AI Analytics

Technical writing

NRC Nuclear Safety Data: The Federal Database Behind Every Reactor Inspection and Incident Report

· AI Analytics
Federal DataNRCNuclear SafetyEnergy

The Nuclear Regulatory Commission maintains the most detailed public safety oversight database for any energy technology in the United States. Every operating nuclear reactor generates quarterly performance indicator data, annual inspection reports totaling roughly 2,500 hours of inspector time, and event reports whenever something abnormal occurs — all of it published to the public through NRC's document systems. The result is an unbroken quantitative record of how every reactor in the US fleet has performed, what problems inspectors found, and how the NRC responded — a record that makes nuclear safety legible at the level of individual plants and individual events.

The Nuclear Regulatory Commission

The NRC was established in 1975 when Congress split the Atomic Energy Commission into two separate agencies: the Energy Research and Development Administration (which later became the Department of Energy) to promote nuclear power, and the NRC to regulate it. The structural separation was deliberate — Congress concluded that combining promotion and regulation in a single agency created an inherent conflict of interest that compromised safety oversight.

The NRC's jurisdiction covers civilian nuclear applications: the 99 commercial power reactors operating at approximately 60 plants as of 2024, the research and test reactors at universities and national laboratories, and roughly 20,000 specific licenses for radioactive material use in medicine, industry, and academia. The NRC does not regulate the Department of Energy's defense nuclear facilities or the commercial nuclear weapons complex — those fall under DOE's own safety programs.

The agency operates through four regional offices: Region I in King of Prussia, Pennsylvania (the Northeast); Region II in Atlanta (the Southeast); Region III in Lisle, Illinois (the Midwest); and Region IV in Arlington, Texas (the South and West). Regional offices conduct the bulk of reactor inspections, materials inspections, and enforcement actions. Headquarters in Rockville, Maryland sets policy and handles major licensing proceedings.

The NRC publishes an unusually large volume of its regulatory work as public documents. Inspection reports, enforcement orders, hearing dockets, environmental impact statements, license applications and amendments, and policy papers are all available through the Agency-wide Documents Access and Management System (ADAMS), which contains more than seven million publicly accessible documents. This transparency is partly statutory (the Atomic Energy Act and the Administrative Procedure Act require it), partly the product of post-Three Mile Island reforms that mandated greater public access to regulatory information, and partly a practical consequence of NRC's quasi-judicial licensing process, which requires a public record.

The Reactor Oversight Process

The Reactor Oversight Process (ROP) is the systematic framework the NRC uses to monitor the safety performance of operating nuclear power plants. Implemented in 2000 after the NRC concluded that its previous oversight approach was too subjective and inconsistent across the four regions, the ROP replaced a largely judgment-based inspection program with a structured, quantitative system that maps inspector findings and plant-reported performance data to a standardized response framework.

The ROP is built around three cornerstones of safety: Initiating Events, Mitigating Systems, and Barrier Integrity. The Initiating Events cornerstone addresses reactor shutdown events — the transients and accidents that challenge reactor safety systems. Mitigating Systems covers the equipment and procedures that prevent or limit core damage if an initiating event occurs: emergency core cooling systems, reactor coolant makeup, residual heat removal. Barrier Integrity covers the physical barriers that contain radioactivity: the fuel cladding, the reactor coolant system pressure boundary, and the containment structure.

Each cornerstone has a set of Performance Indicators — quantitative metrics derived from data that licensees report to the NRC quarterly. The PIs are color-coded on a four-level scale: Green (very low significance, acceptable performance), White (low to moderate significance), Yellow (substantial significance), and Red (high significance). A plant with all Green PIs is performing within acceptable bounds. A plant with White, Yellow, or Red PIs in any cornerstone triggers an escalating NRC response.

The action matrix is the ROP's decision framework. It maps each plant's combination of PI status and inspection finding significance to one of five columns:

The action matrix status for every operating US reactor is published quarterly on the NRC's public website. A plant in Column 2 or higher appears in NRC's “Reactor Oversight Process Action Matrix Summary” quarterly report, which names the plant, describes the finding or PI that triggered the escalation, and lists the NRC's planned response.

Performance Indicators

Licensees submit Performance Indicator data to the NRC quarterly via the NRC's Industry-Wide Nuclear Power Plant Performance Indicator Reporting System (INPO-coordinated). The NRC then publishes the data on its website, updated within weeks of each quarter's end. The data is presented as color-coded maps and plant-by-plant tables; underlying XML data files are also available for download.

The key performance indicators under the Initiating Events cornerstone include:

Under the Mitigating Systems cornerstone, the key PI is Safety System Unavailability — the percentage of time that high-pressure injection, emergency diesel generators, or auxiliary feedwater systems are unavailable due to planned or unplanned maintenance. A system that is unavailable when needed provides no margin. The Barrier Integrity cornerstone tracks reactor coolant system leakage rates (unidentified leakage rate in gallons per minute) and the number of fuel cladding failures detected through coolant activity monitoring.

All PI data for all operating plants going back to the ROP's 2000 implementation is publicly available for download at nrc.gov. The data structure allows analysts to track individual plants' PI histories over time, identify plants with persistent elevated indicators versus temporary anomalies, and compare fleet-wide performance across years.

Inspection Findings and the Significance Determination Process

NRC resident inspectors live and work at each nuclear plant site. Two to four resident inspectors are permanently assigned to each site and conduct continuous oversight, supplemented by specialist inspection teams from regional and headquarters offices who conduct detailed inspections of specific programs. The NRC's baseline inspection program calls for approximately 2,500 inspector-hours per plant per year, covering emergency preparedness, radiation protection, maintenance, engineering, and operations programs.

When an inspector identifies a condition that failed to meet NRC regulatory requirements or plant technical specifications, the finding enters the Significance Determination Process (SDP). The SDP is a structured methodology — essentially a simplified probabilistic risk assessment — that evaluates each finding against the plant's probabilistic risk model to estimate its contribution to core damage frequency. The result maps to the same four-color scale as the PIs:

All inspection findings that are not of very low significance (i.e., White, Yellow, or Red) are published in NRC enforcement actions and documented in plant-specific inspection reports available through ADAMS. Inspection reports for every operating plant, typically issued quarterly after resident inspector review cycles, are publicly available and contain detailed descriptions of all findings, the inspector's assessment of licensee corrective action, and the NRC's planned follow-up.

Event Reports and Licensee Event Reports

Nuclear power plant licensees are required to notify the NRC Operations Center of abnormal events within timeframes that depend on the event's significance. A four-hour reporting requirement covers events like emergency declarations, certain safety system actuations, and loss of safety function. Eight-hour reporting applies to unplanned scrams while the plant was in a condition that disabled safety features. Twenty-four-hour reporting covers a broader range of abnormal conditions. Thirty-day reporting covers less significant conditions identified through surveillance testing.

The NRC Operations Center publishes Event Notification Reports daily on the NRC website. These are brief summaries — a paragraph or two — of each call received, typically available within hours of the licensee's report. The daily event notification page is a starting point for identifying significant events across the fleet; an analyst monitoring this page has near-real-time visibility into abnormal events at every US reactor.

The more substantive document is the Licensee Event Report (LER). Within 60 days of an event, the licensee must file a formal written LER that provides a detailed description of the event, the plant conditions at the time, the cause analysis, the safety significance, the corrective actions taken, and any previous similar events that suggest a generic or systemic issue. LERs are filed in ADAMS and also searchable through the dedicated LER search interface at nrc.gov. Historically the US fleet generates approximately 100–200 LERs per year. The LER database going back to the 1970s provides a longitudinal record of equipment failures, human errors, procedural deficiencies, and external events that have challenged operating reactors.

LER analysis is one of the NRC's primary tools for identifying generic safety issues that affect multiple plants. If multiple plants file LERs describing similar failures in a particular type of valve, pump, or digital control system, the NRC's Office of Nuclear Regulatory Research may initiate a generic issue review that leads to an information notice, generic letter, or ultimately a regulatory requirement applied across the entire fleet.

Three Mile Island, Fukushima, and Post-Accident Reforms

The modern regulatory framework for US nuclear power bears the imprint of two major accidents: Three Mile Island Unit 2 in 1979 and Fukushima Daiichi in 2011.

The TMI-2 accident on March 28, 1979, began with a loss-of-feedwater event that caused reactor coolant temperature to rise. A stuck-open pressurizer relief valve — whose status was not clearly indicated to operators — allowed coolant to escape. Operators, misled by inadequate instrumentation, took actions that made the situation worse. The result was a partial core melt: roughly half the reactor core was damaged, with significant fuel melting. Radioactive gases were released, but no significant radiation injuries resulted. The physical consequences were contained; the regulatory consequences were sweeping.

Post-TMI NRC reforms were extensive. Emergency operating procedures were fundamentally redesigned: the old event-based procedures (identify what happened, follow the procedure for that event) were replaced with symptom-based procedures (monitor key safety parameters and take action based on what the plant is doing, regardless of cause). Control room design standards were revised: the inadequate human factors engineering that contributed to operator confusion at TMI led to systematic control room design reviews across the fleet. Operator training requirements were significantly strengthened, with formal licensed operator requalification programs and simulator training mandated. Emergency Planning Zones were established: a 10-mile EPZ for immediate protective actions and a 50-mile EPZ for ingestion pathway protection — defined areas with detailed emergency plans for evacuation, sheltering, and potassium iodide distribution that must be exercised in drills evaluated by FEMA.

Fukushima Daiichi on March 11, 2011 demonstrated that station blackout — complete loss of alternating current power — combined with prolonged loss of ultimate heat sink could lead to core damage even in reactors that had safely shut down. Three reactors at the plant suffered core damage and hydrogen explosions. The NRC convened a Near-Term Task Force within weeks, producing twelve orders and recommendations. US plants were required to obtain portable FLEX equipment — pumps, generators, hoses, and batteries — that can provide cooling and power for 72 hours without any fixed plant equipment or AC power. Enhanced flooding protection and improved hardened containment vent systems were also required. By 2017 all operating US plants had implemented the FLEX strategy, significantly extending the time available for operator response in a beyond-design-basis event.

Probabilistic Risk Assessment and Risk-Informed Regulation

The NRC requires each nuclear plant to maintain an Individual Plant Examination (IPE) and, for external events, an IPE for External Events (IPEEE) — collectively, probabilistic risk assessments (PRAs) that model the frequency and consequences of possible accident sequences for each plant's specific design and operating conditions.

The two primary PRA metrics are Core Damage Frequency (CDF) and Large Early Release Frequency (LERF). CDF is the annual probability of severe core damage — the result of accident sequences that overcome the plant's safety systems. LERF is the annual probability of a release of radioactive material so early and large that protective actions cannot prevent significant offsite doses. The industry-average CDF for the US nuclear fleet is approximately 1E-5 per reactor-year — one chance in 100,000 per reactor per year of core damage. LERF is typically an order of magnitude lower.

NRC uses PRAs to support risk-informed regulation: allocating regulatory attention and resources based on the safety significance of specific systems, components, and procedures. Risk-informed regulation has influenced the Significance Determination Process, the development of Risk-Informed Technical Specifications, and the approach to license renewal for aging reactors. Plant-specific PRAs are available in ADAMS, although some security-sensitive elements of PRA models may be withheld from public access.

Nuclear Materials Licensing and Agreement States

Beyond power reactors, the NRC regulates approximately 20,000 specific licenses for the use of radioactive material in medicine, industry, and research. These include licenses for radiopharmaceutical manufacturers, hospitals using radioactive isotopes for diagnostic imaging and cancer treatment, industrial radiographers, well-logging operators, and university research programs. The NRC also licenses uranium recovery operations (uranium mills and in-situ recovery facilities) and the temporary storage and disposal of low-level radioactive waste.

Thirty-eight states have entered Agreement State agreements with the NRC under Section 274 of the Atomic Energy Act. Under these agreements, Agreement States assume NRC's regulatory authority over most low-risk radioactive material uses within their borders, while NRC retains authority over reactors, high-level waste, and certain other categories. Agreement States regulate their own materials licensees under programs that must be at least as stringent as NRC's. The NRC periodically reviews Agreement State programs through the Integrated Materials Performance Evaluation Program (IMPEP) to verify they maintain adequate regulatory standards.

Post-September 11, 2001, NRC significantly strengthened nuclear security requirements. Design Basis Threat regulations specify the adversary characteristics that nuclear facilities' physical protection systems must be capable of defeating. Security inspections evaluate whether facilities meet these requirements. Cybersecurity regulation under 10 CFR 73.54 requires power reactor licensees to implement a cybersecurity program protecting digital systems from attack. Security inspection findings, unlike most NRC enforcement actions, are generally not publicly disclosed in detail due to their sensitive nature.

ADAMS — The NRC Document System

The Agency-wide Documents Access and Management System (ADAMS) is the NRC's electronic document management system and the primary repository for publicly available NRC records. ADAMS contains more than seven million documents accumulated since the system's implementation in 2000, supplemented by retrospective scanning of older paper records. The public-facing interface is the ADAMS Public Documents search at nrc.gov/reading-rm/adams.

ADAMS includes every publicly available NRC inspection report, enforcement action, LER, license application and amendment, environmental impact statement, safety evaluation report, and rulemaking document. The search interface supports full-text search across document content as well as metadata searches by docket number, document type, author, date range, and accession number. Each document has a unique ADAMS accession number (format: ML followed by year, day-of-year, and a sequence number, e.g. ML24100A123) that serves as a stable citation.

For reactor regulatory analysis, the most useful ADAMS search strategies are docket searches. Each licensed reactor has a facility docket number (typically 05000XXX format). Searching by docket number returns all NRC documents associated with that reactor, including the entire inspection history, licensing correspondence, and enforcement record. Combined with date range filters, docket searches make it possible to assemble a complete regulatory history for any operating or decommissioned reactor.

Documents in ADAMS are predominantly PDFs, including both machine-readable PDFs for documents created digitally and scanned PDFs for older paper records. The NRC's public API does not provide structured data extraction from document content — analysts working with ADAMS data typically rely on full-text search to identify documents of interest and then parse specific fields from the PDF text. For structured data analysis, the NRC's dedicated data tools (PI XML downloads, LER search, enforcement action tables) are more efficient than ADAMS full-text search.

Nuclear Capacity Factors and Plant Economics

The United States nuclear fleet operates at the highest capacity factors of any generation technology in the country. Capacity factor is the ratio of actual generation over a period to the maximum generation possible if the plant ran at full nameplate output continuously. From 2015 through 2024, the US nuclear fleet averaged capacity factors of approximately 92–93 percent — meaning US reactors generated roughly 92 percent of their theoretical maximum output year after year. By comparison, the US wind fleet averages roughly 35 percent and the solar fleet roughly 25 percent. Combined cycle gas plants typically run at 50–60 percent capacity factors depending on market conditions.

This performance is the product of sustained investment in reliability improvements since the 1980s, when US nuclear capacity factors averaged closer to 60 percent. Extended power uprates — license amendments allowing plants to operate at power levels above their original design ratings, typically by 5–20 percent — have added the equivalent of several new reactors' worth of generation capacity without constructing new plants. Refueling outage durations have declined from an average of roughly 100 days in the 1980s to around 30–35 days today, maximizing the proportion of each year a reactor spends generating electricity.

Despite this technical performance, approximately 15 nuclear plants retired between 2013 and 2023. The retirements were driven primarily by economics: the shale gas revolution produced persistently low natural gas prices, which depressed wholesale electricity prices in deregulated markets below the level at which some nuclear plants could recover their costs. Several states — Illinois, New Jersey, New York, Connecticut, and Ohio — implemented zero-emission credit programs that provide additional revenue to nuclear plants based on their carbon-free generation, successfully preventing further closures in those states. Individual plant capacity factor data is published by the Energy Information Administration through the EIA-923 Form and the EIA-860 plant inventory, as well as through the EIA's electric power statistics portal.

Python: Downloading and Ranking Plants by Unplanned Scram Rate

The following script downloads NRC quarterly Performance Indicator data from the NRC public website, parses the XML for each plant's unplanned scrams per 7,000 critical hours indicator, and produces a ranked table showing which plants have the highest scram rates and which have non-Green status. The script handles the two most common XML structure variants in NRC PI publications and prints a summary of plants outside the Green performance band.

import requests
import xml.etree.ElementTree as ET
from collections import defaultdict

# NRC Reactor Performance Indicators are published quarterly as XML.
# The public data endpoint serves the most recent four quarters combined.
# URL pattern for quarterly PI data:
#   https://www.nrc.gov/reading-rm/doc-collections/reactor-status/pi/current-pi.xml
# Each <plant> element contains <pi> children with name, value, and status attributes.
# Status maps to: G = Green, W = White, Y = Yellow, R = Red

PI_URL = "https://www.nrc.gov/reading-rm/doc-collections/reactor-status/pi/current-pi.xml"

# The indicator we want: unplanned automatic scrams per 7000 critical hours.
# NRC calls this "Unplanned Scrams per 7000 Critical Hours" in the XML.
# Abbreviation varies; match on a substring of the name attribute.
SCRAM_SUBSTRING = "Scrams per 7000"

STATUS_ORDER = {"G": 0, "W": 1, "Y": 2, "R": 3}
STATUS_LABEL = {"G": "Green", "W": "White", "Y": "Yellow", "R": "Red"}


def fetch_pi_xml(url=PI_URL, timeout=60):
    """Download NRC quarterly PI XML. Returns raw bytes."""
    resp = requests.get(url, timeout=timeout, headers={"User-Agent": "pi-analysis/1.0"})
    resp.raise_for_status()
    return resp.content


def parse_scram_indicators(xml_bytes):
    """
    Parse scram-per-7000-critical-hours PI values for each plant unit.

    Returns a list of dicts with keys:
        plant_name  - string, plant and unit identifier from XML
        pi_value    - float, the reported PI value
        status      - string, one of G / W / Y / R
        status_label - string, Green / White / Yellow / Red
    """
    root = ET.fromstring(xml_bytes)
    records = []

    # XML structure varies slightly by NRC publication year.
    # Try both <plant> top-level and <unit> nested within <facility>.
    for plant in root.iter("plant"):
        name = plant.get("name") or plant.get("unit") or "Unknown"
        for pi in plant.iter("pi"):
            pi_name = pi.get("name") or pi.get("indicator") or ""
            if SCRAM_SUBSTRING.lower() not in pi_name.lower():
                continue
            raw_value = pi.get("value") or pi.text or ""
            status_code = (pi.get("status") or pi.get("color") or "G").strip().upper()
            try:
                value = float(raw_value)
            except ValueError:
                value = float("nan")
            records.append(
                {
                    "plant_name": name.strip(),
                    "pi_value": value,
                    "status": status_code,
                    "status_label": STATUS_LABEL.get(status_code, status_code),
                }
            )

    return records


def rank_plants(records):
    """
    Sort plants by scram rate descending (highest scram rate = most events).
    Within equal rates, non-Green status sorts first.
    """
    def sort_key(r):
        pi = r["pi_value"] if not (r["pi_value"] != r["pi_value"]) else -1.0
        return (-pi, STATUS_ORDER.get(r["status"], 99))

    return sorted(records, key=sort_key)


def print_table(ranked):
    header = ("Plant / Unit", "Scrams/7000 hr", "Status")
    col_w = (40, 16, 10)
    sep = "  ".join("-" * w for w in col_w)
    fmt = "  ".join("{:<" + str(w) + "}" for w in col_w)
    print(fmt.format(*header))
    print(sep)
    for r in ranked:
        val_str = "{:.3f}".format(r["pi_value"]) if r["pi_value"] == r["pi_value"] else "N/A"
        print(fmt.format(r["plant_name"][:col_w[0]], val_str, r["status_label"]))

    non_green = [r for r in ranked if r["status"] != "G"]
    print()
    print("Total units with PI data:", len(ranked))
    print("Non-Green status units:  ", len(non_green))
    if non_green:
        print("Non-Green plants:")
        for r in non_green:
            print("  " + r["plant_name"] + " - " + r["status_label"])


if __name__ == "__main__":
    print("Fetching NRC quarterly performance indicator data...")
    xml_bytes = fetch_pi_xml()
    records = parse_scram_indicators(xml_bytes)

    if not records:
        print("No scram PI records found. The XML structure may have changed.")
        print("Inspect the raw XML to update SCRAM_SUBSTRING and element names.")
    else:
        ranked = rank_plants(records)
        print_table(ranked)

The NRC publishes PI data in XML format with a structure that has evolved over successive quarters; the parser above attempts to accommodate the most common variants. If the NRC changes its XML schema, the SCRAM_SUBSTRING string and the element attribute names may need updating — inspecting the raw XML at the published URL with any text editor will reveal the current structure. The output table ranks all reporting plant units by scram rate and separately lists any units in White, Yellow, or Red status, which are subject to escalated NRC oversight under the action matrix described above.

Nuclear plants are the largest source of carbon-free electricity in the United States, operating at capacity factors above 90 percent. EIA tracks nuclear generation alongside all other fuel types in its mandatory Form 923 reporting. See EIA Electricity Data: The Federal Dataset Behind Every Kilowatt-Hour Generated, Sold, and Priced.

Facilities that handle radioactive material are among the industrial sites tracked under EPA's Toxic Release Inventory, which requires annual reporting of hazardous substance releases to air, water, and land. See EPA Toxic Release Inventory: 35 Years of Industrial Chemical Releases and Environmental Justice Patterns.

Nuclear plants sell into wholesale electricity markets regulated by FERC, which oversees pricing, market rules, and anti-manipulation enforcement across organized wholesale power markets. See FERC Enforcement: The Federal Watchdog Over Energy Market Manipulation.