The three timescales
Dispatch decisions happen continuously at three distinct timescales, each with different decision authorities and different consequences. The architecture handles all three by separating their concerns cleanly: the controller owns sub-second protection autonomously; the controller and EstateAI share authority at seconds-to-minutes; EstateAI and the operator share authority at minutes-to-days.
Sub-second. Protection actions, transition decisions, voltage and frequency stability. The microgrid controller acts autonomously against pre-established rules and protective settings. Inverters respond to grid faults in cycles measured in milliseconds; the controller transitions between grid-connected and islanded operation in sub-cycle timescales when the macro grid faults. No human or higher-level system is in this loop. The controller is engineered, tested, and certified to make these decisions correctly, and the rest of the system depends on it doing so silently.
Seconds to minutes. The operational dispatch layer. Decisions about which generation sources contribute, how much storage discharges or absorbs, which loads are served at full capacity and which are throttled, when fuel-based generation engages. The controller executes against dispatch rules; EstateAI proposes adjustments based on conditions the controller does not have visibility into. This is the timescale at which the day-to-day character of the system is determined.
Minutes to days. The planning and optimization layer. Forecasting solar generation, anticipating household loads, scheduling discretionary loads against generation availability, maintaining storage state-of-charge against islanded-readiness targets, planning the cadence of any grid-supplementary engagement. The optimization runs against four forecast streams treated as a joint forecast with uncertainties properly propagated — generation forecast (how much the portfolio will produce, when, with what confidence), load forecast (the household’s typical patterns against current schedule and learned behavior), weather forecast (the conditions driving both generation and load), and grid forecast on grid-connected estates (time-of-use intervals, demand-response signals, outage probabilities). This is the timescale at which the family’s objectives are translated into operational behavior, and where EstateAI is doing its most consequential work.
The objectives
Sovereign-estate orchestration optimizes against a multi-dimensional objective function whose composition is itself a deliberate choice the family makes. Five objectives dominate, with priorities set explicitly by the family rather than assumed by the vendor.
Reliability. The household never experiences a power interruption. Not minimized; eliminated. This is the primary objective; the others are weighted against it, not balanced equally with it. The system that occasionally surprises the family with a flicker or a brief dropout has not been engineered to the standard the discipline requires.
Sovereignty. The macro grid is treated as a buffer, not a backbone. Dispatch maintains the system in a posture where grid loss does not change household operation and where the option to operate fully independent is always available. Sovereignty is pursued within the reliability constraint, not at its expense.
Endurance. Storage state-of-charge is maintained at levels that satisfy the family’s islanded-design intent. If the family has specified seven days of endurance at full operation, the dispatch logic maintains storage state-of-charge sufficient to deliver that whenever islanded operation begins. Endurance is the second-order reliability metric — reliability against not just the current moment but against scenarios the family has chosen to be ready for.
Asset preservation. Battery cycle life, equipment longevity, the multi-decade horizon. Dispatch chooses storage cycling depth, discharge rate, and operating temperature ranges with the 15-to-20-year service life in mind. A dispatch that meets the other objectives by cycling the batteries hard for short-term gain produces a system that needs replacement sooner than the engineering intent.
Cost. Relevant but ranked below the four above. Where the utility offers meaningful time-of-use rates, dispatch shifts loads to favorable intervals; where the utility provides demand-response payments, dispatch can participate. Sovereign-estate orchestration is a self-supply operation in which cost-optimization is opportunistic, not a utility-arbitrage operation built around cost as the primary objective.
The family’s household preferences (climate setpoints, scheduled loads, vehicle timing, hospitality patterns) and environmental priorities are configured as constraints the dispatch operates against, not as primary optimization targets. The family does not wait for the energy system to figure out what they want; the system serves what the family has said they want while optimizing within that intent. The objective function is set during commissioning and refined over the first year of operation; the discipline is to set the weights deliberately with the family’s active participation rather than accepting vendor defaults.
EstateAI as the optimization layer
The microgrid controller is the dispatcher; EstateAI is the optimization layer above it. The distinction matters because it is where the human-AI-controller architecture across the energy system becomes most consequential.
The controller executes dispatch rules at the sub-second-to-minute timescale. The rules are well-defined deterministic logic — if generation exceeds load, charge storage; if storage approaches full charge, curtail solar; if grid frequency deviates beyond threshold, transition to islanded operation; if the operator has authorized vehicle fast-charging, allocate storage capacity to absorb the peak. The controller is good at executing rules reliably; it does not know what the rules should be in a given context.
EstateAI operates above the controller, at the minute-to-day timescale, optimizing the rules and proposing adjustments based on the broader context the controller cannot reason about — the household’s schedule for tomorrow that affects which vehicles will be charging, the weather forecast that affects whether solar will be producing, the principal’s upcoming travel that affects the load profile, the household’s learned preferences about which loads matter most. The interaction takes three forms: rule updates (EstateAI adjusts the controller’s dispatch parameters within authorized bounds, visible on the operations console as a stream of system adjustments), action proposals (EstateAI surfaces decisions outside its autonomous envelope for operator review — running the fuel backstop ahead of a forecast severe-weather event, scheduling an eVTOL turnaround to coincide with peak solar, pre-conditioning a vehicle when the principal’s meeting moved), and forward-looking analysis (tomorrow’s expected generation, the week’s anticipated load patterns, the trends worth watching).
The architectural rule that holds: the controller acts at speed against rules it does not question; EstateAI optimizes the rules and proposes the work beyond rules; the operator approves what EstateAI proposes and sets the policies that constrain what both can do autonomously. This separation is what allows the system to be fast (controller), intelligent (EstateAI), and accountable (operator) at once.
Grid interaction as tactic, not strategy
Grid interaction — selling electricity back, participating in demand-response programs, time-of-use arbitrage, virtual power plant participation — occupies a different position on a sovereign estate than it does in the residential solar industry’s framing. There it is the central organizing concept; here it is a tactic where the utility supports it economically, never the primary objective the system is optimized against. A residence that organizes its energy management around grid interaction is a residence whose energy posture is dictated by the utility’s rate structures and program rules — the utility changes its tariffs (which utilities do regularly), the economics change, the dispatch rules have to be reconfigured. A residence that organizes around self-supply with grid interaction as opportunistic is a residence whose energy posture is its own.
Three modes appear. Self-supply with surplus export is the baseline — the estate generates and consumes its own electricity, with surplus exported where net metering or feed-in tariffs make it economically meaningful. Time-of-use arbitrage shifts discretionary loads and storage discharge against meaningful rate differences between intervals. Demand response and grid services apply where the utility offers payments for demand reduction or for storage to provide ancillary services; the opt-in is deliberate and bounded, because an estate that allows the utility to discharge its storage for grid stabilization during a heat wave is an estate that has compromised its own islanded readiness for grid revenue. Grid interaction is configured during commissioning, reviewed annually as the utility’s offerings change, and adjusted as the family’s priorities evolve.
The decisions worth surfacing
Four decisions in orchestration warrant principal-level engagement.
Objective function configuration. The relative weighting of the objectives is set by the family during commissioning, with the operator and the energy designer translating expressed priorities into the configuration the dispatch logic actually uses. A family that values absolute energy independence configures differently than a family that treats the grid as a useful tool. Default configurations vendors provide are rarely wrong for typical residential use; they are rarely right for the sovereign estate’s actual posture.
EstateAI’s autonomy envelope. What dispatch adjustments EstateAI is authorized to make autonomously versus what must be proposed for operator approval. The envelope is set conservatively at first — minor adjustments within established bounds, with anything consequential requiring approval — and expanded over time as the family develops confidence in EstateAI’s judgment.
Forecasting integration. The forecast streams must be reliable, current, and properly weighted in the optimization. Weather services with reliable forecasting at the estate’s specific location, load forecasting that learns the household’s actual patterns rather than relying on generic residential models, grid-condition forecasting where the utility provides data feeds. The integration is set during design and refined during commissioning.
Platform decision. Orchestration platforms range from integrated stacks (the microgrid controller vendor provides the dispatch logic, the platform, and the integration with EstateAI as one product) to open architectures (controller, optimization layer, EstateAI, and operations console as best-of-breed components integrated by the estate’s technical team). Integrated-stack reduces integration risk and accelerates commissioning; open architecture has more flexibility to evolve. The choice depends on the family’s posture toward technology dependence and is worth being explicit about during design.
EstateOps
Orchestration is where the energy system becomes operational rather than just installed. The controller executes; EstateAI optimizes; the operator approves and tunes. A well-running day looks like this: storage discharging through the night against minimum residence consumption; generation rising through the morning as solar comes online with surplus charging storage; the principal’s vehicle pre-conditioned to coincide with rising solar; discretionary loads (pool heating, working-space pre-cooling) scheduled against midday peak generation; storage providing peak power for evening vehicle charging and household activity; storage approaching its overnight minimum target by 2 AM. The dispatch logic has executed thousands of small decisions, the operator has reviewed perhaps four EstateAI proposals, the family has noticed nothing because nothing has happened that warrants their attention. That nothing-to-notice is orchestration operating correctly. The work is invisible because the engineering is right.