PaperlayerView exact revision on arXiv

Source-linked AI summary

Resource-Aware Rolling-Horizon Controller for Renewable-Based Virtual Power Plants Providing Automatic Frequency Restoration Reserves

Marco Vinicio Avendaño-Caiza, Juan Diego Rios-Peñaloza, Javier Roldán-Pérez, José Luis Rodríguez-Amenedo, Milan Prodanovic

arXiv:2609.31017v1eess.SYcs.SY

TL;DR

The paper addresses the limited treatment of real-time VPP operation alongside market scheduling and frequency control. It proposes a centralized rolling-horizon controller that allocates resources while following day-ahead commitments. Simulations report frequency restoration under ideal forecasts and forecast uncertainty, with resource allocation reflecting availability and operating constraints.

  • Problem

    Existing approaches leave the simultaneous treatment of renewable availability, market commitments, battery degradation, and operational restrictions in real-time VPP control insufficiently explored.

    Source: §I · p. 2

  • Method

    A centralized rolling-horizon controller allocates VPP power and reserve commitments among units while considering real-time operating aspects.

    Source: §I · p. 2

  • Results

    Frequency was restored after reference or load steps under both ideal forecasts and forecast uncertainty, while VPP power followed its day-ahead reference.

    Source: §VI · p. 9

  • Takeaways & Limitations

    The controller allocates VPP resources according to availability and operates renewable generators near maximum capacity except under technical constraints.

    Source: §VI · p. 10

Abstract

from arXiv · show

Virtual power plants (VPPs) are an effective solution for increasing the penetration of renewable energy sources (RES) in power systems. Moreover, VPPs can provide ancillary services if generators and loads are closely connected and properly coordinated. This aspect has been addressed in several studies, yet they mostly focus on the dispatch optimisation level, leaving real-time operation aspects unresolved. To close this gap, a central controller for VPPs based on a rolling-horizon optimisation is proposed in this work. Its main objective is to maximise revenues while delivering frequency restoration services. A day-ahead optimisation firstly defines the power setpoint and the upward and downward reserves of the VPP. Then, the rolling-horizon optimiser, executed every minute, distributes the power and reserve references between the VPP units to achieve their close tracking while taking into account resource availability (wind speed, irradiance, etc.), operational constraints and battery degradation. The effectiveness of the controller is tested using a two-area interconnected power system under different RES generation and demand profiles. The simulations are performed in MATLAB/Simulink and include detailed dynamic models of the VPP elements, demonstrating the applicability of the algorithm to real systems. The day-ahead and the rolling-horizon optimisation problems are modelled using YALMIP and solved using Gurobi. The obtained results demonstrate the VPP supporting grid frequency restoration by optimally allocating its resources, even under the presence of forecast uncertainty.

Cost Functions

The section defines the costs and tracking penalties used in VPP optimisation, alongside rolling-horizon operating variables and the motivation for coordinating economic scheduling with real-time control.

Cost Functions

  • The cost terms cover renewable operating costs, wind-speed-dependent degradation, battery charge and discharge degradation, and penalties for power and reserve tracking errors.The penalty formulation uses curve breakpoints and slopes, penalty factors, and tracking-error variables.Source: Cost Functions · p. 1

Rolling-Horizon Optimisation

  • Rolling-horizon variables include unit power, battery charge and discharge, renewable operating factors, and unit-level and VPP reserve capacity.The passage also defines maximum available power and reserve availability factors.Source: Rolling-Horizon Optimisation · p. 1

I. INTRODUCTION

  • VPPs coordinate distributed generation, flexible loads, and storage as one entity, but real-time resource availability and reserve activation complicate service delivery.Prior scheduling work considered uncertainty and battery degradation, while the cited discussion distinguishes scheduling from real-time operation.Source: §I · pp. 1–2
  • Existing approaches often simplify VPP or grid dynamics, assume predetermined reserves or commands, or leave simultaneous resource availability, market commitments, degradation, and operational constraints insufficiently explored.The work proposes a centralized rolling-horizon controller to address this gap in a two-area system using real production and demand profiles, including forecast differences.Source: §I · p. 2

II. OVERVIEW

The overview describes a two-area system with a renewable-based VPP and a controller that shares power and reserve commitments among its units according to real-time resource availability.

A. System Description

  • The test system comprises two interconnected areas and a VPP with two 60 MW wind plants, a 60 MW PV plant, a 60 MW battery, and distributed internal loads.The wind and PV plants use grid-following converters, while the battery uses a grid-forming converter.Source: A. System Description · pp. 2–3, Figure 1 · p. 3
  • Day-ahead optimisation sets VPP power and reserve references; during real-time operation, secondary control assigns frequency-restoration contributions and the VPP optimiser allocates its reference among sources.The system-level secondary controller uses frequency bias factors and PI controllers; the VPP allocation uses source availability factors.Source: A. System Description · p. 3

B. Optimisation Framework and Virtual Power Plant Controller Overview

  • The VPP controller computes unit power references and reserve allocations, using resource availability to support efficient operation.Its rolling-horizon optimiser determines operating conditions and upward and downward availability factors using forecasts of load, irradiance, wind speed, and battery state of charge; it accounts for degradation and improper-operation penalties.Source: B. Optimisation Framework and Virtual… · pp. 3–4

III. POWER SYSTEM MODELLING

The paper describes dynamic and electrical models for the interconnected areas, VPP generation and storage, and network components, then introduces the rolling-horizon controller and its inputs.

1) Equivalent Model of Areas:

  • The modelling section presents system models for completeness, while each network area is represented by a first-order swing equation with droop control and loads as an input.The load input is later used to model load profiles.Source: §III · p. 4, 1) Equivalent Model of Areas · p. 4

2) Wind Power Generators:

  • Wind farms are represented by a generator shaft with equivalent inertia, wind-speed-dependent operating-point tracking, and a grid-following converter with an LCL filter.The power coefficient uses an approximation cited as.Source: 2) Wind Power Generators · p. 4

3) PV Generator:

  • PV output is modelled as a function of irradiance, with an operating-point tracker using irradiance and the VPP power reference.The PV grid interface includes dc-dc and dc-ac converters and an LCL filter.Source: 3) PV Generator · p. 4

4) Battery Energy Storage System:

  • The battery model uses series resistance and an RC branch, with open-circuit voltage dependent on state of charge and grid-forming control for grid support.Its controller includes virtual impedance, power controllers, and dq current and voltage regulators.Source: 4) Battery Energy Storage System · p. 4

5) Transformers, Lines, and Loads:

A. Optimisation Framework

  • The rolling-horizon optimiser distributes VPP power and reserve references while accounting for resource availability, operational constraints, and degradation.Its inputs include wind speed, solar irradiance, state of charge, the VPP power reference, and committed reserves.Source: A. Optimisation Framework · p. 4

B. Day-Ahead Optimisation

Day-ahead optimisation sets the VPP’s market power and reserve commitments, and rolling-horizon optimisation defines the operating decisions used to follow them.

B. Day-Ahead Optimisation

  • The day-ahead stage maximises VPP profits from energy and secondary-reserve markets over a one-day horizon with one-minute steps.Power and reserve commitments remain constant within each quarter-hour market period.Source: B. Day-Ahead Optimisation · p. 4
  • The day-ahead optimisation uses forecast load, wind speed, and irradiance and includes power-balance, power, and energy constraints.Its purpose is to obtain realistic dispatch and reserve-commitment profiles; further details are referenced in.Source: B. Day-Ahead Optimisation · p. 4

C. Rolling-Horizon Optimisation

  • Rolling-horizon decision variables include VPP active power, upward and downward reserves, and renewable-unit operating factors.Tildes denote forecasts and hats denote rolling-horizon decision variables.Source: C. Rolling-Horizon Optimisation · p. 4

1) Objective Function:

The objective sums operating, degradation, and tracking-penalty costs to guide VPP resource allocation. Its terms account for wind conditions, battery state of charge, PV operation, and deviations from scheduled power and reserves.

1) Objective Function:

  • The objective function sums the VPP unit cost terms and penalties for power and reserve tracking errors.The terms include wind operating and wind-speed-related costs, battery discharge and charge degradation costs, PV operating cost, and tracking penalties.Source: 1) Objective Function · pp. 4–5
  • Wind costs model speed-dependent degradation and penalize renewable curtailment, with wind operating cost decreasing as Kop,w increases.The wind-speed cost uses threshold-defined intervals, while the curtailment penalty is included to encourage use of available resources.Source: 1) Objective Function · p. 5
  • The BESS discharge cost decreases as SOC rises and penalizes low SOC to discourage deep discharge, while charging cost penalizes high SOC to discourage overcharging.The discharge and charge degradation costs depend on operating mode and SOC range.Source: 1) Objective Function · p. 5
  • PV operating cost is modeled similarly to wind operating cost, while PV degradation costs are omitted as negligible.Source: 1) Objective Function · p. 6
  • An active-power tracking penalty is added to keep the VPP power reference close to its scheduled setpoint.The penalty is based on the active-power tracking error.Source: 1) Objective Function · p. 6
  • Reserve tracking penalties apply only when actual upward or downward reserve is below the scheduled reserve.Source: 1) Objective Function · p. 6

2) Constraints:

The constraints allocate forecast-dependent renewable generation, balance VPP power, and ensure committed reserves remain available. Battery operation is bounded by charging, discharging, and state-of-charge constraints.

2) Constraints:

  • Renewable generation forecasts and rated capacities bound available wind, PV, and battery power.Wind availability uses location-specific wind-speed forecasts; PV availability uses irradiance forecasts.Source: 2) Constraints · p. 6
  • Renewable operating factors stay between their minimum values and one, with an additional wind-farm rated-power limit at high wind speeds.The active-power operating point is constrained not to exceed wind-farm rated power when wind speed is high.Source: 2) Constraints · p. 6
  • VPP active power equals the combined wind, PV, and battery injections minus the forecast internal load.The battery uses the generator sign convention: positive power indicates discharge and negative power indicates charging.Source: 2) Constraints · p. 6
  • Upward and downward reserves are calculated for each unit and aggregated, with VPP reserve capacities required to meet their references.Reserve shortfalls are treated as deviations and penalised.Source: 2) Constraints · p. 6
  • The battery model prevents simultaneous charging and discharging and constrains state of charge within its limits.State of charge accounts for charging and discharging, battery efficiency, nominal capacity, and the optimisation sampling period.Source: 2) Constraints · p. 7

V. RESULTS

The results section evaluates the controller in a two-area system under perfect-forecast and forecast-deviation cases. The day-ahead references show major changes around sunrise and sunset, alongside smaller changes associated mainly with battery operating-condition changes.

V. RESULTS

  • The controller is evaluated in a two-area power system simulated in MATLAB/Simulink, with optimisation modelled in YALMIP and solved using Gurobi.The system parameters are reported in Table I.Source: §V · p. 7
  • The day-ahead optimisation uses a one-day horizon with one-minute steps, while rolling-horizon optimisation uses one-minute steps over ten minutes.Source: §V · p. 7
  • Two cases compare perfect forecasts with more realistic operation in which forecasted wind, irradiance, and load differ from actual values.Source: §V · p. 7
  • The day-ahead power profiles have large variations around 9 h and 18 h, corresponding to sunrise and sunset.Less pronounced changes around 2 h and 6 h are associated mainly with battery operating-condition changes.Source: §V · p. 7

A. Case 1: Operation Under Ideal Forecast Conditions

Under perfect forecasts, the controller restores frequency after load steps while VPP power tracks its reference; unit dispatch and reserve availability adapt to resource operating conditions.

A. Case 1: Operation Under Ideal Forecast Conditions

  • After 0.2 pu load steps in both grid areas, frequency quickly returns to its reference and VPP active power follows its reference while compensating for the load changes.The largest frequency variations coincide with the largest DAM reference changes.Source: A. Case 1: Operation Under Ideal… · p. 7, Figure 5 · p. 8
  • Wind farm 1 operates near maximum capacity except at 23 h, while wind farm 2 departs more from its theoretical maximum and PV stays near available power during daytime.The BESS remains close to its DAM schedule; the wind-farm allocation reflects higher wind speed at node 2.Source: A. Case 1: Operation Under Ideal… · p. 7, Figure 6 · p. 8
  • Downward availability generally follows renewable generation; BESS downward availability is high while discharging and zero when charging at nominal power.The downward availability profiles are shown for both wind farms, PV, and the BESS.Source: A. Case 1: Operation Under Ideal… · p. 7, Figure 7 · p. 8
  • Upward availability factors for wind turbines at node 1 and the PV plant remain close to zero; the battery and wind farm at node 2 vary intermittently, without intermittency in output power.A possible mitigation is to incorporate moderate penalty terms.Source: A. Case 1: Operation Under Ideal… · p. 8

B. Case 2: Operation Under Forecast Uncertainty

With actual RES generation and load differing from DAM forecasts, frequency and VPP output generally return to reference after transients, but a late-day resource shortfall reveals a battery power limit.

B. Case 2: Operation Under Forecast Uncertainty

  • With actual operating values, frequency and VPP output return to their references after transients, as in the perfect-forecast case.From about 22.5 h, both deviate from their references.Source: B. Case 2: Operation Under Forecast… · p. 9
  • From about 22.5 h, lower-than-forecast wind generation causes undergeneration that the BESS cannot compensate for at nominal power, ending only when node-2 wind recovers.The late-day frequency and VPP output deviate from their references during this interval.Source: B. Case 2: Operation Under Forecast… · p. 9, Figure 9 · p. 9
  • The BESS departs further from its DAM schedule to compensate for RES differences, while wind farm 2 and PV spend longer below rated capacity than in Case 1.Wind farm 1 operates at rated capacity all day without reaching saturation.Source: B. Case 2: Operation Under Forecast… · p. 9

VI. CONCLUSION

The centralized rolling-horizon controller combines economic scheduling with unit-level resource allocation for frequency-restoration commitments. Simulations report reference tracking and frequency restoration under forecast uncertainty, with storage compensating for renewable forecast deviations.

VI. CONCLUSION

  • The centralized rolling-horizon controller allocates committed energy and reserve references across units using resource availability and operating constraints, accounting for unit-level technical and economic aspects while minimizing VPP operating cost.The approach combines economic scheduling with real-time allocation between units.Source: §VI · p. 9
  • Under ideal and uncertain forecasts, frequency was restored after VPP-reference or system-load steps, while VPP power tracked its day-ahead commitment and unit burden reflected availability and technical constraints.The system-frequency and VPP-power response under forecast uncertainty is shown in Fig. 10; unit power profiles under uncertainty are shown in Fig. 11.Source: §VI · pp. 9–10, Figure 10 · p. 9, Figure 11 · p. 9
  • Downward availability factors were more stable than upward factors; renewables ran near capacity except under technical constraints, while BESS adjusted for renewable forecast deviations to preserve VPP reference tracking.Renewables also provide upward reserves when required for commitments or when more profitable.Source: §VI · p. 10
  • Upward availability can become intermittent near source maximum, but this generally affects only the VPP’s smaller frequency-restoration contribution.Renewable sources provide upward reserves when needed to meet commitments or when that option is more profitable.Source: §VI · p. 10
  • Future work will include VPP power flows and losses, plus internal constraints such as node voltages and line loading.Source: §VI · p. 10

Related papers

Loading 2609.31017v1…