Stochastic programming for home energy management system optimization

The European Union set the 2020 goals to reduce GHG emissions, increase energy efficiency and d ecarboniz e the energy supply . Home Energy Management Systems (HEMS) are a source of flexibility and a cost - effective strategy to pursue the integration of Renewable Energy Sources (RES) However, the behavior of the component s involved in the residential energy consumption is highly uncertain which poses a challenge for HEM S . A strategy to tackle such uncertain parameters is s tochastic optimization . Even though its formulation dates back decades, only the recent rise of adequ ate technology and software have made possible its implementation. This work proposes to use stochastic optimization in HEMS. The requirements for Stochastic Programming (SP) consist of a mathematical model, a scenario tree and different data instances. The mathematical model is based o n the INVADE project and was implemented using Pyomo. The stochastic formulation of the problem was done with PySP, the Pyomo extension for SP. The scenario tree was created using NetworkX and the data instances were formul ated with Dplyr . A household with an inflexible load, PV generation, a battery and a connection to the grid was considered. Three case studies were analyzed, to gain insight on the impact of different stochastic parameters. Furthermore, a sensitivity analy sis regarding the nature of the stochastic input data was performed. Afterwards , a detailed description and analysis regarding the use of the software was done . Finally , the environmental impact of the project was assessed. In conclusion the stochastic and deterministic formulations are equivalent for the present work due to the high flexibility of the grid. The Value of Stochastic Solution (VSS) was of around - 0.04 € . However, this value increases with increasing standard deviation ( σ) of the input data. T he algorithm schedules the HEMS components in response to the market price signals. The main source of flexibility is the grid, followed by the battery. Generation curtailment is also attractive and is scheduled in all simulations. Lastly, feed in to the g rid is the least attractive flexibility mechanism. The expected flexibility was of 0.381 € per day. R and Python proved to be simple and powerful. Furthermore, Pyomo is ideal to translate models into python objects. PySP has its advantages and drawback, al though some of the last ones were circumvent by using NetworkX and Pandas. This algorithm has an estimated mitigation potential of 1gCO 2eq /per day for the case studies analyzed. On the other hand, the environmental impact created during the realization of this project was of 115 kgCO 2eq