The goal of this project is to develop an energy management method for cities, specifically for managing electrical energy in buildingBuildings. The proposed method involves detecting roofs using deep learningDeep Learning techniques and determining their photovoltaic solar potential. The estimated theoretical solar potential is 363.33 GWh/year, with a technical potential of 291 GWh/year that can be exploited by installing solar panels on the roofs. A sample of 100 roofs provides a technical photovoltaic power of 3 MW equivalent to 24 MWh over eight hours. The study includes an optimization of energy consumption for the sample, comparing two scenarios and resulting in energy savings of 117,720 DH/day (11,735.62 USD/day) in electricity purchases from the National Office of Electricity and Water (ONEE Electricity Branch). The optimization produces an electrical energy management program, which can be uploaded onto an embedded system such as an Arduino card. As a proof of concept, the operation of the Arduino board was simulated on two buildings. To further improve the project, an economic analysis is necessary to assess its profitability. The analysis shows an initial investment of 24.7 million DH (2.40 million USD) and a payback time of 7 months for the scenario with intelligent energy management. Taking into account the Carbon tax avoided, the net present value over a 20-year project duration is estimated to be 682 million DH (68.15 million USD).