Pre-processing : For our analysis , we have utilized the Open Fluvius dataset as our primary data source . This dataset is of high quality , but still contains some missing data , duplicates , and time intervals that do not correspond accurately to quarterhour values . These erroneous values constitute approximately 0.23 % of the data . A preprocessing algorithm is introduced , which iterates through the data per household , starting at the oldest timestamp value , and ( 1 ) deletes duplicate rows by retaining only the first occurrence of a timestamp , ( 2 ) implements Last Observation Carried Forward ( LOCF [ 21 ]) on rows where a single kWh value is missing ( with or without a missing timestamp ), ( 3 ) adjusts timestamps not corresponding to a 15-minute interval to their closest quarter neighbor , and ( 4 ) applies single imputation based on mean substitution [ 22 ], with the time of day and day of the week as categories to fill in multiple consecutive missing values . After the preprocessing step , the dataset is split into a training and test set , where the test data consists of the last 90 days of the dataset .
IV . OPTIMAL BATTERY MANAGEMENT SYSTEM
By employing predicted PV output and load data , we can leverage an optimal battery controller to reduce the prosumer ’ s electricity bills . It is crucial to examine this methodology under various circumstances in order to gain a thorough understanding of the advantages offered by this approach . Two different electricity prices have been considered in this study : the hourly electricity tariff ( variable ) and the retail energy-flat ( fixed ). It should be noted that in addition to this power rate , additional costs including taxes and capacity charges will also be incurred . Capacity tariff is determined based on the peak power consumption , and it works well to alter the behavior of energy prosumers by preventing high electricity usage peaks . Households that use a traditional electricity meter will be required to pay a fixed amount as a capacity tariff . The prosumer must pay 96 EUR / year for the yearly minimum peak consumption of 2.5 kW ( CTminimal ), with the capacity tariff increasing linearly for each additional kW peak ( 48 EUR / kW ). The economic evaluation of a PV panel system with battery storage while considering different tariff structures is one of the main goals here . The findings of this study might be very important in choosing the right route for pricing in the future ( grid electricity price , feed-in tariff , and battery price ). The considered electricity tariffs are shown in Fig . 1 .
The chosen load profile exhibits an annual consumption of 3.5 MWh . Based on this consumption , the sizes of the PV system and battery have been determined , with 1 kWpPV and 0.75 kWh battery capacity allocated for every 1 MWh of load . Residential battery storage ( GivEnergy ECO 2.6 kWh / 2.88 kW ) is selected in this study . The injection limit is set to 3 kW . In the first step , a maximizing self-consumption approach is implemented in which the battery will be charged by the surplus PV generation and discharged when the demand is greater than the PV generation [ 23 ]. This control system is simulated in the MATLAB environment . The early findings demonstrate that the inverter control system installed in the residential unit under study is regulated in accordance with the maximizing selfconsumption method by correlating the results of this method with smart meter measurements .
Fig . 1 Electricity price ( a ) hourly electricity tariff , and ( b ) fixed tariff .
The state-of-charge of the battery is identified based on the conventional linearized method . The capacity tariff should be determined according to the maximum power consumption . In the next step , optimization is provided in the Julia programming language . The Gurobi solver is chosen as the optimizer . The objective function for this optimization is defined as Eq . ( 1 ).
91 min @ - AE 234 ) 53 ! 6 ( t ) ∗ Price 78 ' ( t ) H
:#$
( a )
( b )
− ( E : 453 ! 6 ( t ) ∗ Price ;<== ( t )) + ( U > ( 2 ) ∗ ( maxP − 2.5 ) ∗ € 48 ) Q
Thus , it is essential to find the maximum absorbed energy from the grid ( maxP ) by linearizing a “ Max ” function in the constraints . The capacity tariff is then executed as follows :
"
( 1 )
! P ! ( l ) = max P ( 2 )
#$%
CT &'( ( l ) × U ! ( l ) ≤ P ! ( l ) ≤ CT &)* ( l ) × U ! ( l ) ( 3 )
CT &'( = [ 0 ( 2.5 + eps )] ( 4 ) CT &)* = [ 2.5 inf ] ( 5 )
+
! U ! ( l ) = 1 ( 6 )
#$%
Peak consumption exceeding 2.5 kW will result in the value of UL ( 2 ) being equal to 1 , at which point the additional linear capacity tariff will be applied .