Optimal assortment of methods to mitigate the imbalance power in the day-ahead market

Amid an increase in global electricity demand and concerns over climate change, fossil fuels are driven primarily by a lack of supply. To overcome these concerns, distributed energy resources (DERs) are considered an alternative. However, DERs produce uncontrollable and stochastic power output, which creates imbalances in generation and demand profiles. Further, errors introduced due to forecasting of power output from photovoltaic (PV), wind turbine (WT), load profile, and electricity prices also create an imbalance power (IP). During the forecasting stage, the presence of outliers in the historical datasets impacts the accuracy and performance of the forecasting model; therefore, there is a need to detect and impute these outliers into the actual data before forecasting. A two-stage scheduling methodology was proposed to mitigate the IP arising in the real-time market due to uncertainties and errors in forecasting. In the first stage, the outliers are detected using K-means and autoencoder, imputing these outliers with the neighbors using K-nearest neighbor, and in the next stage, dispatch schedule of each generator was found to mitigate the difference between forecasted and actual values. The hybrid approach starts with data segmentation and applies K-means clustering, which divides data points that are associated with clusters and helps in the identification of underlying patterns and anomalies. After that, anomalies within each cluster are found using an autoencoder neural network, which becomes more adept at identifying intricate nonlinear relationships in the data. The reconstruction error of the autoencoder is used to identify abnormalities. The KNN technique makes sure that values that are imputed are relevant to the context and do not add bias to the dataset. The approach is validated by using an IEEE-33 bus system where the analysis of the impact of electricity price variation on the operational expenditure and the impact of variation in load demand on operational cost has been analyzed.