Journal articleIFPRI3; ISI; CRP2; CRP5; CRP7EPTD; DSGD; PIMPRCGIAR Research Programs on Policies, Institutions, and Markets (PIM); CGIAR Research Program on Water, Land and Ecosystems (WLE); CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS
Efficient and transparent operation of power generation systems is critical to ensuring economic stability within electricity markets; however, anomalies in pricing and dispatch frequently remain undetected, particularly in environments challenged by limited data availability or governance concerns. To mitigate this pressing issue, we propose a novel, data-driven anomaly detection framework that integrates Autoencoder-based reconstruction error analysis, and linear programming optimization to systematically identify and quantify inefficiencies. Our methodology integrates data preprocessing, anomaly scoring, and cost comparison within a scalable and interpretable pipeline. Moreover, this semi-automated workflow seamlessly combines data cleansing, optimization modeling, and anomaly visualization, thereby delivering precise and actionable insights tailored for both technical practitioners and policy decision-makers. Experimental results using real data from the Bangladesh Power Development Board (BPDB) convincingly demonstrate the framework’s ability to detect anomalies in electricity pricing data while substantially enhancing transparency and fostering stakeholder trust. The framework was also validated by experimenting with different plant factors and calculating optimum per unit and total electricity cost for the energy generated by 20 real power plants. The best optimal average per unit cost is found to be approximately 33% lower than the actual average per unit cost.