The European energy market has experienced heightened volatility driven by geopolitical tensions, infrastructure disruptions, and an accelerated transition toward renewable energy. This thesis investigates the use of Generative Artificial Intelligence (GenAI), specifically Large Language Models (LLMs), to predict price movements in critical energy commodities, including electricity and natural gas. In collaboration with A2A S.p.A., as part of the Fast Intelligence project — an initiative led by the Trading and Execution office, designed to integrate GenAI-powered algorithms into trading strategies within the energy futures markets — this study leverages LLMs—such as GPT-4, Claude, Gemini, and Llama—to analyze real-time news events and anticipate market dynamics. The research utilizes a dataset from January to May 2023, a period marked by significant geopolitical and market-shaping events, to identify a validation model suitable for realtime analysis. The AI models assign impact scores to news items, correlating them with price fluctuations, offering insights into how external factors drive market behavior. A comprehensive evaluation of model performance revealed that GPT-4 and Claude deliver the most accurate and actionable predictions, providing critical support for traders in decision-making and risk management during volatile conditions. This work highlights the transformative potential of AI in processing vast amounts of unstructured data to generate reliable predictions, paving the way for more sophisticated energy trading strategies. The results demonstrate that GenAI can be effectively integrated into futures trading operations, optimizing market analysis and improving forecast precision. These findings suggest that further advancements in AI could significantly enhance risk management and decision-making in energy markets, facilitating broader AI adoption across the sector.
Negli ultimi anni, il mercato energetico europeo ha vissuto una forte volatilità, influenzato da tensioni geopolitiche, interruzioni infrastrutturali e dalla transizione verso le energie rinnovabili. L’Intelligenza Artificiale Generativa (GenAI), con i suoi modelli linguistici avanzati (LLMs), rappresenta un’innovazione fondamentale per prevedere le oscillazioni dei prezzi nelle principali commodities energetiche, come l’elettricità e il gas naturale. Nell’ambito del progetto Fast Intelligence, sviluppato in collaborazione con A2A S.p.A. e guidato dall’ufficio Trading and Execution, che mira a integrare algoritmi basati su GenAI nelle strategie di trading sui mercati dei futures energetici, sono stati impiegati modelli come GPT-4, Claude, Gemini e Llama per analizzare una vasta gamma di notizie in tempo reale, inclusi eventi geopolitici e sviluppi tecnici nel settore energetico, al fine di anticipare le dinamiche di mercato. L’analisi ha considerato un periodo particolarmente significativo, da gennaio a maggio 2023, segnato da eventi geopolitici di grande impatto sul mercato energetico europeo. I modelli hanno assegnato punteggi di impatto alle notizie, correlando queste informazioni con le oscillazioni dei prezzi e fornendo indicazioni precise su come fattori esterni possano influenzare i movimenti di mercato, migliorando così l’accuratezza delle previsioni. I risultati hanno evidenziato l’efficacia di GPT-4 e Claude, che hanno offerto un supporto strategico fondamentale per le decisioni di trading e la gestione del rischio in contesti ad alta volatilità. L’uso dell’intelligenza artificiale per la gestione di grandi volumi di dati non strutturati apre nuove prospettive per strategie di trading sempre più sofisticate e mirate. I risultati confermano che l’integrazione di GenAI nel trading di futures energetici e nell’analisi di mercato ottimizza l’efficienza decisionale, fornendo previsioni tempestive e accurate. Questo approccio non solo migliora la comprensione delle dinamiche del mercato energetico europeo, ma offre anche un modello metodologico adattabile a contesti di mercato diversi. Infine, si propongono ulteriori sviluppi per perfezionare la precisione predittiva e si esplorano le implicazioni future dell’intelligenza artificiale nel settore energetico.
The Next Frontier in News-Based Market Predictions: Generative AI Applications and Challenges in the European Energy Sector - A Case Study in A2A
PELLEGATTA, ALESSANDRO;Parravicini, Pierpaolo
2023/2024
Abstract
The European energy market has experienced heightened volatility driven by geopolitical tensions, infrastructure disruptions, and an accelerated transition toward renewable energy. This thesis investigates the use of Generative Artificial Intelligence (GenAI), specifically Large Language Models (LLMs), to predict price movements in critical energy commodities, including electricity and natural gas. In collaboration with A2A S.p.A., as part of the Fast Intelligence project — an initiative led by the Trading and Execution office, designed to integrate GenAI-powered algorithms into trading strategies within the energy futures markets — this study leverages LLMs—such as GPT-4, Claude, Gemini, and Llama—to analyze real-time news events and anticipate market dynamics. The research utilizes a dataset from January to May 2023, a period marked by significant geopolitical and market-shaping events, to identify a validation model suitable for realtime analysis. The AI models assign impact scores to news items, correlating them with price fluctuations, offering insights into how external factors drive market behavior. A comprehensive evaluation of model performance revealed that GPT-4 and Claude deliver the most accurate and actionable predictions, providing critical support for traders in decision-making and risk management during volatile conditions. This work highlights the transformative potential of AI in processing vast amounts of unstructured data to generate reliable predictions, paving the way for more sophisticated energy trading strategies. The results demonstrate that GenAI can be effectively integrated into futures trading operations, optimizing market analysis and improving forecast precision. These findings suggest that further advancements in AI could significantly enhance risk management and decision-making in energy markets, facilitating broader AI adoption across the sector.File | Dimensione | Formato | |
---|---|---|---|
2024_10_Pellegatta_Parravicini_Tesi.pdf
non accessibile
Descrizione: testo della tesi
Dimensione
6.54 MB
Formato
Adobe PDF
|
6.54 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/226855