Optimal Residential Battery Management System Using Artificial Intelligence
Omid Alavi *† , Jacopo Sala *† , Robin Schrijvers § , Philippe Haldermans § , Tom Quareme § , Michael Daenen *† , Koen Gilissen §
* IMO-IMOMEC , Hasselt University , Wetenschapspark 1 , 3590 Diepenbeek , Belgium † EnergyVille , Thor Park 8310 , 3600 Genk , Belgium
§
PXL Smart ICT , PXL University of Applied Sciences and Arts , Elfde Liniestraat 24 , 3500 Hasselt , Belgium michael . daenen @ uhasselt . be , koen . gilissen @ pxl . be
Abstract — This study proposes an optimal residential battery scheduling system that combines the optimization problem for an energy management system with artificial intelligence for PV yield and load predictions . The most effective artificial intelligence algorithm has been chosen for this study after assessing several artificial intelligence algorithms for PV yield and load forecasting in Belgium . To find the best battery scheduling and grid-tied PVbattery system control , the output data from artificial intelligence is used as input to the optimization algorithm in the next phase . The capacity tariff , which was not given much consideration in prior studies , has also been rigorously used in this analysis . The results show that combining Sandia and NREL PVWatts models accurately predicted the PV yield , while the Elastic Net model gives the best results for load forecasting . The optimization results also show that by applying a variable rate of hourly electricity , the profitability of residential battery storage can be significantly increased .
Keywords — Artificial intelligence , capacity tariff , maximizing self-consumption , optimization , time series prediction .
I . INTRODUCTION
A key component of smart grid technology and a prerequisite for the transition to a low-carbon economy is the forecasting of electric load and photovoltaic ( PV ) production . The demand for more efficient and risk-averse energy management is only growing as a result of the global sustainable energy transition and crises . The escalating electrification and decentralization of power production present the Distribution System Operator ( DSO ) with serious issues related to power grid congestion . The growing need for innovative storage technologies demands the introduction of advanced digital assistants . The design of the next generation of intelligent energy storage systems ( ESSs ) is made possible by the digital transformation occurring in the energy sector [ 1 ]. Historical and current power consumption data are made available to prosumers , energy providers , and DSOs through digital metering , which is the crucial component for digital transformation . Due to its unpredictable and intermittent nature , it is challenging to match the electrical load with the PV output [ 2 ]. The ability to forecast PV yield at the level of the individual installation is made possible by accurate local weather predictions . Household energy storage solutions , such as batteries and electric vehicles ( EVs ) with the ability to charge in both directions , can reduce power costs for users by utilizing forecasts of electric load and yield [ 3 ].
Numerous studies have evaluated how grid tariffs influence household energy expenses . Azarova et al . [ 4 ] examined the consequences of 11 theoretical network tariffs on 765 households ’ energy costs . By carefully analyzing socioeconomic data , the researchers determined which demographic groups could gain or lose from various tariff models . They stressed the growing disparity between regular consumers and wealthier prosumers . Focusing on the possible advantages for residential solar panel owners , Dargouth et al . [ 5 ] explored the effects of alterations in wholesale market structures on electricity market prices and how this impacts the savings on energy bills for solar panel owners . Ren et al . [ 6 ] assessed the financial rewards of employing a solar panel and battery system under nine distinct tariff scenarios , incorporating three network components ( fixed charges , capacity-based , and peak demand ) and three retail energy components ( flat , time-of-use , and critical peak price ). They discovered that the most significant savings on energy bills were obtained through capacity-based and critical peak price energy rates . These investigations highlight the significance of comprehending how grid tariffs and market configurations affect household energy expenditures , especially for solar panel owners and those considering such installations . A study in [ 7 ] proposes a mixed-integer-linear programming approach for the efficient design and operation of PV and battery systems . It assesses the effects of various tariff scenarios on the economic viability of privately-owned energy systems and their grid usage intensity . Five different tariff scenarios , such as real-time pricing , capacity-based tariffs , and block rate tariffs , are benchmarked . The findings suggest that block rate tariffs are the most promising approach . Nonetheless , examining Belgium ’ s grid tariff scheme for Flemish PV-battery system owners is crucial to determine whether the current capacity-based tariff can contribute to electricity savings for residential households .
In this study , an intelligent ESS is proposed to support users in optimally planning their electricity consumption and peak shaving . The framework comprises an integrated structure for data pre-processing , a module for forecasting PV yield and electric load , an energy management module based on current and projected tariffs , and a battery management system . The data predicted by the artificial intelligence ( AI ) algorithm is used as input for the optimization program to schedule battery charging / discharging , and thus , lower electricity bills . Introducing a variable price in this scenario for the electrical component and adding a variable " Capacity Tariff " to the power
XXX-X-XXXX-XXXX-X / XX /$ XX . 00 © 20XX IEEE