When the storage stops, who fixes it?
The licensed chief engineer does not fix the battery

Talk with people involved in grid-scale battery storage and you will often hear that "there is hardly anything to do for O&M". Build the facility, energize it, connect it — after that the aggregator schedules charge/discharge, the licensed chief engineer performs statutory inspections, and the OEM repairs whatever breaks; with these three parties, the thinking goes, the storage facility runs. But these three leave out one decisive point. When something goes wrong on the DC side (EMS, PCS, BESS, BMS), who is the first to triage it and begin recovery? The aggregator faces the market on the assumption that the equipment is running; it does not maintain the equipment. The chief engineer is the guardian of the receiving/transforming (AC) side. The OEM looks only as far as its own devices. Between the three, a gap remains: first response on the DC side.

This article does not call that gap a "regulatory gap". Look closer and the fact is the opposite: the law clearly imposes statutory safety-supervision responsibility for the DC side on the chief engineer. What is empty is not the law but who, on the ground, executes that statutory responsibility — a mismatch of "responsibility is statutory, execution is absent". This may be somewhat more awkward than "the regulation has a hole". On paper the hole is filled, yet in reality it is not. So no one picks it up as their own problem.

This article first fixes the structure of that mismatch in law, then draws the whole picture with a responsibility matrix of equipment × party. Next it organizes the economic consequence — that an outage is not merely opportunity loss, but bites in two stages as institutional penalties in the balancing market and the capacity market. Finally it draws on international practice to show how mature markets solved this gap, connecting to decision axes that storage owners, EPCs, and O&M providers can use in contract design. If our companion piece, "Contract design splits JPY 200 million over 20 years", was a map of the 21 contracts across 5 layers that support economic operation after COD, this article is its counterpart: a map of responsibility that supports physical operation after COD.

01 — The "three parties run it" misconception

Operation of a storage facility is often described in terms of three parties: the aggregator, the licensed chief engineer, and the OEM. But each of the three has a scope it takes on and a scope it does not. Stacking them does not add up to maintaining the health of the whole facility.

The aggregator uses the owner's battery to schedule charge/discharge in power markets and earn revenue. Its role is the economic operation of the battery; equipment maintenance is simply not its domain. It faces the market assuming the equipment is running, and is not the party that goes to site when things stop.

What about the licensed chief engineer? Here many people assume that "if there is a chief engineer, the equipment will be looked after", but this mistakes the scope. For HV facilities and above, the Electricity Business Act requires appointing a chief engineer, and there is no doubt they are electrical professionals. They handle operation, fault investigation, and shutdown/restart of the cubicle (receiving/transforming equipment). But few chief engineers can restore the EMS or PCS underneath it on their own. As a rule they are the guardian of the receiving/transforming side, not someone who takes on, in practice, the role of guardian of the DC-side equipment. Even when the work is outsourced to a safety management organization, it is common to draw the line so that the cubicle is covered but the DC side is not.

Then does the OEM fix it? A storage facility is made up of devices from multiple makers — EMS, PCS, BESS — and each company handles maintenance of its own devices (periodic inspection, call-out, correction). But the makers do not each hold enough field personnel to instantly dispatch a technician to every one of the storage facilities now multiplying nationwide. Query each company before it is settled which device the cause lies in, and time simply drains away over where responsibility sits. There is also an easily overlooked constraint. Maker warranties, as a rule, lapse if anyone outside the maker's certification touches the device — even a license holder. In other words the DC side is not "untouchable"; rather, "touching it voids something else (the warranty)". It is more accurate to see it as fenced in practice by three things: skills, maker certification, and warranty scope. This is not a legal limit but, as we will see later, a constraint that can be moved by design.

If three parties do not complete the picture, then a party that integrates and manages the whole facility — O&M — has no choice but to step to the front. Move the chief engineer, move the OEM, use insurance, or treat communications as the cause. Taking on that judgment and first response, and becoming the single point of contact for the owner: this is the starting point of this article.

02 — Responsibility is statutory, execution is absent

This is where the heart of the article begins. The field's line that "the DC side is outside the chief engineer's scope" is not because the law defines it so. In law, the DC side is squarely included in the object of safety supervision.

Article 43(1) of the Electricity Business Act requires a person who installs business-use electrical facilities to appoint a chief engineer to supervise safety regarding their construction, maintenance, and operation (Article 42 also requires preparing and filing safety management rules). The question is whether "business-use electrical facilities" include the DC side, and here the legal definition of a storage facility answers it. Under the Ministerial Ordinance on Technical Standards for Electrical Equipment, a storage facility is a place that stores, by means of an electricity-storage device and other electrical equipment installed on the premises, electricity transmitted from outside, and transmits it back outside at the same voltage and frequency. The inverter (PCS) that drives the storage device and the protective devices are included in this electrical equipment. There is no wording anywhere that excludes the DC side from the object of supervision.

The qualification tiers are also drawn by voltage, not by equipment type. Under Article 56 of the Enforcement Regulation, the supervision scope by license class lets a Third-class chief electrical engineer supervise business-use electrical facilities below 50,000 V, but power plants and storage facilities of 5 MW (5,000 kW) and above are excluded. Second-class covers below 170,000 V, First-class everything. The structure simply was never built to divide supervisability by AC vs. DC or by switchgear vs. battery. METI's materials on electrical-safety personnel also state the Third-class scope as "below 50 kV (excluding power plants or storage facilities of 5,000 kW and above)", so the fact that storage facilities of 5 MW and above are outside the Third-class scope is confirmable even in agency documents. Note that under the 2022 amendment to the Electricity Business Act (effective April 2023), discharge above 10 MW (10,000 kW) was positioned as a power-generation business.

But here a further degree of precision is needed. What is imposed on the chief engineer is supervision of safety — confirming conformity to technical standards and reporting to the installer if there is a risk of nonconformity — not correcting the PCS or BESS themselves. Supervisory responsibility extends by statute to the whole facility (DC side included), but there is no provision that names who actually fixes the devices underneath it (correction, first response). Pushed to its conclusion, the law names an execution party for only two things: "safety supervision of the receiving/transforming side" and "cybersecurity and technical-standards conformity at grid interconnection", and DC-side correction spills outside them. So the mismatch takes the form "supervision is statutory / correction is absent".

Let us lay this discrepancy on a single page, splitting the responsibility the law defines from on-site execution.

What the figure shows is a simple fact. The law imposes whole-facility safety supervision on the chief engineer, yet on-site execution stops at the receiving/transforming side, and correction of the DC side and communications belongs to no one. It is not that "the regulation has a hole"; it is that "on paper the hole is filled, but on the ground it is not" — and that is exactly why it is so hard to pick up.

03 — Responsibility matrix: equipment × party

Figure 1 viewed the discrepancy between law and field vertically; expanding the same structure across the plane of "which equipment, held by which party, on what basis" makes the location of the gaps far more concrete. Equipment domains run down the side, the parties involved run across the top, and each intersection is colored by one of four bases — defined by statute, defined by contract, the de-facto custom, or a gap that belongs to no one.

Follow the colors and the structure is visible at a glance. Dark (statute) lands only on the receiving/transform row and the part the owner bears via cybersecurity / technical standards. The large remaining area is occupied by contract (gold), custom (gray), and gap (red). The red in particular — the gaps — clusters in the owner column on the DC side (as-built docs, procurement, evidence data) and on the safety side of the comms row (chief engineer, safety org). Operating a grid-scale storage facility is nothing other than deciding which party, under which contract, owns these contract / custom / gap areas. Made concrete, each intersection looks like this.

There is one way to read it. The law clearly names a party for only "safety supervision of the receiving/transforming side" and "cybersecurity / technical-standards conformity at interconnection". Correction of the EMS, PCS, and BESS, and the design, documentation, and recovery of communications, lie outside them. To whom these gaps are assigned by contract becomes the heart of the project design. From here we dig, in order, into the two areas where the gaps cluster — DC-side correction, and communications.

04 — Why the hole does not get filled

If responsibility is statutory, why does execution go unfilled? The reason lies in two institutional structures. One is the system for "how to assign" the chief engineer; the other is the system for "what the safety management rules require to be written".

4-1 The constraint of the outsourcing-approval vessel

For HV storage facilities and above, the chief engineer is either appointed in-house or outsourced. METI's "Interpretation and Operation of the Chief Engineer System (internal rules)" sets, as a condition of outsourcing for the sites a chief engineer covers, that they can in principle reach the site within two hours (the so-called two-hour rule). Further, METI Notification No. 249 of 2003 limits the sum of conversion coefficients of sites a safety-duty officer may contract to under 33. It is a scheme that caps contracts by "points" set according to equipment capacity/output, and solar power plants and storage facilities carry condition-specific coefficients (0.31–0.33 in the notification; by category, a 0.32 / b 0.31 / c 0.33). The outsourceable range for storage facilities, like solar power plants, is set at output below 5 MW (5,000 kW) and voltage below 7,000 V (METI, "On the form of safety regulation for storage facilities", Apr 15, 2022).

This vessel of "a chief engineer within two hours and with points to spare" is no problem when personnel are plentiful. But reality is the opposite. METI's "On the chief electrical engineer system" (Mar 31, 2023) shows that, absent new measures, there could be a shortage of about 1,000 Second-class and about 800 Third-class engineers in FY2030. Outsourcing accounts for most appointment forms, and the workforce is aging. About 60% of license holders are 50 or older, and the Electrical Safety Subcommittee's materials project that Third-class will face a shortage of about 4,000 against estimated demand of about 18,000 in 2045.

From this, one can read that the ease and cost of securing a chief engineer change structurally by region. When the distance requirement, the points limit, and the personnel shortage overlap, finding a nearby candidate with points to spare, appointing them, and turning that into a quote takes a fair amount of time. An O&M company's needing several weeks to present a quote stems from this institutional vessel — it is not that some company is slow, but that the system is built that way.

4-2 What the safety management rules do not require

The other structure lies in the required contents of the safety management rules. Article 50 of the Enforcement Regulation of the Electricity Business Act enumerates the matters to be set in the safety management rules. A grid-scale storage facility that is a power-generation business (output above 10 MW) sits under paragraph 2, in-house use etc. under paragraph 3 — the paragraph differs by equipment category, but all consist roughly of these items: involvement and organization of the business manager and management-responsible officer; safety education; patrol/inspection/examination; operation; methods of preservation when stopping for a considerable period; measures to take in disasters and other emergencies; records on safety; and the system and record-keeping for statutory self-inspection and pre-use self-confirmation.

Here there is no provision that requires, as an independent item, how to keep the as-built documentation of communications equipment, or who recovers it and how when a router fails. The standard structure of safety management rules is built around the receiving/transforming side and electrical safety, and communications design documents and recovery duties spill outside the system's documentation requirements. When the field in the FIELD NOTE says "as-built docs go un-handed-over on many projects", it is not because they slacked on paperwork, but because the system does not require it to be written. What no one explicitly requires, no one explicitly keeps. This is the true nature of the gap on the communications side.

05 — The real cost of downtime

"If it stops on the DC side, you just can't go to market for a while" — if that is the thinking, the cost may be underestimated by a notch. An outage at a grid-scale storage facility is not limited to opportunity loss. It bites in two stages, as institutional penalties in the balancing market and the capacity market. First, let us pin down the magnitude of the revenue at stake when it stops.

The capacity-market main auction clearing price for delivery FY2028 is roughly JPY 8,785–14,812/kW-yr by area (Hokuriku, Kansai, Chugoku, Shikoku at the floor 8,785; Hokkaido, Tohoku, Tokyo at the ceiling 14,812; OCCTO clearing result, published Jan 29, 2025). The JEPX spot annual average for FY2024 was JPY 12.31/kWh (JEPX "FY2024 Business Report"). The balancing-market ΔkW cap price was cut from the former JPY 19.51/ΔkW·30min to 15 alongside the move to day-ahead trading in March 2026, with a stated policy of stepping it down to 10 and 7.21 if competition does not improve (ANRE, Jan 2026). If a 2 MW-class unit stops participating, the daily revenue tied to these is lost. But what really bites is not the opportunity loss itself, so much as the institutional penalties beyond it.

5-1 Balancing market: fail to respond, get cut, get locked out

A resource that clears in the balancing market (EPRX) bears an obligation to respond to dispatch. EPRX's trading rules define assessment (verification of response performance) and penalties. Failing Assessment II, which verifies response performance, incurs a 1.0x penalty on the cleared amount (this multiplier was revised from 1.5x to 1.0x at the 36th Balancing Market Subcommittee on Mar 2, 2023, to increase the incentive to participate; Assessment I, which checks maintenance of a deliverable state, is graduated by degree of nonconformity, with a maximum strength of 1.5x). Further, three or more nonconformities for the same resource and same product within one calendar month halt new trading of that product. Stopped, you cannot respond; failing to respond, you are cut; repeat it and you are locked out of the market — a three-tier structure.

That said, there is a relief framework for events caused on the grid side, such as output curtailment, that the operator could not foresee at bidding: the penalty multiplier is set to 1.0x and the event is excluded from the nonconformity count (application for grid-cause penalty relief; Forms 24 and 25 apply from the Mar 14, 2026 delivery onward). Whether you stopped yourself or the grid stopped you splits the treatment — here too, triage of the cause matters. The product mix, minimum bid size, and the downward direction of cap prices in the balancing market are organized in the contract-design article (Layer 5).

5-2 Capacity market: an unplanned outage bites at "5x"

A storage facility with a capacity provision contract in the capacity market — especially via the Long-term Decarbonization Auction (classified as a "stable power source") — bears a requirement to keep operating. The capacity provision terms define assessment and penalties, and the economic penalty is levied, based on the capacity contract amount, on the cumulative requirement-shortfall slots minus 8,640 slots (about 180 days) (OCCTO "Capacity Market Operation Manual (requirement/penalty)", "Capacity Provision Terms"). Conversely, up to about 180 days a year is effectively a grace, and shortfalls beyond that are the target.

What deserves attention is how those "shortfall slots" are counted. Maintenance filed in advance as a planned outage is counted at parity, but a sudden unplanned outage is counted at 5x (excluding normal nights and holidays). The system explicitly quantifies, at 5 to 1, the economic difference between planned maintenance that caught the early signs and an unplanned, sudden failure.

This factor of five is what most sharply explains the economic value of O&M. Leave a sudden failure as an "unplanned outage" and it bites at 5x; catch the early signs and convert it into a planned maintenance outage and it is parity. Early anomaly detection, immediate triage of the cause, rapid recovery — what O&M response speed does is convert the sudden into the planned, compressing penalty exposure by up to about one-fifth. This is not about driving outages to "zero". It is about shifting the quality of an outage to the "planned" side. That is exactly why response speed has a clear economic basis even when uptime itself cannot be guaranteed.

5-3 Is there a "going rate" for annual O&M?

So what about the O&M cost itself? Here we write honestly. A public standard unit price for fixed O&M (OPEX) specific to grid-scale batteries does not yet exist. The few public clues are the operating-maintenance cost used for the model plant in the Long-term Decarbonization Auction; METI's review-committee material presents operating-maintenance cost (labor) at JPY 5,000/kW-yr ("Organizing the cost and revenue issues of grid / co-located renewable storage systems", FY2024 3rd Stationary Storage System Deployment Committee, Aug 29, 2024; refer to the model-plant information used to compute the LTDA ceiling price). On the international benchmark, NREL ATB places fixed O&M at about 2.5% of CAPEX (including battery augmentation, on a 15-year operating basis). On the ground, some weight it generously at around JPY 5 million/yr including insurance, while others see it landing at JPY 2.5–3 million/yr excluding insurance, but these are operator-stated figures, not publicly established levels.

When placing an O&M cost in a business plan, it is worth bearing in mind that this is a starting figure, not a going rate. What can be referenced publicly is the unit-price datum "JPY 5,000/kW-yr" and the international benchmark of about "2.5% of CAPEX", both of which swing widely with facility scale (2 MW-class vs. 10 MWh-class) and regional differences. Only after decomposing the line items (chief engineer, monitoring, call-out, communications, insurance) and laying them out can the reasonableness of the level be discussed. Indeed, the very fact that "the public standard unit price does not go beyond a single datum" becomes a point to fold in as uncertainty when seating an O&M cost in a business plan.

06 — Communications: no one's job

Of all the gaps in a storage facility, the hardest to pick up is communications. The reason is clear. Communications fall exactly at the boundary between the receiving/transforming equipment the electrical professional (chief engineer) watches and the control the device professional (OEM) watches. The chief engineer's safety supervision is built around the receiving/transforming side, and the OEM's warranty completes inside its own devices. The network itself — lines, routers, gateways, and what connects each device — is at the center of neither's domain. And as we saw in Chapter 04, the safety management rules do not require the as-built documentation or recovery duty of communications equipment as an independent item. A layer that no one explicitly holds is formed right here.

This gap bares its teeth at the moment of a stop. A storage facility depends on communications even more than solar does; output control and response to balancing dispatch both presuppose real-time communications. Yet even when the EMS screen shows "comms fault", that alone moves no one. Whether the local router went down, the fiber provider's side did, or a setting was lost — depending on where the cause is, the party to call and the recovery procedure change entirely. If the one to perform this triage is written nowhere in any contract, the moment the alert appears, the clock starts on searching for "who should move". Worse, the as-built documentation essential to that triage — which router got which settings, what is connected at which IP — is, on many projects, never handed over. Not because paperwork was neglected, but because the system did not require it written and the contract did not demand it. Communications are no one's problem in normal times and everyone's problem in an emergency.

On top of this gap, from 2027 onward, regulatory requirements arrive. The material of the 20th Grid Code Study Group (Dec 16, 2025; ANRE) set a policy to require, as a technical condition of grid interconnection, that "control systems with communication functions (PCS, EMS, etc.)" adopted by distributed energy resources use products that have obtained ★1 of the JC-STAR scheme. Application is scheduled for HV in April 2027 and for LV below 50 kW in October 2027 (LV is six months later as a transitional measure reflecting distribution inventory; this timeline was also touched on in the subsidy article). As a premise, the Tech Standards Ordinance already obligates business-use electrical facilities — except small business-use electrical facilities (solar etc. below 50 kW) — to ensure cybersecurity (Art. 15-2).

Here there is one practical misconception: the idea that "if the gateway router has obtained ★1, the devices inside it are out of scope". But the scheme's design philosophy is not network-wide; it applies requirements per product (system) that uses IP communication. ANRE's material organizes, in its footnote, that among the scope, products (systems) using IP communication are subject to the ★1-acquisition requirement, and acquisition is also recommended for IP-communication devices outside the scope ("On cybersecurity measures for distributed energy resources", Feb 12, 2026, Material 5). A router's own acquisition does not mean exemption for downstream devices, nor is this something acquired in bulk for a whole battery set. It becomes the work of enumerating the communication/control devices one by one and confirming whether each is a ★1-obtained product. ★1 and ★2 are the maker's self-conformity declaration, ★3 and above are third-party certified, and IPA began operation on Mar 25, 2025 and is accepting ★1 applications.

6-1 The communications required differ by product

Requirements differ by product in steps. Response time ranges from within 10 seconds for primary reserve to within 60 minutes for tertiary reserve ②, and secondary ① — which requires second-by-second control — is limited to dedicated lines only. Meanwhile tertiary ②, whose response is on the order of tens of minutes, can enter via a simplified dispatch system, with monitoring intervals in minutes.

Sources: EPRX "Product requirements and trading schedule for the balancing market", OCCTO Balancing Market Subcommittee reference materials. Secondary ① cannot do second-by-second control on a simplified dispatch system, so its comms line is limited to dedicated.

6-2 So what is communications O&M?

Turn all of the above over and the concrete services of communications O&M come into view. First, preparing and handing over the as-built documentation — receiving, as documents at COD, the line types, router/gateway models and settings, IP assignments, and the inter-device connection diagram, and keeping them updated. Second, securing a pre-configured backup unit — industrial-router lead times are hard to read amid world conditions, and without a spare on site, recovery can stall for weeks (lengthening lead times for network equipment in general are confirmable, but primary data on individual lead times for grid-use routers is not). Third, triage procedure and first response in an emergency — deciding in advance, when "comms fault" appears, whom to contact and in what order among the site, the line provider, and the maker. Fourth, conformity confirmation and device replacement for the ★1 requirement — how to maintain device-level conformity on interconnections and replacements from 2027. None of these is automatically included in the safety management rules, the maker warranty, or the aggregator contract. Whether you leave communications as "no one's job" or have it explicitly taken on in the O&M contract — that single point splits recovery speed when things stop.

Depending on "which product and which line" a storage facility earns on, the quality of communications required, and the loss when communications drop, change by orders of magnitude. The more you promise second-by-second response on a dedicated online line, the more vital communications become. Before describing communications speed with a single number, settle before COD which product and which line your asset runs on, and who fixes communications when they drop.

07 — Insurance does not pay "until the cause is known"

A storage facility is customarily covered by property insurance and by business-interruption (BI) insurance that indemnifies lost profit during a stop. But this insurance does not pay out automatically when an accident occurs. Payment presupposes proof that "the subject suffered damage by a fortuitous accident", and BI insurance indemnifies only the loss arising as a result of that damage. Batteries are still new equipment, and whether the cause is a cell defect, faulty workmanship, or a natural disaster, the burden of proving the cause weighs heavily on the operator; for events like thermal runaway, identifying the cause takes a long time (at the US Moss Landing, the cause was not announced for about five months after the September 2021 stop). Until the cause is fixed, the insurance money does not move. And integrating devices from multiple makers can produce situations where responsibility cannot be assigned to a single company (in Korea, 23 ESS fires occurred in 2017–2019; a public-private joint investigation named four factors but did not reach conclusive proof, and the battery makers contested the cause theories).

* The fine print of coverage, exclusions, payment conditions, and proof requirements depends on each insurer's individual policy terms, and public information is limited. The very fact that the terms are not sufficiently disclosed becomes a point to fold in as uncertainty when placing insurance as a premise of a business plan. The whole picture of insurance design is covered in the insurance-design article.

08 — How mature markets solved this gap

Is the DC-side responsibility-execution gap a phenomenon unique to Japan? Looking at international practice, the answer is: "mature markets solved this gap by consolidating it into a single point of responsibility". Japan's three-party fragmentation is positioned as the eve before a market that has not yet moved toward integration.

8-1 Single point of responsibility

At US grid-scale storage facilities, O&M handling on-site maintenance and asset management overseeing contracts, warranties, and finance are divided, and the owner bundles multiple service contracts with KPIs and penalties. Roles are further layered — the OEM bearing product warranty (roughly 2–3 years), the system integrator filling the gap against required performance, the O&M provider securing operation under a 10–20-year long-term service agreement (LTSA), the asset manager, and the optimizer that optimizes market operation (equivalent to Japan's aggregator). What is described in Japan as "the three parties run it" is only a part of these roughly five roles. US legal practice, too, commonly has the OEM bear performance guarantees under supply and long-term service agreements, and recent organizing notes that as OEMs provide integrated systems, the need for turnkey bundling is fading. Seating a single point of responsibility that takes on the whole facility in one hand is the mature market's solution. Japan's "aggregator + chief engineer + OEM" corresponds to a state where that single point of responsibility, for the DC side, remains absent and fragmented.

8-2 Three layers of guarantee, and their exclusions

In a market where a single point of responsibility holds, three layers — availability, performance, and capacity — are guaranteed by contract. On availability guarantees, the third-party certification body DNV reports that many contracts set availability at around 97% per year as the threshold below which liquidated damages arise. Actual contract clauses also publicly compute the penalty by multiplying the shortfall below guaranteed availability (in points) by capacity and unit price — for example, against a 98% guarantee, an actual 96.5% means the 1.5-point gap is multiplied by capacity and rate. Maintaining a degradation guarantee requires operation that records and retains data such as temperature, current, and SoC at 15-minute intervals (this granularity becomes necessary for warranty claims and for confirming tax-credit eligibility).

Here, Japan's "cannot guarantee" seen through Chapter 07 and this international "97% guarantee" do not contradict; rather they cohere. The international availability guarantee holds after excluding force majeure and grid-caused events. Research bodies also note that monitored power electronics can achieve contractual availability of around 99.9%, but only on condition of excluding external and force-majeure events and not embedding performance into the availability guarantee. Some O&M providers advertise "availability above 99%", but all stand on the premise of carving out external factors as exclusions. What differs is that, having carved out those external factors, a single OEM bundles and guarantees the remaining equipment-caused availability. In Japan, with no party to bundle the whole facility, there is simply no basis from which to guarantee availability in one place.

The same holds for capacity: the OEM bundles a design that guarantees the whole system on a 10–20-year scale and compensates degradation by augmentation. DNV's organizing puts the guaranteed capacity at roughly down to 70% of the initial, and NREL ATB's fixed O&M (2.5% of CAPEX) also includes augmentation cost. In Japan too, the LTDA permits including the cost of additional acquisition and full replacement, so O&M's reach extends from "fix it when it breaks" to "hold capacity throughout 20 years".

8-3 Safety regulation: domestic "3 m separation" vs. overseas "performance-based"

The same contrast of "Japan fixed values, overseas performance-based" appears in how safety regulation is assembled. Domestically, the FDMA standard revision promulgated May 2023 and effective January 2024 changed the regulatory unit from cell capacity to kWh, with a floor of above 10 kWh (10–20 kWh with fire-prevention measures is excluded = effectively notification above 20 kWh), and outdoor installation requires in principle 3 m or more separation from buildings (cubicle types etc. recognized as posing no fire-prevention obstacle are excluded). MWh-class grid use is naturally subject to regulation. By contrast, North America's NFPA 855 sets unit-to-unit separation at 3 feet (~0.9 m) in principle, but allows reduction if safety is demonstrated by UL 9540A large-scale fire testing. Fixed separation (Japan) or performance-based on test data (overseas) — even with the same goal of "preventing fire spread", design freedom changes; either way, maintaining separation, compartmentation, and inspection records throughout the operating period is a service someone must carry.

8-4 Three generalizations

Generalizing from the international comparison: first, the idea of functional division itself is universal. Second, the responsibility gap in the communications layer is strongly Japan-specific. The institutional split between the chief electrical engineer system (electrical safety) and communications/control (OEM / O&M) creates the DC-side and communications gaps. Overseas, because the LTSA comprehensively covers the whole on an availability basis, gaps are structurally less likely to arise. That said, as the Korean ESS fires showed, the ambiguity of responsibility itself when integrating multiple makers is universal. "The split between the electrical trade and the comms trade" may be Japan-specific, but the structure of a gap in integrated responsibility is universal. Third, that the chief engineer does not in practice cover DC-side correction is not an inevitability of system design, but a Japan-specific operating custom rooted in skills, maker certification, and warranty scope. In law, the DC side too is within the scope of safety supervision.

09 — Pre-COD sequence and decision axes

What the organizing so far shows is that O&M is not "something to think about after COD" but a design matter before COD. DC-side correction, communications documentation, warranty boundaries, penalty exposure — these are fixed at the moment of handover and become hard to insert later. In practice, you want to begin assembling O&M roughly six months before grid interconnection.

Placing the sequence for embedding O&M before COD on a timeline alongside the interconnection procedure looks like this.

A common trap here is the expectation that with turnkey bundling (full-wrap EPC), "if you hand it all over, they'll take care of everything". Certainly a turnkey EPC consolidates responsibility up to handover, but post-COD O&M, warranty, and communications documentation are not necessarily included in that contract. US legal practice, too, increasingly sees the need for turnkey bundling fade as OEMs provide integrated systems, with a structure in which the developer procures equipment directly from the OEM and entrusts construction to a separate EPC — the EPC bearing installation and workmanship defects, the battery supplier bearing device warranty — becoming one common type. Bundling does not necessarily gather responsibility to a single point. Rather, you need to explicitly decompose, by contract, "who, which equipment, until when, on what basis" holds it. Whether you can receive the as-built documentation (especially the communications setting documents) at handover, where the warranty boundaries lie, who bears the unplanned-outage penalty exposure — holding these as pre-COD checklist items can substantially reduce later disputes.

We have organized the points storage owners, EPCs, and O&M providers want to confirm before COD into nine axes. Tap to record your confirmation status.

* The above are design-level checkpoints and do not substitute for the legal conformity or contract negotiation of an individual project. Applicability changes with equipment category, output scale, and grid conditions.

10 — Open questions ahead

Grid-scale storage O&M is still a moving field on both the regulatory and market sides. Finally, we list the questions that will firm up from here.

The structural shortage of safety personnel. Renewable and storage facilities requiring Third-class supervision grow year by year, while the workforce faces projected shortages toward 2030 and 2045. The outsourcing vessel of a distance requirement (2 hours) and a points limit (sum under 33) bites harder as personnel thin. The ease and cost of securing a chief engineer change structurally by region — a premise to fold into site selection and O&M-cost estimates.

International alignment of cyber requirements. JC-STAR began ★1 mutual recognition with the UK PSTI Act from January 2026 and with Singapore's CLS from June 2026, and the US Cyber Trust Mark and EU CRA are already in force / operation. How Japan's device-level requirement connects with these zone-level overseas schemes directly bears on device procurement and replacement practice.

Building out maker certification and integrated O&M. The key to moving the current state — the DC side fenced by skills, maker certification, and warranty scope — lies in how to incorporate maker certification and who bundles and guarantees availability and capacity. How to assemble, within Japan's institutions and commercial customs, the problem mature markets solved with a single point of responsibility — this is the area to be made concrete from here.

The exit of augmentation and disposal. How to design the exit of 20-year operation — the cost and responsibility of augmentation to compensate degradation, and the disposal/recycling after operation ends. As the LTDA permits including the cost of additional acquisition and full replacement, and fire/hazardous-material regulation requires a safety regime throughout the operating period, O&M's reach extends from "build and run" to "hold and close out".

For reference, let us look at the market's depth in numbers. In the first Long-term Decarbonization Auction, battery bids were about 4.559 GW and awards about 1.092 GW (award rate ~24%) (OCCTO clearing result; note that batteries and pumped hydro combined bid about 5.397 GW, and this combined figure must not be mistaken for batteries alone). Interconnection-study applications to the grid reach the scale of tens of GW nationwide, and as these storage facilities reach COD one after another, who fills the DC-side gap seen in this article rises as an issue for the market as a whole.

Conclusion — Not "nothing to do", but "no one doing it"

"There is hardly anything to do for O&M" — this article started from that line. But looking closer, the structure was the opposite. It is not that there is nothing to do. The law clearly imposes statutory safety-supervision responsibility for the DC side on the chief engineer. What is empty is not responsibility but who, on the ground, executes that statutory responsibility — a "supervision is statutory, correction is absent" mismatch.

The aggregator faces the market assuming the equipment runs, the chief engineer supervises the receiving/transforming side, and the OEM maintains its own devices. The three each fulfill their scope, but a gap remains — DC-side correction and communications recovery, explicit in no one's contract. And the economic consequence of that gap bites in two stages, as institutional penalties: the balancing market's assessment and the capacity market's 5x unplanned-outage count. Mature markets solved this gap by consolidating it into a single point of responsibility that bundles the whole facility. That such a party has not yet grown in Japan is not a limit of the system, but a matter of contract, custom, and skills. That is exactly why it can be moved by design.

The buy side of a grid-scale storage facility wants to look, beyond price, site, and device specs, at "who holds the DC side, under which contract". The sell side finds that folding O&M and as-built documentation into the design is the shortcut to avoiding post-handover disputes. And the build/maintain side can start by treating the three walls — skills, maker certification, warranty scope — not as walls of law but as design variables. It is not that there is nothing to do; it is that no one is doing it — who fills that gap, under which contract, and by when. Deciding that before COD becomes the heart of the design that supports the physical operation of a storage facility running for 20 years.

Principal primary sources (confirmed as of May 2026)
Law: Electricity Business Act (Art. 43(1), 42, 2), Enforcement Regulation of the Electricity Business Act (Art. 52, 56, 50, 3-4), Ministerial Ordinance on Technical Standards for Electrical Equipment (definition of a storage facility; Art. 15-2), METI Notification No. 249 of 2003 (sum of conversion coefficients under 33; solar power plants / storage facilities 0.31–0.33) / e-Gov, METI.
Safety & personnel: METI "On the form of safety regulation for storage facilities" (Apr 15, 2022, Electrical Safety Division, Material 1), "Interpretation and Operation of the Chief Engineer System (internal rules)", "On the chief electrical engineer system" (Mar 31, 2023), Electrical Safety Subcommittee materials on electrical-safety personnel (FY2030 shortage of ~1,000 Second-class and ~800 Third-class; Third-class ~4,000 shortage in 2045; ~60% of license holders aged 50+).
Markets: OCCTO "Capacity Market Operation Manual (requirement/penalty)", "Capacity Provision Terms" (economic penalty = contract amount × (cumulative shortfall slots − 8,640 slots = ~180 days); unplanned outage counted 5x, excl. normal nights & holidays), capacity-market main auction clearing result (published Jan 29, 2025; by area JPY 8,785–14,812/kW-yr; Net CONE 9,875), Balancing Market Subcommittee materials (Assessment II penalty strength revised from 1.5x to 1.0x = 36th, Mar 2, 2023; Assessment II nonconformity is 1.0x regardless of degree, Assessment I is graduated, max 1.5x), EPRX trading rules and grid-cause relief forms (Forms 24, 25 from the Mar 14, 2026 delivery onward), balancing cap price (19.51 → 15 in Mar 2026, stepping to 10 and 7.21 / ANRE Jan 2026), JEPX "FY2024 Business Report" (spot annual average JPY 12.31/kWh), Long-term Decarbonization Auction clearing result (batteries bid 4.559 GW, awarded 1.092 GW, award rate 24%).
Cyber: 20th Grid Code Study Group Material 4 (JC-STAR ★1 requirement; HV April 2027 / LV below 50 kW October 2027; Dec 16, 2025), ANRE Material 5 (per-product IP-communication = device-level requirement; Feb 12, 2026), IPA JC-STAR (★1/★2 self-conformity declaration; ★3 and above third-party certified; operation began Mar 25, 2025), METI/IPA UK PSTI mutual recognition (Jan 1, 2026; ★1 only) / Singapore CLS mutual recognition (Jun 1, 2026).
Safety: FDMA revision of standards for storage-battery installations (promulgated May 2023, effective Jan 2024; regulatory unit changed from Ah/cell to kWh; floor above 10 kWh; 10–20 kWh excluded with fire-prevention measures = effectively notification above 20 kWh; outdoor requires 3 m or more separation from buildings; cubicle types etc. excluded), NFPA 855 (3 ft separation = ~0.9 m), UL 9540A, NERC CIP / IEC 62443.
O&M & international: Stationary Storage System Deployment Committee Material 3 (operating-maintenance cost JPY 5,000/kW-yr = refer to the LTDA model plant; Aug 29, 2024), NREL ATB (fixed O&M = ~2.5% of CAPEX, including augmentation, 15-year basis), DNV (availability around 97% per year as the liquidated-damages trigger point; guaranteed capacity roughly down to 70% of initial), public contract clauses (penalty computed as shortfall below guaranteed availability × capacity × unit price), research-body organizing (monitored power electronics can achieve ~99.9% contractual availability, on the premise of excluding force majeure / external events), Norton Rose Fulbright (15-minute-granularity data retention to confirm degradation guarantees and tax eligibility), Vistra (Moss Landing Phase I; Sep 2021 stop → Jan 2022 cause announcement = ~5 months).
* The following items are flagged as our observation (❓) because the primary original was not reached or public data is absent: the Korea ESS fire's four factors, USD 32 million in damages, 522 units (~35%) halted, and the makers' rebuttals are consistent across multiple English-language reports, but the Korean-language original of the Ministry of Trade, Industry and Energy (MOTIE) was not reached. There is no public data showing the regional ranking of chief-engineer compensation, the individual lead time of grid-use routers, or actual domestic uptime figures, so "absence of public data" is recorded as our observation. Some details of the outsourcing ratio and the age composition of safety personnel, and the annual increase in facilities requiring Third-class supervision, were not reached in the corresponding primary slides.
* This article is a self-edited piece without external supervision. Regulations and markets are mid-amendment, and figures and effective dates may change. Confirm investment decisions and legal conformity against primary sources and with experts.