Morocco has set ambitious targets for renewable electricity, aiming to supply over half of its power from renewable sources by 2030 [1]. At the same time, the building sector (residential and tertiary) remains a major driver of energy demand, reported at 32% of national final energy consumption, which makes demand-side efficiency a central lever for meeting these transition objectives [2]. Morocco also benefits from strong solar potential: annual global horizontal irradiance commonly falls in the 5 – 6 kWh/m²day range across much of the country [3], positioning rooftop photovoltaic (PV) systems as a credible option to reduce grid dependence and enhance household energy resilience. Realising this potential in the residential sector, however, typically requires PV deployment to be coordinated with envelope-level efficiency improvements and storage, so that generation is better aligned with household load patterns and local market rules.

Recent Morocco-specific transmission analyses further show that higher renewable penetration can significantly reshape power flows and reveal bottlenecks, reinforcing the value of coordinated distributed PV and flexibility resources [4]. Residential PV adoption is therefore shaped not only by technology costs, but also by retail pricing and prosumer regulations. The national utility (ONEE) offers an optional time-of-use (TOU) tariff for eligible low-voltage consumers, introducing a price signal that can increase the value of storage by shifting PV-derived electricity from daytime production toward higher-value evening periods [5]. Recent studies further indicate that the economic value of residential batteries depends strongly on tariff structure, export compensation, and dispatch strategy under time-varying prices [6]. More broadly, scenario-based policy analyses show that incentive design and market rules can materially change system feasibility and the distribution of benefits across configurations, which motivates treating tariff and export assumptions explicitly in optimization [7]. In parallel, PV economics depends strongly on how surplus exports are treated and on the operating conditions that apply to residential storage within the self-production framework. The regulatory environment further complicates PV economics. ANRE Decision No. 04/26 sets the tariff for surplus electrical energy produced under Laws No. 40-19 and No. 82-21, while surplus sales remain limited to 20% of annual production [8]. Implementing decrees were still being drafted at the time of the latest ANRE report, and no standard feed-in tariff has been promulgated for the low-voltage residential case. As a result, households effectively operate under a self-consumption paradigm: exported energy is either curtailed or accepted by the grid without remuneration. In addition, Morocco prohibits charging residential batteries from the grid. These constraints partial TOU participation, capped surplus sales, and no grid-to-battery charging necessitate design methods that maximise self-consumption while respecting regulatory limits.

Building-energy research in North Africa has progressed along two largely independent trajectories: building-envelope optimisation and PV – battery system optimisation. In the first trajectory, meta-heuristic approaches are widely used to explore interacting passive measures and their combined impact on end-use demand. De Oliveira et al. report that retrofit strategies to improve building energy efficiency and sustainability commonly combine envelope insulation, improvements to climatisation and lighting systems, and the integration of renewable energy sources, with the preferred measures varying according to climate and building characteristics [9]. In Morocco, Abdou et al. apply multi-objective optimisation of passive energy-efficiency measures for net-zero energy building design, explicitly balancing demand reduction and performance trade-offs within dynamic simulation [10]. Recent Moroccan studies extend this line of work using climate-zone-specific modelling and multi-objective formulations: Boumlik et al. optimise insulation and glazing across six Moroccan climate zones and discuss pathways toward near/net-zero performance in residential buildings [11], while Benaddi et al. couple TRNSYS with GenOpt to identify envelope solutions that balance economic, environmental, and thermal-comfort criteria across the same climatic diversity [12]. More broadly, recent systematic evidence indicates that GA-based multi-objective retrofit optimisation is becoming increasingly sophisticated, particularly through the continued dominance of NSGA-II and the growing integration of life-cycle analysis and dynamic simulation tools [13]. Nevertheless, this envelope-optimisation stream is rarely coupled to PV – battery sizing and dispatch under time-varying retail prices and prosumer constraints, even though storage value is inherently time-dependent.

The second trajectory focuses on PV–battery sizing and dispatch, most commonly formulated as mixed-integer linear programming (MILP). Early tariff-aware work by Jakus et al. showed that jointly optimising PV and battery capacities with coordinated scheduling across alternative retail tariff structures can deliver substantial bill reductions, particularly under net billing and dynamic pricing [14]. Subsequent residential studies showed that battery economics are highly sensitive to electricity price variability, whereas well-sized PV-only systems remained the more financially robust option under varying electricity-price conditions [15]. Recent building-focused studies further treat PV – battery sizing as a constraint-aware optimisation problem using time-resolved demand: Yu et al. optimise PV-battery configurations for existing buildings under transformer-capacity constraints, while a recent Buildings study applies NSGA-II to size PV-battery systems for detached houses using hourly load profiles and testing robustness under different electricity-price conditions [16], [17]. More recent contributions have strengthened modelling realism by incorporating behind-the-meter uncertainty and degradation-aware cycling limits in the sizing problem [18], and by extending MILP-based home energy management models to represent battery cycle degradation more explicitly, including the influence of operating patterns on battery degradation [19]. In parallel, Moroccan studies have begun to provide locally grounded evidence on residential energy performance and management under national conditions, including the effects of climate-dependent construction practices and thermal-regulation compliance [20], together with forecast-based, tariff-aware optimisation of PV – battery operation in residential smart microgrids under the Moroccan pricing framework [21]. Recent work further shows that day-ahead MILP-based battery scheduling can improve economic outcomes while preserving grid stability when forecasted PV generation, tariff structures, and voltage constraints are incorporated explicitly into the optimisation framework [22]. Beyond single dwellings, recent studies on multi-story and community-scale settings show that coordinated or shared storage can improve solar utilisation, reduce grid dependence, and strengthen the economic value of collective self-consumption [23], [24]. Complementary optimisation-driven work also shows that energy-flow scheduling can materially improve performance indicators in PV-based systems; for instance, Dadjiogou et al. demonstrate PSO-based energy-flow management for a PV microgrid supplying multiple loads [25].

Despite these advances, most PV – battery optimisation studies still treat building demand as an exogenous input and rarely couple envelope-driven load changes with prosumer-policy constraints and operationally feasible battery dispatch, leaving the synergy between passive demand reduction and cost-optimal PV – battery design insufficiently explored.

A further limitation in the PV – battery literature is that “optimisation” is sometimes used loosely: several studies effectively compare a small set of predefined scenarios, while others employ simplified formulations that can produce dispatch patterns that are not operationally feasible. More broadly, this reinforces a key need in residential prosumer studies: PV – battery sizing should be assessed with (i) realistic, time-resolved building demand and (ii) policy- and tariff-consistent operational constraints, because storage value is fundamentally time-dependent.

To address these gaps, this paper proposes a coupled workflow in which a GA-based envelope optimisation generates an hourly demand profile that feeds a policy-consistent MILP for PV – battery sizing and dispatch under time-varying retail prices and prosumer export rules. The framework is designed to produce operationally valid schedules and to represent key policy levers that shape household economics. Using this setup, the analysis quantifies where additional PV or storage yields diminishing marginal benefit and assesses how alternative export-remuneration assumptions shift the cost-optimal design relative to a self-consumption baseline. By integrating envelope optimisation with dispatch-constrained PV–battery optimisation, the resulting capacity choices can differ materially from approaches that treat demand reduction and supply sizing as separate problems.

A two-stage optimisation framework is developed to produce operationally credible and policy-compliant photovoltaic–battery designs by coupling envelope-driven demand reduction with mixed-integer linear programming dispatch and sizing. First, a multi-objective genetic algorithm (GA) implemented in DesignBuilder with the EnergyPlus simulation engine explores feasible combinations of insulation and glazing for a representative dwelling in Oujda. The optimisation seeks to reduce annual electricity demand for heating and cooling while respecting thermal comfort constraints and limiting additional construction cost. This stage produces an hourly electricity load profile for a full year (8,760 hours). Second, a mixed-integer linear programming (MILP) model selects the photovoltaic capacity and the battery capacity and optimises hourly energy routing between photovoltaic generation, the battery, and the grid under Morocco’s time-of-use tariff and the assumed export conditions. Battery operation is represented using state-of-charge dynamics with a cyclic boundary condition, charging and discharging efficiency losses, and mutually exclusive operating modes enforced through binary variables with a big-M formulation linked to the battery power limit. Policy switches represent alternative regulatory settings, including disabling grid-to-battery charging and changing export remuneration assumptions. A post-optimisation capacity-perturbation analysis is then applied by incrementally increasing photovoltaic and battery sizes and re-optimising dispatch for each case. This analysis is used to identify cost –performance plateaus and quantify the sensitivity of optimal designs to export and tariff settings.

Figure 1.

Overview of the approach

Figure 1 illustrates the general methodological framework adopted in this study, integrating building energy simulation, envelope optimization using GA, and renewable energy system optimization via MILP.

Case Study Description

The case study involves a two-story, single-family urban residential building located in Oujda, within Morocco’s Oriental region (climatic Zone 3), as shown in Figure 3. The building exhibits a typical compact Moroccan residential architecture with a central courtyard designed for daylighting and natural ventilation [2]. It covers approximately 114 m², with the ground floor consisting of a garage, kitchen, reception area, hall, and living room, while the upper floor accommodates bedrooms, a lounge, and bathroom facilities Figure 2.

Figure 2.

Ground floor and first-floor layouts of the case study building

Climate data for Oujda was sourced from the Meteonorm database (Oujda-Angads station, 34.787°N, – 1.924°E, elevation 467.9 m) [29], [30]. The climate characteristics include:

• Peak summer temperatures frequently exceeding 38 °C.

• Annual mean temperatures ranging between 17.5 and 19.2 °C.

• Low annual precipitation (<350 mm/year).

Figure 3.

Morocco climatic zones, indicating the study area within Zone 3 [2]

The baseline energy simulation indicated an Energy Use Intensity (EUI) of approximately 115.84 kWh/m²/year, consistent with regional benchmarks [2], [34]. The monthly electricity demand for heating and cooling, Figure 4 show that space heating is concentrated in the winter months (December – February), while cooling dominates during the summer (June – September), with peak demand in July and August. This seasonal pattern reflects the climatic conditions of Oujda and the thermal characteristics of the uninsulated building.

Figure 4.

Monthly Heating and Cooling Electricity Demand for the Uninsulated Building

Envelope Optimization

Envelope optimisation was performed using a multi-objective GA to explore insulation and glazing alternatives and identify robust trade-offs between annual HVAC electricity consumption and incremental construction costs. In total, 245 envelope configurations were evaluated Figure 5, producing a Pareto set from which a compromise solution was selected for subsequent PV – battery optimisation. The selected envelope consists of 3 cm polystyrene insulation applied across the envelope elements considered in the optimisation and double glazing filled with air (U=2.95 W/m²K).

Figure 5.

Energy Consumption and Construction Cost – GA Optimization Results

The selected envelope reduces annual HVAC electricity consumption by 26.9% relative to the uninsulated baseline Figure 6. The reduction is most pronounced during winter (December – February), reflecting lower transmission losses, while a modest increase in cooling electricity may appear during the warm season (May – September). This trade-off is consistent with the “anti-insulation” effect reported for warm and semi-arid climates when night ventilation and external shading are not enabled, as increased thermal resistance can limit nocturnal heat release and increase the retention of internal and solar gains [35], [36]. Enabling night ventilation and shading would be expected to mitigate this summer effect while preserving winter savings.

Figure 6.

Monthly Total HVAC Electricity Consumption Uninsulated and Optimized Insulated

The annual hourly electricity-demand profile obtained from the selected envelope configuration is used as the input load time series for the PV – battery optimisation model described in the Method section.

Renewable Energy System Optimization

Following envelope optimization, the MILP model was applied using the optimized building load profile to identify optimal sizing and operational strategies for a hybrid PV-battery storage system Figure 7.

Figure 7.

Hourly Electricity Demand Throughout the Year

Optimal system sizing: the mixed-integer linear programme co-sizes capacities with hourly dispatch.

Table 4 reports the cost-optimal designs for S1–S3 and shows a consistent hierarchy aligned with the incremental flexibility introduced across scenarios. Moving from S1 to S2 yields the dominant service improvement because storage relaxes the coincidence constraint between PV generation and demand: SSR increases by +0.302, while annual total cost decreases by 18.6% and LCOE decreases by 18.1%. Moving from S2 to S3 delivers a smaller additional SSR gain but a clear economic improvement through surplus value recovery: annual total cost decreases by 9.8% and LCOE decreases by 9.6%. This ranking reflects the underlying mechanisms: storage increases on-site supply by shifting midday PV into evening demand, whereas paid export primarily reduces net cost by monetising residual PV surplus that cannot be absorbed by the load or battery.

Table 4.

Optimal PV–BESS sizing and KPIs by scenario

Scenario	PV capacity (kW)	BESS capacity (kWh)	SSR	Annual total cost (USD/y)	LCOE (USD/kWh)
S1	8.00	0.00	0.464	1839.13	0.127
S2	10.4	11.8	0.766	1497.17	0.104
S3	10.9	13.5	0.813	1350.92	0.094

A complementary annual energy-balance interpretation reinforces this comparison. For the same annual load (14.45 MWh), the S1 optimum supplies 6.70 MWh/y directly to the load, with the remainder (7.74 MWh/y) purchased from the grid. In S2, storage creates a second on-site supply pathway: PV supplies 7.09 MWh/y directly and the battery contributes 3.98 MWh/y, reducing grid purchases to 3.38 MWh/y. In S3, the same time-shifting channel is strengthened (battery-to-load rises to 4.51 MWh/y) and grid purchases fall further to 2.70 MWh/y, while an additional outlet appears in the form of remunerated exports (2.98 MWh/y). These flow shifts explain why the major SSR increase occurs between S1 and S2, whereas the additional cost reduction in S3 is achieved mainly through monetising surplus rather than through a proportional expansion of on-site supply.

Figure 8.

Daily PV flow components: load, battery, export, curtailment (S1–S3)

The fractional distribution of PV energy destinations Figure 8 provides a mechanism-level explanation for the diminishing returns observed as capacities scale. In S1, the productive sink is strictly limited to contemporaneous demand, leading to a rapid saturation of on-site supply and a corresponding surge in curtailment as capacity increases. S2 introduces the battery as a secondary sink, which effectively captures a significant share of midday excess; however, this time-shifting channel remains bandwidth-limited, and residual surplus re-emerges once the storage absorption capacity is reached. S3 demonstrates the highest resource utilization by incorporating grid export as a systematic outlet for energy that cannot be absorbed by the load or battery. This routing logic illustrates that as the system expands, it progressively shifts from converting incremental PV into avoided grid purchases toward managing increasing volumes of surplus, which inherently drives the observed concavity in SSR gains.

The temporal manifestation of these routing differences is captured in the year-round daily operational profiles Figure 9, which highlight when scenario constraints matter most. S1 is characterized by a "self-sufficiency ceiling" during high-production months, where massive midday surpluses coexist with persistent nocturnal grid reliance due to the lack of intertemporal flexibility. S2 fundamentally restructures this relationship, substituting battery discharge for a large share of non-solar demand and substantially reducing grid interaction during the high-insolation mid-year period. S3 preserves this time-shifting contribution while improving net economics through export monetization; notably, the load-supply structure of S3 remains largely similar to S2, confirming that the principal distinction between these two scenarios is expressed more strongly in cost-based indicators (Total Cost and LCOE) than in physical service delivery. Collectively, these operational dynamics establish a coherent causal chain where scenario-specific constraints determine the PV routing, which in turn governs the achievable SSR and the ultimate economic performance of the system.

Figure 9.

Daily load supply mix: PV, battery, grid (S1–S3)

The cost-optimal photovoltaic and battery capacities obtained in this study (photovoltaic 10.4 – 10.9 kW; battery 11.8 – 13.5 kWh) are larger than the photovoltaic capacities reported for net-zero annual balance in a prototypical Moroccan single-family dwelling with a 90 m² roof, where required photovoltaic capacity ranges from 1.92 to 6.00 kW across six climate zones[11]. This difference is consistent with the higher electrified annual demand represented in this study and with a cost-minimisation formulation that does not impose a net-zero annual constraint. In terms of storage magnitude, the battery energy capacity is comparable to the off-grid residential configuration evaluated for Fez by Mekila Mbayam and Bounahmidi [37], which combines 3.6 kW of installed photovoltaic capacity with approximately 11.9 kWh of lithium iron phosphate battery storage, and this value falls within the optimal battery range identified in this study. By contrast, their photovoltaic capacity is substantially smaller than the range obtained in this study, reflecting differences in modelling objectives and constraints: their sizing is driven by meeting an annual energy-balance target with simulation-based design choices, whereas the present work endogenizes photovoltaic capacity through dispatch-constrained, full-year mixed-integer optimisation under time-of-use pricing and explicit operational limits, which can justify larger photovoltaic installations until marginal value is eroded by surplus and system constraints.

To assess robustness and marginal returns beyond the cost-optimal designs (step 0), a coordinated capacity-expansion sweep is performed.

Figure 10.

Sensitivity summary of PV energy destinations across incremental PV and BESS capacity combinations for Scenarios S1, S2, and S3

Figure 10 provides a mechanism-level explanation for the diminishing returns and clarifies cross-scenario contrasts. At the cost optima (step 0), annual PV generation increases from 12.92 MWh/y (S1) to 16.79 MWh/y (S2) and 17.68 MWh/y (S3). However, additional PV does not translate one-for-one into useful on-site supply because the binding limitation is temporal mismatch rather than annual resource availability. In S1, the productive sink is restricted to contemporaneous demand, so surplus emerges once daytime demand is saturated (approximately 6.22 MWh/y at the optimum, computed as PV generation minus PV-to-load). In S2, storage introduces a second sink: PV-to-battery reaches 4.15 MWh/y at the optimum and reappears as 3.98 MWh/y delivered from the battery to the load (net of charge/discharge losses), which increases on-site supply; nevertheless, the time-shifting channel remains bandwidth-limited and residual surplus persists (approximately 5.56 MWh/y at the optimum, computed as PV generation minus PV-to-load and PV-to-battery). In S3, export provides an additional outlet once the battery absorption saturates: PV-to-grid reaches 2.98 MWh/y at the optimum, reducing residual surplus to approximately 2.77 MWh/y and improving net economics without requiring a proportional increase in on-site self-supply. Across the sweep, these scenario-specific sinks explain the observed concavity: as capacities increase, the system progressively shifts from converting incremental PV into avoided grid purchases toward managing increasing volumes of surplus via export (S3) and/or residual surplus (all scenarios), which inherently yields diminishing SSR gains.

Table 5.

Sensitivity summary of technical and economic performance indicators across PV and battery sizing increments for Scenarios S1–S3

Scenario	Step	PV (kW)	Battery (kWh)	Annual total cost (USD/y)	Savings vs baseline (USD/y)	SSR (–)	LCOE (USD/kWh)
S1	0	8.0	0.0	1,839.13	378.90	0.464	0.127
	3	11.0	0.0	1,893.92	324.11	0.519	0.131
	6	14.0	0.0	1,996.08	221.95	0.545	0.138
	9	17.0	0.0	2,115.06	102.98	0.560	0.146
	11	19.0	0	2,198.12	19.91	0.569	0.152
S2	0	10.4	11.8	1,497.17	720.86	0.766	0.104
	3	13.4	14.8	1,559.28	658.75	0.882	0.108
	6	16.4	17.8	1,700.94	517.08	0.951	0.118
	9	19.4	20.8	1,901.03	317.00	0.985	0.132
	12	22.4	23.8	2,137.62	80.41	0.996	0.148
S3	0	10.9	13.5	1,350.92	867.11	0.813	0.094
	4	14.9	17.5	1,461.01	757.01	0.939	0.101
	8	18.9	21.5	1,714.81	503.21	0.987	0.119
	12	22.9	25.5	2,034.78	183.25	0.998	0.141
	14	24.9	27.5	2,202.77	15.26	0.999	0.153

Table 5 quantifies the cross-scenario trade-offs implied by Figure 8 and highlights where marginal returns begin to collapse. In S1, PV upsizing produces progressively smaller service gains while economic performance weakens: increasing PV from 8 to 17 kW raises SSR by only +0.096, while annual total cost increases by 276 USD/y and LCOE rises by +0.019 USD/kWh; savings compress sharply over the same range, indicating that additional PV increasingly converts into low-value surplus rather than avoided purchases. In S2, early co-upsizing yields large SSR gains at moderate cost for example, from step 0 to step 3, SSR increases by +0.116 for a cost increase of 62 USD/y whereas later steps exhibit clear saturation: from step 9 to step 12, SSR increases by only +0.011 while annual total cost increases by 237 USD/y and LCOE rises from 0.132 to 0.148 USD/kWh. In S3, paid export shifts the economic frontier outward at high SSR levels for example, step 0 to step 4 increases SSR by +0.126 with 110 USD/y additional cost and a small LCOE change (0.094 to 0.101 USD/kWh) but at larger sizes the marginal service gain collapses (step 12 to step 14 adds only +0.001 SSR) and savings approach breakeven (15 USD/y at step 14), signalling that further upsizing is dominated by surplus management rather than additional useful on-site supply. Overall, SSR improvements are concave in all scenarios, and the economic indicators track this saturation: once SSR is high, additional capacity yields limited avoided purchases, so annual total cost and LCOE increase and savings compress toward zero.

Policy sensitivity analysis: feed-in tariff and export allowance.

Figure 11 summarises how the cost-optimal PV–battery design responds to the feed-in tariff (P_exp) and the export allowance (S_exp). Two thresholds are apparent. First, curtailment is rapidly eliminated as S_exp increases: at P_exp = 0.2 and S_exp = 20%, 2,767 kWh of 16,855 kWh PV generation is curtailed (16.4%), whereas curtailment reaches 0.0% for multiple cases from export capacity share in range 35%–40%. Second, a regime shift occurs near S_exp = 70%, beyond which design and operating indicators stabilise and the marginal gains in savings progressively saturate.

Net savings relative to the baseline increase monotonically with export capacity share and with P_exp Figure 11a. At high P_exp, the savings response saturates beyond S_exp = 0.70; at P_exp = 0.8, savings increase from approximately 920 USD/y (S_exp = 0.20) to 2,075 USD/y (S_exp= 0.70), while export revenue plateaus at 2,268.55 USD/y for S_exp ≥ 0.70. Consistently, PV capacity increases with export_cap_share and reaches a stable plateau by export capacity share in range 0.65–0.70 in the high-P_exp cases Figure 11c.

Battery sizing exhibits a discontinuous adjustment at the regime boundary. At P_exp = 0.8, battery capacity decreases from 16.02 kWh at S_exp = 0.65 to 10.53 kWh at S_exp = 0.70, coincident with PV saturation at 25 kW. This shift is mirrored in operating shares: PV -to-battery share decreases from 21.6% at S_exp = 0.65 to 9.2% at S_exp = 0.70 (P_exp = 0.8), while PV-to-grid share approaches its upper plateau (70%) for S_exp ≥ 0.70. The reduced storage contribution increases grid purchases despite higher export monetisation: at P_exp = 0.8, annual grid cost drops to 82.53 USD/y at S_exp = 0.65 but rises to 312.61 USD/y at S_exp = 0.70 and remains stable thereafter Figure 11b.

The post-0.70 behaviour reflects a structural reallocation in the optimal investment and dispatch. Once the permitted export fraction is sufficiently high, the export-share constraint no longer limits the optimum, and remunerated exports provide a strong marginal value for additional photovoltaic generation. The optimiser therefore prioritises PV expansion until it reaches its saturation level, while battery energy capacity is reduced because the incremental benefit of time-shifting surplus PV into evening demand becomes smaller than the benefit of exporting surplus PV directly at the prevailing P_exp. This battery step-down reduces the system’s ability to cover non-solar hours with stored PV, which explains the discontinuous increase in grid purchases. Simultaneously, because PV export revenue already operates near its plateau, further increases in export capacity share yield limited additional value, producing the observed flattening of the savings curve. In effect, S_exp = 0.70 marks the transition to an export-oriented optimum characterised by PV capacity saturation, lower storage dependence, and diminishing marginal gains in annual savings.

Overall, the results show that envelope optimisation materially reduces the space-conditioning electricity demand and establishes the hourly load basis for system sizing. The cost-optimal photovoltaic-only case reaches moderate self-supply, whereas adding battery storage produces the largest increase in self-supply by shifting midday generation to evening demand. Introducing paid export further improves the economic outcome mainly through surplus value recovery, with a smaller additional increase in self-supply. Capacity-expansion sweeps confirm diminishing marginal returns: beyond the optima, additional capacity increasingly converts into surplus rather than avoided grid purchases. Finally, the policy sensitivity indicates that export remuneration and export allowance can materially shift the optimal photovoltaic – battery balance, highlighting the dependence of sizing outcomes on regulatory conditions.

Figure 11.

Sensitivity of the cost-optimal photovoltaic–battery design to export allowance and feed-in tariff: (a) Net savings vs baseline; (b) Annual grid cost.; (c) Photovoltaic capacity; (d) Battery energy capacity; (e) Photovoltaic-to-load share of photovoltaic generation; (f) Photovoltaic-to-battery share of photovoltaic generation

This study provides a policy-aware framework for reducing electricity use and emissions in Moroccan dwellings by coupling envelope demand reduction with cost-optimal rooftop photovoltaic generation and battery storage. Envelope optimisation via a multi-objective genetic algorithm reduces annual space-conditioning electricity demand by 26.9% relative to the baseline, after which a mixed-integer linear programme co-sizes photovoltaic and battery capacities and optimises hourly dispatch under the National Office of Electricity and Drinking Water time-of-use tariff structure, the no grid-to-battery charging rule, and a capped paid-export option.

The results indicate that demand reduction is a prerequisite for effective renewable-system sizing because it lowers and reshapes the hourly load profile. Under zero export remuneration, adding battery storage yields the largest increase in self-supply by shifting midday photovoltaic generation toward evening demand, although gains saturate as storage energy capacity and power limits become binding. Export conditions emerge as the dominant economic lever when remuneration is introduced: policy sensitivity shows a distinct regime transition once the permitted export allowance approaches 70%, after which photovoltaic capacity and operating indicators become comparatively stable while the cost-optimal battery capacity drops discontinuously. This behaviour reflects a shift in value from time-shifting toward direct sale of surplus electricity to the grid, and it explains both the stabilisation of the design and the reduced marginal benefit of further export-allowance increases once export revenues approach a plateau.

The findings support an “efficiency-first, then right-size” strategy. For the Oujda case study, key performance indicators are most favourable near the identified sizing region (about 10 kW photovoltaic and about 12 kWh battery energy capacity), whereas further upsizing delivers diminishing self-supply gains while increasing annual total cost and the levelized cost of electricity. From a regulatory perspective, finalising a bankable export remuneration scheme and clarifying export allowances would shift the economic frontier more strongly than marginal adjustments elsewhere, because these terms determine whether surplus photovoltaic generation is curtailed, stored, or sold to the grid.

Several limitations should be noted in line with the scope and policy context of the analysis. The results are derived from a single representative dwelling in Oujda and therefore quantify system behaviour under one climatic setting and one load profile, rather than providing a nationally representative estimate. In addition, the economic conclusions are policy-contingent: they reflect the tariff structure, export remuneration assumptions, and export-allowance rules assessed, which remain subject to regulatory clarification and future adjustment. Accordingly, the reported optimal sizing and the identified export-driven regime shift should be interpreted as indicative of how design incentives respond to the prevailing regulatory framework, rather than as fixed values that would apply under alternative market conditions or revised grid-integration rules. Nevertheless, the modelling–optimisation workflow is transferable: applying it to other regions requires substituting the relevant climate files (weather data) and regional load profiles, and re-running the sizing optimisation under the local tariff/export scheme and grid-integration rules. Thus, while the numerical optima are case-specific, the approach provides a decision framework that can be adapted to other climates and policy regimes.

Future work should extend the framework to additional Moroccan building archetypes and climates, incorporate multi-year and stochastic inputs, represent flexible loads and demand response, integrate grid-aware constraints, and evaluate robustness to tariff and policy uncertainty (including alternative export remuneration levels and export-allowance rules). Within these bounds, the proposed genetic-algorithm-to-mixed-integer-linear-programming workflow provides a reproducible approach for producing physically credible designs that link envelope efficiency, photovoltaic generation, and battery storage decisions to Morocco’s evolving tariff and export policy environment.