After a DER dispatch event, there is always a moment of accounting: what was dispatched, what was delivered, what gets paid, and what gets reported. For a utility operating a portfolio of distributed energy resources that includes rooftop solar inverters, behind-the-meter batteries, electric vehicle chargers, and water heaters enrolled in a virtual power plant, the post-event documentation is not a clerical task. It is the evidence layer that supports settlement payments, program reporting, and the regulatory audit trail. AI can compress what used to take days of spreadsheet work into structured, verifiable summaries, but only if the person producing the summary understands what needs to be verified and what the model cannot know on its own.

What DER Orchestration Actually Produces

DER orchestration is the process by which a utility or third-party aggregator coordinates the dispatch of multiple distributed energy resources to achieve a system-level outcome: peak shaving, frequency regulation, voltage support, or a demand-response capacity commitment. Distributed energy resources (DER) is the umbrella term covering behind-the-meter generation and storage, including rooftop photovoltaic (PV) systems, battery energy storage systems (BESS), smart thermostats, water heaters, and EV chargers that can be remotely dispatched or modulated.

A DER orchestration event produces several categories of data that flow into downstream processes:

• Dispatch records: The control signals sent to each enrolled device, with timestamps, the requested action (charge, discharge, curtail, defer), and the expected magnitude. These come from the DER management system (DERMS) or the aggregator's virtual power plant (VPP) platform.
• Response records: The actual state change observed at each device, measured either through direct device telemetry, interval meter data from the advanced metering infrastructure (AMI), or derived from the distribution system operator's (DSO's) feeder-level measurements. These tell you what the aggregated resource actually did.
• Settlement data: The calculated performance for each enrolled participant, comparing delivered response against the baseline or the committed service level, multiplied by the applicable incentive rate from the program tariff or aggregation agreement.
• Program reporting data: Aggregated summaries showing total MW dispatched, total MWh delivered, total participants responding, average response time, and total program cost for submission to a state program administrator, ISO, or corporate program management team.

AI-assisted summarization and reporting enters this workflow after the data is collected. The AI tool does not collect the data, calculate baselines from scratch, or determine incentive rates. It organizes, synthesizes, and drafts readable summaries from data that the program manager provides. This distinction is the operational ground rule that prevents AI-assisted DER reporting from becoming a source of errors in regulatory documents.

AI-Assisted Dispatch and Settlement Summaries

Settlement is where DER orchestration becomes financial. Each enrolled participant receives a payment or bill credit reflecting their contribution to the dispatch event. For a large residential VPP program with 10,000 enrolled customers, individual settlement statements must be accurate, must reflect the program's tariff baseline methodology, and must be producible at scale without manual error in each one.

AI can assist in the following settlement steps, with the specified human verification required at each:

Baseline calculation setup: The program manager defines the baseline methodology from the program tariff (for example, a ten-similar-day average adjusted for weather), provides the relevant meter data, and asks the AI to structure the calculation template. The AI does not define the methodology; it formats the calculation from the methodology the manager provides.

Settlement statement drafting: The program manager provides the measured response data for each participant, the baseline values (from the DERMS or MDMS), and the tariff rate schedule. The AI drafts individual settlement statements showing the baseline, the measured response, the calculated reduction in kWh, and the incentive calculation. The manager spot-checks at least a 5% sample of statements against the raw meter data before the batch is processed.

Exception identification: After the batch draft, the manager asks the AI to flag any statements where the calculated reduction is negative (suggesting the participant's load went up during the event), zero (no response), or where the incentive calculation exceeds a plausibility threshold (suggesting a data error). These exceptions are reviewed manually before the settlement batch is processed.

Settlement narrative: For program reporting purposes, the manager uses AI to draft a narrative summary of the settlement results: total participants settled, total MWh delivered across the portfolio, average per-participant reduction, and total incentive payments made. The AI drafts from the summary statistics the manager provides; the manager verifies the totals against the DERMS settlement report before the narrative is finalized.

Settlement statements are not a place for AI confidence. Every number in a settlement statement comes from a data source: the meter, the tariff, the baseline calculation. If the AI cannot cite the source for a number, the number requires manual verification before it goes to a customer or a program administrator.

DER Dispatch Notes for Program Reporting

Program reporting covers the documentation a utility produces for its DER programs at the event level and the season or annual level. Event-level reports typically go to the operations team, the program management team, and in some cases to a state public utilities commission program administrator. Annual reports may go to the commission as part of the utility's required program reporting, to a corporate sustainability team, or to an ISO/RTO if the DER portfolio participates in a wholesale market product.

AI is particularly useful in compressing the narrative-writing step for program reports. A program manager who has the summary statistics in front of them can produce the event report narrative in seconds rather than hours. The critical discipline is providing the correct numbers to the AI before asking it to write the narrative. A prompt like "write a summary of our June 15 VPP event performance" without supplying the actual performance data will produce a plausible-sounding narrative that may contain entirely invented statistics.

The correct approach: the manager exports the event summary from the DERMS (total MW dispatched, average response by device class, total MWh delivered, response rate as a percentage of enrolled devices, any anomalies), provides this summary to the AI as input, and asks the AI to draft the report narrative. The AI organizes the numbers into readable prose. The manager reviews the draft to confirm that every number in the narrative matches the DERMS export and that the narrative does not introduce qualitative claims (such as "the program exceeded expectations") that go beyond what the data supports.

What DER Program Reporting Must Address

A complete DER program report for a state commission or an ISO program administrator typically must address:

1. Event summary: Date, time, duration, and triggering condition for each dispatch event in the reporting period.
2. Portfolio performance: Total MW dispatched versus total MW committed, with the performance ratio and any variance explanation.
3. Device class breakdown: Performance by technology type (BESS, PV inverter curtailment, HVAC, EV charging, water heaters), because different device classes have different response characteristics and the program may have class-specific performance targets.
4. Participant response rates: The percentage of enrolled participants who responded to each event, noting any systematic non-responders for follow-up.
5. Settlement summary: Total incentive payments made, broken down by participant class if applicable.
6. Program health indicators: Enrollment trends, attrition rate, and any customer service issues arising from event communications or settlement disputes.

AI drafts the narrative sections efficiently. The numbers in each section come from the program's authoritative data systems. The program manager's sign-off on the report attests that the numbers are accurate; the AI's role is producing readable prose around those numbers, not producing the numbers themselves.

Handling Anomalies in DER Dispatch Data

Real DER dispatch events produce messy data. Devices fail to respond. Meters report implausible values. Communication links time out during the event, leaving gaps in the response record. Battery systems that were dispatched to discharge show charging behavior in the meter data because a local control algorithm overrode the dispatch signal. These anomalies are normal, and their handling in the settlement and reporting workflow is a place where AI can help, but requires clear operational rules.

The most common DER dispatch anomalies and the recommended AI-assist approach for each:

Non-responding devices: When a device receives a dispatch signal but no response is recorded, the settlement for that participant is zero (no delivery, no payment). The report should note the non-response count and the fraction of enrolled devices that did not respond. AI can flag these accounts from the dispatch and response records and draft the non-response summary. The program manager verifies that the non-response count in the AI summary matches the DERMS dispatch log.

Communication failures: When a device's telemetry is unavailable during the event due to a communication outage, the program must decide whether to use estimated performance, feeder-level proxy measurements, or simply exclude the affected accounts from settlement. The decision is governed by the program's M&V (measurement and verification) protocol, not by the AI. The AI can draft the explanation of the communication failure and its settlement treatment once the program manager has determined the correct treatment from the M&V protocol.

Implausible meter readings: When a meter reports a value that is physically impossible (a residential battery showing 500% of rated discharge capacity) or inconsistent with the pre-event baseline (a device showing zero consumption when it was operating normally before the dispatch), the reading must be flagged for manual review by the meter operations team before settlement. AI can identify statistical outliers in a batch of meter readings, but the determination of whether an outlier is a meter error, a data transmission error, or an actual operating anomaly requires human judgment and meter-data-system review.

Override conflicts: When a battery's local control algorithm overrides the DERMS dispatch signal, the device may appear to behave inconsistently with the dispatch instruction. This can happen when the local algorithm's safety rules (protecting battery state of charge) conflict with the dispatch request. These events require documentation in the program report: the dispatch instruction, the observed behavior, the likely cause, and the program's treatment in settlement. AI can draft this exception narrative from the event log data.

Building the Settlement-to-Reporting Pipeline with AI

A DER program that operates multiple events per month and serves thousands of enrolled participants needs a reproducible, auditable pipeline from raw dispatch data to settled accounts to program report. Building that pipeline with AI assistance requires defining the workflow components clearly.

The pipeline has five steps, each with a defined human checkpoint:

Step 1: Data collection and validation. The DERMS exports the dispatch records, response records, and interval meter data for the event period. The program manager confirms the data export is complete and flags any gaps. This step is entirely human; the AI does not collect data.

Step 2: Exception screening. The program manager provides the raw data to the AI and asks it to screen for anomalies: non-responding devices, implausible meter values, communication-failure accounts. The AI produces an exception list. The program manager reviews each exception and applies the M&V protocol's treatment rules.

Step 3: Settlement calculation and drafting. The manager provides the validated data, the baseline values, and the tariff rate to the AI. The AI drafts individual settlement statements. The manager spot-checks a sample, resolves any exceptions from Step 2, and approves the settlement batch for processing.

Step 4: Event report drafting. The manager provides the settlement summary statistics to the AI and asks for the event report narrative. The AI drafts the report. The manager verifies the numbers match the DERMS summary and approves the draft.

Step 5: Archive and distribution. The settlement batch, exception log, event report, and M&V documentation are archived as a unit in the program record. The event report is distributed to required recipients (operations team, program management, commission administrator if required). The archive must be dated, versioned, and accessible for the program's audit retention period.

This pipeline produces an auditable record for every event. When a commission audit requests the documentation for a specific event, the program manager can produce the complete five-step package: data export, exception log, settlement batch with spot-check sign-off, event report, and archive receipt. The AI's role in the pipeline is Steps 2 and 4: exception screening and narrative drafting. Steps 1, 3, and 5 require human action and sign-off.

DER Program Reporting for FERC Order 2222 Aggregations

DER programs that participate in ISO wholesale markets under FERC Order 2222 aggregation rules face an additional reporting layer beyond the retail program report. FERC Order 2222, effective since 2020 but still being implemented at the tariff level by individual RTOs and ISOs as of 2026, allows DER aggregations to participate in all organized wholesale market products: energy, capacity, and ancillary services. The ISO market requires performance measurement against market product specifications that typically differ from retail program baselines in three material ways: the performance period is aligned to ISO market intervals (usually five-minute or hourly settlement), the baseline methodology follows the ISO's tariff-specified approach (not the retail program's tariff), and the settlement currency may be capacity MW-days or ancillary service MW for the relevant product rather than the retail energy payment metric.

A single DER event may trigger both a retail settlement for the enrolled customers and a wholesale market settlement with the ISO. Concretely: a BESS aggregation dispatched to reduce load during a summer peak hour settles with the ISO for the capacity commitment it fulfilled (measured in MW against the registered capacity value) and separately settles with enrolled residential customers for the energy reduction each household delivered (measured against the retail tariff baseline methodology). The two settlements use different baselines, different measurement periods, and different rates. Getting these confused in an AI drafting session produces inaccurate filings in both directions.

AI-assisted reporting for dual-settlement programs requires two explicitly separate workflows: one for the retail settlement using the retail tariff baseline methodology, and one for the wholesale market settlement using the ISO's market protocol. Each AI session must be grounded on its own source document set: the retail tariff rider for the retail workflow, and the applicable ISO tariff section and market rule for the wholesale workflow. The program manager runs both sessions separately, reconciles the results, and confirms that the aggregate retail payment plus the wholesale market receipt does not double-count any MW of delivered performance.

A common AI failure mode in dual-settlement contexts: the program manager pastes both the retail tariff and the ISO market rule into a single prompt and asks for a unified settlement summary. The model produces a blend that applies one methodology to some accounts and the other to others, or produces a single aggregate number that cannot be traced to either authoritative source. The correct prompt disciplines are: one source document, one settlement workflow, one verification pass. Then reconcile the two outputs as a human-judgment step before filing either.

Key Takeaways

• AI-assisted DER settlement and reporting compresses hours of data organization and narrative drafting into minutes, but only when the program manager provides the authoritative data as input. The AI drafts from what you give it; it does not produce settlement numbers or performance statistics independently.
• The five-step settlement-to-reporting pipeline (data collection, exception screening, settlement drafting, report drafting, archive) has two AI-assisted steps and three human-only steps, with a defined sign-off at each stage.
• Anomaly handling in DER dispatch data requires the M&V protocol, not the AI, to determine the correct settlement treatment for non-responding devices, communication failures, implausible readings, and override conflicts.
• Program reports must address event summary, portfolio performance, device class breakdown, participant response rates, settlement summary, and program health indicators; AI drafts the narrative efficiently when provided with verified statistics from the DERMS.
• For DER programs participating in wholesale markets under FERC Order 2222, retail and wholesale settlements require parallel AI drafting workflows grounded on separate source documents: the retail tariff and the ISO market protocol.
• Every settlement statement number must be traceable to a data source: the meter, the tariff, or the baseline calculation. An AI that cannot cite the source for a number in a settlement document has produced a number that requires manual verification.
• The audit trail for a DER program event is a unit: dispatch records, response data, exception log, settlement batch, spot-check documentation, event report, and archive receipt. That unit is what a commission audit or an ISO program review will request.