Direct Current Architectures to Eliminate Power Conversion Losses

22 concepts16 papers1 proposal

The foundational architecture of the modern electrical grid was built on a compromise. In the early days of electrification, Alternating Current (AC) won the "war of the currents" because it could be easily stepped up to high voltages for long-distance transmission and stepped down safely for consumer use via simple transformers.

Today, however, the nature of both power generation and consumption has fundamentally flipped. Our most critical energy assets—both on the generation side (photovoltaic solar arrays) and the consumption side (data centers, electric vehicles, and smart home electronics)—operate natively on Direct Current (DC).

Forcing this modern ecosystem to interface with a legacy AC grid creates two severe vectors of energy waste: transmission energy loss across centralized infrastructure, and power conversion loss across redundant distribution stages. A growing body of research focuses on bypassing the legacy AC grid entirely by deploying native Direct Current architectures and localized microgrids.

The Multi-Stage Conversion Tax

To understand why modern power distribution is inefficient, we must look at the continuous conversion tax paid by high-performance computing environments and modern households.

Inside an exascale data center, standard efficiency metrics often obscure a massive drain: the continuous conversion of AC grid power to the DC power required by silicon components. As detailed by Wesley Brewer et al., these conversion overheads can easily exceed a megawatt of pure waste in large-scale facilities. This represents a massive operational cost that goes unaccounted for in standard software-centric metrics.

Why is this loss so high? Traditional architectures rely on a chain of redundant conversions. In a typical data center power delivery path, power transitions from AC grid voltage, to DC to charge Uninterruptible Power Supply (UPS) backup batteries, back to AC for internal distribution, and finally back to DC at the server motherboard. Mengshu Sun et al. point out that centralized UPS systems suffer from efficiency drops between 15% and 30% due to these exact stages. This penalty is further compounded by the physical distance between the centralized energy storage units and the actual point of load at the server level.

Traditional AC Data Center Path:
[AC Grid] ──> [DC (UPS Battery)] ──> [AC Internal Distribution] ──> [DC Server Load]
               ▲                  ▲                                 ▲
               └─ Conversion 1 ───┴────────── Conversion 2 ─────────┘

This problem scales down to our homes. Modern households are packed with consumer electronics that natively require DC power. Manja Görner et al. highlight that traditional smart home systems incur continuous, silent energy penalties because every single appliance must embed its own independent AC-to-DC converter. These small internal converters are notoriously inefficient, radiating heat and driving up aggregate electricity demand.

Direct DC Coupling: Bypassing the Inefficiencies

The most elegant way to solve a conversion loss is to eliminate the conversion entirely. This is the core intuition behind Direct DC Coupling.

Consider green hydrogen production via solar energy. In a standard setup, a photovoltaic (PV) array generates DC electricity, an inverter converts it to AC for the grid, and a rectifier converts it back to DC to power the electrolyzer. Yan Ren demonstrates that by directly coupling the PV arrays to the electrolyzers, the entire intermediate AC-DC conversion apparatus can be bypassed. This architectural simplification directly minimizes energy conversion losses and strips out costly, complex power electronics interfaces.

Direct DC Coupling Path:
[PV Array (DC)] ─────────────────────────────────────────────────> [Electrolyzer (DC)]
                               (Zero Conversion Stages)

At a system-wide scale, this concept manifests as high-voltage DC (HVDC) power distribution. Instead of stepping up and down through AC transformers, facilities can distribute power natively at high DC voltages. As Kejian Cao et al. demonstrate, high-voltage DC distribution systems eliminate the redundant AC-to-DC and DC-to-AC transitions within the UPS chain entirely. By delivering high-voltage DC directly to the racks, energy loss as heat drops dramatically, which simultaneously alleviates the cooling load required to keep the facility stable.

Similarly, at the residential scale, Manja Görner et al. show that implementing a dedicated DC low-voltage net directly powers home appliances through specialized plugs, bypassing device-level conversion losses entirely.

Localized Microgrids and the Transmission Barrier

Eliminating conversion losses goes hand-in-hand with eliminating transmission losses. Centralized grid infrastructures face steep economic and logistical barriers when trying to extend transmission lines to geographically isolated or rural areas, as noted by Baoying Li et al. and Amam Hossain Bagdadee et al.. In remote regions, resource-limited areas, or scattered island environments, traditional grid expansion is physically or financially impractical, making localized autonomy a necessity (N. M. A. Hadia et al.; Lingfei Qi et al.).

N. K. Anand explains that decentralized power generation at the point of use completely bypasses the need for expensive long-distance transmission lines. However, localized generation introduces a new problem: balancing supply and demand without a massive centralized grid to absorb shocks.

The solution is the deployment of interconnected DC microgrids. By keeping energy distribution local and native to DC, communities can share power across short distances with minimal friction. Zhichuan Huang et al. propose a DC distribution network specifically designed for inter-home energy sharing. Because the local transfers happen entirely within a DC framework, the system avoids the heavy AC-DC conversion penalties that occur when homes try to trade power back and forth through a standard AC utility line.

Algorithmic Coordination of Decentralized Networks

If we remove the centralized grid infrastructure, how do we prevent localized networks from collapsing under volatile renewable generation? The answer lies in decentralized, peer-to-peer coordination algorithms.

Relying on a distant, centralized grid to balance local energy fluctuations introduces massive transmission overhead, as highlighted by Shangyu Xie et al.. Instead, modern architectures use hierarchical and algorithmic management to balance the load locally. For instance, Jing Li et al. demonstrate a decentralized dynamic power management framework where localized grid assets optimize their power flows through neighbor-to-neighbor communication. This completely eliminates the need for a vulnerable central controller while significantly reducing total power losses.

To scale this up, researchers look toward multi-tiered optimization. Fuxing Zhang et al. outline a hierarchical framework that optimizes power distribution across multiple interconnected Local Energy Networks (LENs), balancing volatile renewable generation and storage across local, neighborhood, and wide-area scales.

Hierarchical Optimization Flow:
[Wide-Area Scale] ──> [Neighborhood Scale] ──> [Local Energy Network (LEN)]

Similarly, Hao Wang 0016 et al. propose a decentralized bargaining-based market that allows interconnected microgrids to cooperatively trade energy without a central authority, achieving up to a 22% cost reduction compared to operating in isolation. By combining native DC hardware architectures with these intelligent, decentralized control layers, modern power networks can route energy directly from source to load with minimal conversion or transmission waste.

Key Foundational Work

Implementation of User-Oriented Smart Services into an Innovative DC Low Voltage Net (2015) — contributed 4 concept elements to this direction.
Research and practice of energy saving and consumption reducing in Data center (2015) — contributed 2 concept elements to this direction.
Defining and solving the energy allocation problem with continuous prosumers (2015) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_power_conversion_loss_d621707d,
  title = {power conversion loss},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/power-conversion-loss/rdr_d621707d22044a6e924af25301e6a1be},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_power_conversion_loss_d621707d_v1,
  title = {power conversion loss},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/power-conversion-loss/rdr_d621707d22044a6e924af25301e6a1be/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake