Conclusion. In the “2026 AI-Based Distributed Energy Special Zone Support Program” announced by the Ministry of Climate, Energy and Environment on September 8, Jeju was selected as the demonstration site. Three business models will undergo on-site due diligence at the end of this year and then be demonstrated over the next five years: an energy-zero housing complex (around 20 households in Hallim-eup or Daejeong-eup) that ties solar, heat pumps, EVs, ESS and AI appliances into a single system; an EV charging station that links a VPP, ESS and ultra-fast EV charging and goes as far as bidirectional charging and discharging (V2G); and a smart farm using non-lithium ESS.
The conclusion from an electrical equipment engineer’s standpoint is this. The core of this demonstration is not how precisely AI optimizes charging and discharging, but designing the connection boundary so that the AI’s charge/discharge schedule recommendations must pass through a distribution grid fault detection interlock before they turn into actual switching commands. And if non-lithium ESS is treated with the same protection parameters as lithium ESS, the safety advantage of this battery is lost precisely where it matters: in the interlock design.
However accurately AI predicts when to charge and discharge, that accuracy becomes meaningless the moment its output sits on a path that can bypass the distribution grid interlock.
Some background. The Ministry of Climate, Energy and Environment selected Jeju as the demonstration site for the AI-based distributed energy special zone on the basis of the region’s expanding deployment of residential heat pumps, an EV penetration rate that reached 12.2% as of the end of August, and a grid with a high share of renewable energy. The energy-zero housing complex connects solar, heat pumps, EVs, ESS and AI appliances into a single system to manage power generation and consumption in real time at the village level, and surplus power is shared and traded within the village in time slots when electricity rates are low or demand is low.
The EV charging station project links a virtual power plant (VPP), ESS and ultra-fast EV charging service, uses AI to forecast renewable generation and optimize ESS charging and discharging, and installs bidirectional V2G chargers at key locations to store energy in EVs when renewable output is high and return it to the grid when demand is high. The smart farm project uses an AI-based environmental control and energy management system to forecast renewable generation and farm energy use and apply them to heating, cooling, lighting and nutrient solution supply.
In particular, non-lithium ESS will be applied to the EV charging station and smart farm projects to verify long-duration operating performance, safety and economics and to confirm commercial viability.
Premise (based on an operating time chart). The facts disclosed in this announcement are the outline of the three business models, Jeju’s EV penetration rate (12.2%), the demonstration period (five years after on-site due diligence), and where non-lithium ESS will be applied (the EV charging station and the smart farm). The actual discharge enable conditions of the V2G chargers, the communication method between the AI schedule signal and local protection relays, and the specific protection parameters of the non-lithium ESS have not been disclosed. The interlock order and I/O map below are a conceptual example that translates general V2G and distributed ESS control principles for this kind of demonstration into an electrical configuration.
Safety interlock verification order (top priority)
- Check distribution grid fault detection first — distribution grid fault signals such as overvoltage, undervoltage and frequency deviation (DI_GRID_FAULT) are assigned as the top-priority input that immediately blocks V2G discharge enable, regardless of whether the AI server has approved a schedule.
- Limit AI schedule signals to advisory grade — the charge/discharge schedule produced by the AI VPP optimization server is reflected only as one of the enable conditions (an AND condition), and no path is provided through which it can drive a switch on its own.
- Verify connector lock and EMO hardwiring — the EV connector physical lock confirmation signal and the Emergency Off (EMO) signal must operate over hardwiring regardless of communication or AI server status, and any configuration that bypasses them in software is excluded.
- Set dedicated protection parameters for non-lithium ESS — non-lithium (for example, vanadium-based) batteries differ from lithium-ion in thermal runaway risk and response characteristics, so the temperature and voltage trip thresholds of existing lithium ESS are not transplanted as-is; separate protection parameters suited to the battery chemistry are set instead (specific values to be verified against supplier specifications).
- Predefine fallback behavior on communication failure — before the demonstration begins, define fallback logic so that V2G chargers and ESS return safely to default states (discharge disabled, charging only, and so on) when the AI server or communication network fails.
V2G and non-lithium ESS interlock I/O map (example)
| Signal | I/O address | Type | Operating condition |
|---|---|---|---|
| Distribution grid fault detection | DI_GRID_FAULT | Digital In | TRUE on overvoltage, undervoltage or frequency deviation; top-priority trip |
| AI charge/discharge schedule approval | DI_AI_SCHEDULE_OK | Digital In | Recommendation signal from AI VPP server (advisory grade, cannot drive alone) |
| EV connector lock | DI_EVSE_CONNECTOR_LOCK | Digital In (hardwired) | Bidirectional power flow allowed only after lock is confirmed |
| Emergency Off (EMO) | DI_EMO_TRIP | Digital In (hardwired) | Immediately cuts off the power converter regardless of AI or communication status |
| Non-lithium ESS SOC | AI_ESS_SOC_PCT | Analog In | 0–100%, separate curve reflecting long-duration discharge characteristics |
| Non-lithium ESS temperature | AI_ESS_TEMP_C | Analog In | Lower thermal runaway risk than lithium, but separate threshold needed (to verify) |
| V2G discharge enable | DO_V2G_DISCHARGE_ENABLE | Digital Out | ON only when all of the above conditions are met |
The takeaway from this I/O map fits in one line. Among the inputs feeding DO_V2G_DISCHARGE_ENABLE, only DI_GRID_FAULT and DI_EMO_TRIP are hardwired paths fully independent of the AI server’s state, and a configuration in which discharge cannot start as long as either of these two signals is active, however normal the other signals are, is the minimum condition for protecting the distribution grid from human and communication errors in a V2G demonstration.
PLC logic snippet (IEC 61131-3 Structured Text pseudocode)
(* V2G discharge enable logic - conceptual example, unrelated to any actual manufacturer specification *)
(* Note: specific thresholds not in the program announcement are not included *)
VAR
DI_GRID_FAULT : BOOL; (* distribution grid fault detection, top priority *)
DI_AI_SCHEDULE_OK : BOOL; (* AI VPP server recommendation, advisory grade *)
DI_EVSE_CONNECTOR_LOCK : BOOL; (* connector lock confirmation, hardwired *)
DI_EMO_TRIP : BOOL; (* Emergency Off, hardwired *)
AI_ESS_SOC_PCT : REAL; (* non-lithium ESS SOC, 0-100 *)
DO_V2G_DISCHARGE_ENABLE : BOOL;
END_VAR
(* Discharge enable: grid normal AND connector locked AND AI approved AND EMO not active AND SOC headroom *)
DO_V2G_DISCHARGE_ENABLE := (NOT DI_GRID_FAULT)
AND DI_EVSE_CONNECTOR_LOCK
AND DI_AI_SCHEDULE_OK
AND (NOT DI_EMO_TRIP)
AND (AI_ESS_SOC_PCT > 20.0);
(* On a grid fault, cut off immediately regardless of AI approval *)
IF DI_GRID_FAULT OR DI_EMO_TRIP THEN
DO_V2G_DISCHARGE_ENABLE := FALSE;
END_IF;Boundary conditions: dealing with field variables
There is a point where the advantage of this design principle flips. Non-lithium ESS is still at an early demonstration and mass-production stage at a small number of companies such as Standard Energy, so real C-rate and response speed data had not been sufficiently published before this demonstration. How well the fast charge/discharge demands of V2G actually match the long-duration, relatively low power density characteristics of non-lithium batteries is something this demonstration will have to confirm. In addition, the effect of communication delay or noise between the AI server, chargers and distribution grid on when the DI_GRID_FAULT signal is recognized requires waveform analysis and field measurement. There is a single selection criterion: whether the V2G discharge enable signal can be cut off immediately through a hardwired path fully independent of the AI server’s software state. Until that path is confirmed, it is safer not to scale up the demonstration.
References
Electric Times, “Jeju to become a ‘testbed’ for AI-based distributed energy demonstration” (posted 2026-09-08 12:00, reporter Kim Bu-mi, citing the Ministry of Climate, Energy and Environment announcement of the “2026 AI-Based Distributed Energy Special Zone Support Program”)
Patent KR20230171367A, Standard Energy Co., Ltd., “ESS/UPS Combined System that supports wide C-rate coverage” — priority date 2022-06-13, filed 2023-02-24, published 2023-12-20; legal status per Google Patents: Active (KR). It claims a combined ESS/UPS charge/discharge control unit that uses a vanadium-based (non-lithium) battery to support a wide C-rate range. It is cited as prior art related to the non-lithium ESS response characteristics discussed in this article; whether this supplier has actually been adopted for the Jeju demonstration has not been confirmed.
