase may need to be installed. This means dramatically increased costs and installation size to accommodate the cables, which can cause routing problems.

If the new cable system is connecting to an existing high-voltage (69- to 161-kV) or extra high-voltage (230-kV and above) system, it would be ideal to wait until the installation is as close geographically as possible to the new data center to step down the voltage to a level the data center can handle. This is why the data center typically has its own on-site substation. By keeping the voltage class at a higher level for longer distances, the total installation width can be kept to a minimum for a greater distance. This is especially important in urban and suburban areas where space can be a limited resource. In more rural areas, larger cable system installations may be allowed because of greater available space.

One advantage of using a lower voltage class is that utilities often have stored distribution voltage–level cables ready to be used, potentially saving time and money in procurement.  

Powering high-load consumers such as data centers presents a unique set of challenges for underground transmission systems. Among these, the selection between radial tap and LILO configurations is a primary driver of overall system performance, influencing reliability, outage exposure and downstream design decisions.

While underground cable systems typically have much higher initial costs, they also typically have much lower long-term maintenance costs and higher reliability than overhead lines, which is a valuable quality for data centers.

Radial tap configurations offer advantages in cost and constructability but introduce greater risk due to a lack of inherent redundancy. LILO configurations, while more complex, provide improved reliability and are often better suited for high-consequence loads.

Once the system configuration is determined, careful consideration must be given to ampacity calculations and thermal performance. The near-constant loading typical of data centers reduces thermal recovery and limits emergency loading capability, making conservative design assumptions and accurate modeling essential.

Ultimately, successful underground transmission design for data centers requires a balancing of cost, constructability and reliability, with particular emphasis on system configuration and thermal constraints. As data center demand continues to grow, adapting traditional underground transmission design practices to account for continuous, high-load conditions will be essential for maintaining long-term system reliability.