In order to plan and manage low-carbon investments in wide real estate assets, strategic approaches should be considered to act on building stocks as a whole, with the aim of overcoming the single-building perspective and identify the energy retrofit level leading to the maximum possible benefit.

However, decarbonization programs of urban compartments require highly detailed information about the energy-use and energy-efficiency-potentials of a huge amount of buildings, which ends up being a problem of mass-appraisal and screening evaluation.

The subject of this study is, therefore, the development of a decision support system for the planning and management of energy retrofit operations in large building portfolios. Energy improvement is here treated as an optimisation problem in which conflicting objectives and constraints are balanced. In order to develop the decision-making model, several techniques from various disciplines including statistics, economics, energy simulation, computer programming, optimisation and risk analysis were combined.

First, a set of four neural networks is developed to assess the energy consumption of buildings due to heating, cooling, hot water and electricity, based on deep learning and artificial intelligence procedures. Next, different energy-retrofit options are suggested, and different possible alternative intervention scenarios are determined, where a scenario represents any combination of the retrofit options on each building in the building stock. Three performance indices are then estimated to assess the benefits produced by each possible retrofit scenario in energy, economic and cultural terms. The energy savings are estimated using the neural networks, the monetary benefits are calculated on the basis of a Life Cycle Costing approach, while the cultural aspects are evaluated in terms of material and architectural compatibility of the retrofit measures with the building; in particular, it is with an Analytic Hierarchy Process, developed by interviewing a panel of ten experts in the field of energy retrofit, that the architectural compatibility of interventions is quantified through the estimation of a 'compatibility score'. It is then with a multi-attribute optimisation strategy that an evolutionary algorithm tests all possible retrofit scenarios until the optimal configuration is identified, i.e. the one that simultaneously maximises the three performance indices, respecting the domino of feasibility. Finally, a Monte Carlo simulation verifies the risk associated with the chosen retrofit configuration.