A machine-learning study sorts Moroccan households by energy use and links the groups to income and poverty status. It gives planners a way to work without widespread smart-meter data, while installed capacity remains distinct from renewable electricity generated.
Northern coastal regions carry Morocco's strongest household demand. Much of the country's renewable potential lies in the southeast. Industrial development and desalination also shape the power system. Transport electrification adds pressure, as do major projects beyond household use.
Researchers Safaa Safouan and Karim El Moutaouakil of Sidi Mohamed Ben Abdellah University in Taza used household survey data to sort households into three consumption profiles. One group has low consumption; another has medium consumption. The third has high consumption. The profiles track with income and poverty status. The analysis also distinguishes urban from rural households. The data came from Morocco's Household Consumption and Living Conditions Survey, collected by the High Commission for Planning, or HCP, the national statistical agency.
That survey records household characteristics, expenditure and living conditions, not continuous electricity readings. The model can organize the information available, but it cannot turn those records into smart-meter data or establish the precise electricity use of each household. HCP's survey infrastructure can help researchers study differences between households. Direct measurement would still be needed to validate consumption estimates.
Published in Neural Computing and Applications, the study introduces the Enhanced Fractional Probabilistic Self-Organizing Map, or EF-PRSOM. The framework clusters patterns in complex data. It also adds a fractional derivative that gives its optimization process memory of earlier steps. A genetic algorithm selects the fractional-order parameter automatically.
Before applying EF-PRSOM in Morocco, the researchers tested it against a French household electricity dataset from the UCI machine learning repository. The model recorded a Silhouette score of 0.6709 and a Davies-Bouldin Index of 0.8769. Against the plain probabilistic self-organizing map, its Silhouette score rose by 124.8 percent. Its Davies-Bouldin Index fell by 84.2 percent. The study also reports better results than K-means. Standard self-organizing maps scored lower, as did Gaussian mixture models.
In Morocco, the three clusters were significantly associated with income, poverty status and urban or rural residence, at p < 0.001. Poor and rural households make up a disproportionate share of the low-consumption group. Low use alone does not explain why a household consumes less. The poverty link gives policymakers a stronger basis for identifying where affordability or access deserves attention. These are associations, not proof that poverty or location causes a particular level of consumption.
National power figures show why household analysis needs to be read alongside system-level measures. By the end of 2025, renewables accounted for 46.1% of Morocco's installed generating capacity but supplied 24.3% of national electricity demand, according to reported figures. An independent World Power Monitor estimate puts low-carbon generation at 23.2% of 2025 output. Its estimate for fossil-fuel generation is 76.8%. Those measures are not interchangeable: installed capacity does not show how much electricity was generated or delivered.
Total installed capacity reached 12,314 MW at the end of 2025, including 5,630 MW of renewable capacity. Wind accounted for 2,452 MW. Solar accounted for 1,086 MW, split between 546 MW of photovoltaic capacity and 540 MW of concentrated solar capacity. Morocco's 2025-2030 national equipment plan envisages about 120 billion dirhams in additional investment. It targets total capacity of 27,646 MW by 2030, with roughly 80% of new capacity expected to come from renewables and storage. The stated goal is to exceed 52% renewables in installed capacity by 2030. Medi1News reported on the target and MASEN's role as a strategic integrator of the transition. Capacity growth alone does not guarantee a matching rise in renewable electricity supplied, particularly as demand expands.
The regional mismatch has consequences for transmission planning. Demand is strongest in the north, while major renewable resources are concentrated in the southeast. Connecting supply potential to demand means accounting for where households use energy, not relying only on broad regional averages. The study uses Moroccan survey data; its clusters should not be assumed to describe households in Algeria, Tunisia, Libya or Mauritania without comparable country-level evidence.
That household view complements the infrastructure questions in Morocco infrastructure analysis. Profiles pairing consumption with poverty status, while accounting for location, could help authorities target subsidies and efficiency programmes more precisely. They could also inform clean-cooking initiatives. The study presents a planning tool; it does not report that authorities have introduced any of these measures.
