Batteries represent the beating heart of global energy and economy transition. This concept is well illustrated in Chapter 1 with a brief overview of batteries between past, present and future. In Chapter 2 a battery testing system with software/hardware interface is presented and accurately described in each phase of its realization so as to be easily reproduced in a laboratory as if this chapter were a tutorial. Such a system allows to study the charging and discharging behavior of batteries according to customized cycles, by acquiring data such as voltage and current, as well as taking into account the effect of temperature. In Chapter 3 a real-time estimation algorithm including a proportional-integral (PI) based observer and an equivalent electric circuit model (ECM), is improved exploiting the state of charge (SOC) dependence on the open circuit voltage (OCV). Particularly, it is described how the SOC can be reliably deduced from relaxation voltage in a short time thanks to a prediction function. The integration of such function into the PI-based observer allows to get a good trade-off between accuracy and complexity. The same prediction function is also used for the purpose of estimating the state of health (SOH) obtaining a satisfactory accuracy while limiting the computational burden. In particular, two basic estimation methods, namely the coulomb counting method (CCM) and the ECM, are combined in order to get a real-time mixed algorithm; correctness of results is verified performing experimental comparisons with other common SOH estimation methods. With special reference to the automotive field, Chapter 4 presents a straightforward solution to estimate the SOC of battery packs used to supply low voltage (LV) electric drives integrated in hybrid and electric vehicles. The main idea is exploiting the electric drive to generate suitable DC bus current profiles to estimate the battery pack parameters, and thus the SOC, whenever the electric drive is not used as propulsion unit. In particular, the drive is exploited as controlled variable load to generate current steps and monitor the transient response of the battery pack through the measure of the DC bus voltage and motor phase currents. No additional external circuits are required, and a limited computational burden is required. Simulations and experimental tests confirm the feasibility of the proposed method. In virtue of the attention that the study of Li-ion batteries degradation deserves, Chapter 5 provides a review of the most important battery SOH estimation methods. After sorting all methods into three groups, two approaches were followed for each, highlighting theoretical aspects, strengths, weaknesses and model accuracies. In particular, electrochemical impedance spectroscopy (EIS) and incremental capacity analysis (ICA) were firstly considered among experimental methods. Secondly, equivalent ECMs and aging models (AMs) among model-based methods. Finally, neural networks (NNs) and support vector regression (SVR) among data-driven methods. With the same attention to the degradation process and, therefore, to the prognostic evaluation of a battery, Chapter 6 consists of an in-depth study in the literature on how to identify the RUL according to the so-called change point analysis. Contextually, with a view to an increasingly efficient integration of batteries into renewable plants, aspects related to photovoltaic (PV) plants have been studied. Hence, Chapter 7 focuses on the performance assessment of large PV plants using an integrated state-space average modeling approach. A comparison between central and string inverters is also presented as well as a study on mismatch losses. Distributed power converters represent a technical solution to improve the performance of large or utility-scale PV plants. Unfortunately, evaluation of the yield obtained in large PV fields by using distributed converters is a difficult task because of recurring partial unavailability, inaccuracy of power analyzers, operating constraints imposed by the power plant controller (PPC) and so on. To overcome such issues in real operating scenarios, a new modeling strategy has been introduced and validated in terms of computational complexity and accuracy. This approach is based on the state-space averaging technique which is applied to large PV plants with multiple conversion stages by performing some elaborations in order to get a final integrated model. The new modeling strategy has been tested in MatLab/Simulink environment using data coming from a 300 MW PV plant located in Brazil representing the case study of this work. In this plant, one subfield is equipped with central inverters while another is with string inverters. The proposed model, whose accuracy is in the range from 2.2 to 2.7% with respect to the measured energy, effectively supports data analysis leading to a consistent performance assessment for the distributed conversion system. Final results highlight that string inverters ensure a gain of about 2% in terms of produced energy. This particular power configuration allows also the calculation of mismatch at string level as well as at array level at the DC side input of each string inverter. The mismatch effect is of fundamental importance in the operation of PV power plants because it causes significant losses in energy production. The mismatch originates from several factors such as non-uniform modules aging, shading, dust accumulation, faults in tracker systems and so on. Evaluation of mismatch level is a difficult task because it is necessary to take into account many factors: power configuration, geographical position, weather conditions, quality of measurements provided by dataloggers, etc. In this case, a multi-criteria approach has been implemented with the purpose to evaluate the coherency of results. Finally, Chapter 8 shows how to measure active power as the difference between the peak value of instantaneous power and the apparent power. The traditional approach to calculate active and reactive power in AC power systems requires the measurement of phase shift between voltage and current for the evaluation of power factor. To do this, power analyzers can implement several methods. In principle, it is always necessary to identify specific points of waveforms (e.g., using a zero-crossing detection technique) and get their time shift. In a similar way, frequency value must be evaluated in order to calculate angular frequency. Unfortunately, this kind of common methods exhibits some issues as the large sensitivity to noise. Moreover, inaccuracies in the evaluation of power factor have a big impact on the final estimation of electric power. On the contrary, the practical implementation of the proposed formulation in power analyzers guarantees several benefits without reducing accuracy.
Le batterie rappresentano il cuore pulsante della transizione energetica ed economica globale. Questo concetto è ben illustrato nel Capitolo 1 con una breve panoramica sulle batterie tra passato, presente e futuro. Nel capitolo 2 viene presentato e descritto accuratamente, in ogni fase della sua realizzazione, come se fosse un tutorial, un sistema per testare le batterie con un’interfaccia software/hardware in modo da poter essere facilmente riprodotto in laboratorio. Tale sistema permette di studiare il comportamento di carica e scarica delle batterie secondo cicli personalizzati, acquisendo dati quali tensione e corrente e tenendo conto anche dell'effetto della temperatura. Nel Capitolo 3 viene implementato un algoritmo di stima in tempo reale che include un osservatore basato sull'integrale proporzionale (PI) e un modello di circuito equivalente (ECM), sfruttando la dipendenza dello stato di carica (SOC) dalla tensione a circuito aperto (OCV). In particolare, viene descritto come il SOC possa essere dedotto in modo affidabile dalla tensione di rilassamento in breve tempo grazie ad una funzione di predizione. L'integrazione di tale funzione nell'osservatore PI consente di ottenere un buon compromesso tra accuratezza e complessità. La stessa funzione di predizione è utilizzata anche allo scopo di stimare lo stato di salute (SOH), ottenendo un'accuratezza soddisfacente con sforzo computazionale limitato. In particolare, due metodi di stima di base, vale a dire il metodo coulomb counting (CCM) e l'ECM, vengono combinati per ottenere un algoritmo misto in tempo reale; la correttezza dei risultati viene verificata eseguendo confronti sperimentali con altri comuni metodi di stima dell’SOH. Con particolare riferimento al settore automobilistico, il Capitolo 4 presenta una soluzione semplice per stimare il SOC dei pacchi batteria utilizzati per alimentare azionamenti elettrici a bassa tensione (LV) integrati in veicoli ibridi ed elettrici. L'idea principale è sfruttare l'azionamento elettrico per generare adeguati profili di corrente del bus DC per stimare i parametri del pacco batteria e, quindi, il SOC, ogni volta che l'azionamento elettrico non viene utilizzato come unità di propulsione. In particolare, l’azionamento viene sfruttato come carico variabile controllato per generare gradini di corrente e monitorare la risposta ai transitori del pacco batteria attraverso la misura della tensione DC del bus e delle correnti di fase del motore. Non sono richiesti circuiti esterni aggiuntivi ed è richiesto uno sforzo computazionale limitato. Simulazioni e prove sperimentali confermano la fattibilità del metodo proposto. In virtù dell'attenzione che merita lo studio del degrado delle batterie al litio, il Capitolo 5 fornisce una rassegna dei più importanti metodi di stima dell’SOH delle batterie. Dopo aver suddiviso tutti i metodi in tre gruppi, sono stati seguiti due approcci per ciascuno, evidenziando aspetti teorici, punti di forza, punti di debolezza e accuratezza dei modelli. In particolare, inizialmente sono stati considerati la spettroscopia di impedenza elettrochimica (EIS) e l'analisi della capacità incrementale (ICA) tra i metodi sperimentali. In secondo luogo, gli ECM ed i modelli di invecchiamento (AM) tra i metodi basati su modelli. Infine, le reti neurali (NN) e la regressione vettoriale di supporto (SVR) tra i metodi data-driven. Con la stessa attenzione al processo di degrado e, quindi, alla valutazione prognostica di una batteria, il Capitolo 6 consiste in un approfondimento in letteratura su come identificare il RUL secondo la cosiddetta change point analysis. Contestualmente, nell'ottica di una sempre più efficiente integrazione delle batterie negli impianti rinnovabili, sono stati studiati gli aspetti legati agli impianti fotovoltaici. Pertanto, il Capitolo 7 si concentra sulla valutazione delle prestazioni di grandi impianti fotovoltaici utilizzando una rappresentazione in spazio di stato di tipo average. Viene inoltre presentato un confronto tra inverter centralizzati e di stringa nonché uno studio sulle perdite di mismatch. I convertitori di potenza distribuiti rappresentano una soluzione tecnica per migliorare le prestazioni di impianti fotovoltaici di grandi dimensioni o su scala industriale. Sfortunatamente, la valutazione del rendimento ottenuto in grandi campi fotovoltaici utilizzando convertitori distribuiti è un compito difficile. Per superare tali difficoltà in scenari operativi reali, è stata introdotta e validata una nuova strategia di modellazione in termini di complessità computazionale e accuratezza. Quest’approccio si basa sulla rappresentazione in spazio di stato di tipo average che viene applicata a grandi impianti fotovoltaici con più stadi di conversione eseguendo alcune elaborazioni per ottenere un modello integrato finale. La nuova strategia di modellazione è stata testata in ambiente MatLab/Simulink utilizzando i dati provenienti da un impianto fotovoltaico da 300 MW situato in Brasile che rappresenta il caso studio di questo capitolo. In questo impianto, un sottocampo è dotato di inverter centralizzati mentre un altro è dotato di inverter di stringa. Il modello proposto, la cui accuratezza è compresa tra 2,2 e 2,7% rispetto all'energia misurata, supporta efficacemente l'analisi dei dati portando ad una valutazione coerente delle prestazioni per il sistema di conversione distribuito. I risultati finali evidenziano che gli inverter di stringa assicurano un guadagno di circa il 2% in termini di energia prodotta. Questa particolare configurazione di potenza consente anche il calcolo del mismatch a livello di stringa oltre che a livello di array all'ingresso lato DC di ciascun inverter di stringa. L'effetto mismatch è di fondamentale importanza nel funzionamento degli impianti fotovoltaici perché provoca perdite significative nella produzione di energia. Il mismatch è originato da diversi fattori come l'invecchiamento non uniforme dei moduli, l'ombreggiamento, l'accumulo di polvere, i guasti nei sistemi tracker e così via. La valutazione del livello di mismatch è un compito difficile perché è necessario tenere conto di molti fattori: configurazione di potenza, posizione geografica, condizioni meteorologiche, qualità delle misure fornite dai datalogger, ecc. In questo caso è stato implementato un approccio multi-criteriale con lo scopo di valutare la coerenza dei risultati. Infine, il Capitolo 8 mostra come misurare la potenza attiva come differenza tra il valore di picco della potenza istantanea e la potenza apparente. L'approccio tradizionale per calcolare la potenza attiva e reattiva nei sistemi di alimentazione in AC richiede la misurazione dello sfasamento tra tensione e corrente per la valutazione del fattore di potenza. Per fare ciò, gli analizzatori di potenza possono implementare diversi metodi. In linea di principio, è sempre necessario identificare punti specifici delle forme d'onda (ad esempio, utilizzando una tecnica di rilevamento zero-crossing) e ottenere il loro spostamento temporale. In modo simile, il valore della frequenza deve essere valutato per calcolare la frequenza angolare. Sfortunatamente, questo tipo di metodi comuni presentano alcuni problemi come la grande sensibilità al rumore. Inoltre, le imprecisioni nella valutazione del fattore di potenza hanno un grande impatto sulla stima finale della potenza elettrica. Al contrario, l'implementazione pratica della formulazione proposta negli analizzatori di potenza garantisce numerosi vantaggi senza ridurre l'accuratezza.
Analisi e gestione dei sistemi di accumulo dell’energia a batteria e degli impianti fotovoltaici / Vasta, Ester. - (2023 Jun 28).
Analisi e gestione dei sistemi di accumulo dell’energia a batteria e degli impianti fotovoltaici
VASTA, ESTER
2023-06-28
Abstract
Batteries represent the beating heart of global energy and economy transition. This concept is well illustrated in Chapter 1 with a brief overview of batteries between past, present and future. In Chapter 2 a battery testing system with software/hardware interface is presented and accurately described in each phase of its realization so as to be easily reproduced in a laboratory as if this chapter were a tutorial. Such a system allows to study the charging and discharging behavior of batteries according to customized cycles, by acquiring data such as voltage and current, as well as taking into account the effect of temperature. In Chapter 3 a real-time estimation algorithm including a proportional-integral (PI) based observer and an equivalent electric circuit model (ECM), is improved exploiting the state of charge (SOC) dependence on the open circuit voltage (OCV). Particularly, it is described how the SOC can be reliably deduced from relaxation voltage in a short time thanks to a prediction function. The integration of such function into the PI-based observer allows to get a good trade-off between accuracy and complexity. The same prediction function is also used for the purpose of estimating the state of health (SOH) obtaining a satisfactory accuracy while limiting the computational burden. In particular, two basic estimation methods, namely the coulomb counting method (CCM) and the ECM, are combined in order to get a real-time mixed algorithm; correctness of results is verified performing experimental comparisons with other common SOH estimation methods. With special reference to the automotive field, Chapter 4 presents a straightforward solution to estimate the SOC of battery packs used to supply low voltage (LV) electric drives integrated in hybrid and electric vehicles. The main idea is exploiting the electric drive to generate suitable DC bus current profiles to estimate the battery pack parameters, and thus the SOC, whenever the electric drive is not used as propulsion unit. In particular, the drive is exploited as controlled variable load to generate current steps and monitor the transient response of the battery pack through the measure of the DC bus voltage and motor phase currents. No additional external circuits are required, and a limited computational burden is required. Simulations and experimental tests confirm the feasibility of the proposed method. In virtue of the attention that the study of Li-ion batteries degradation deserves, Chapter 5 provides a review of the most important battery SOH estimation methods. After sorting all methods into three groups, two approaches were followed for each, highlighting theoretical aspects, strengths, weaknesses and model accuracies. In particular, electrochemical impedance spectroscopy (EIS) and incremental capacity analysis (ICA) were firstly considered among experimental methods. Secondly, equivalent ECMs and aging models (AMs) among model-based methods. Finally, neural networks (NNs) and support vector regression (SVR) among data-driven methods. With the same attention to the degradation process and, therefore, to the prognostic evaluation of a battery, Chapter 6 consists of an in-depth study in the literature on how to identify the RUL according to the so-called change point analysis. Contextually, with a view to an increasingly efficient integration of batteries into renewable plants, aspects related to photovoltaic (PV) plants have been studied. Hence, Chapter 7 focuses on the performance assessment of large PV plants using an integrated state-space average modeling approach. A comparison between central and string inverters is also presented as well as a study on mismatch losses. Distributed power converters represent a technical solution to improve the performance of large or utility-scale PV plants. Unfortunately, evaluation of the yield obtained in large PV fields by using distributed converters is a difficult task because of recurring partial unavailability, inaccuracy of power analyzers, operating constraints imposed by the power plant controller (PPC) and so on. To overcome such issues in real operating scenarios, a new modeling strategy has been introduced and validated in terms of computational complexity and accuracy. This approach is based on the state-space averaging technique which is applied to large PV plants with multiple conversion stages by performing some elaborations in order to get a final integrated model. The new modeling strategy has been tested in MatLab/Simulink environment using data coming from a 300 MW PV plant located in Brazil representing the case study of this work. In this plant, one subfield is equipped with central inverters while another is with string inverters. The proposed model, whose accuracy is in the range from 2.2 to 2.7% with respect to the measured energy, effectively supports data analysis leading to a consistent performance assessment for the distributed conversion system. Final results highlight that string inverters ensure a gain of about 2% in terms of produced energy. This particular power configuration allows also the calculation of mismatch at string level as well as at array level at the DC side input of each string inverter. The mismatch effect is of fundamental importance in the operation of PV power plants because it causes significant losses in energy production. The mismatch originates from several factors such as non-uniform modules aging, shading, dust accumulation, faults in tracker systems and so on. Evaluation of mismatch level is a difficult task because it is necessary to take into account many factors: power configuration, geographical position, weather conditions, quality of measurements provided by dataloggers, etc. In this case, a multi-criteria approach has been implemented with the purpose to evaluate the coherency of results. Finally, Chapter 8 shows how to measure active power as the difference between the peak value of instantaneous power and the apparent power. The traditional approach to calculate active and reactive power in AC power systems requires the measurement of phase shift between voltage and current for the evaluation of power factor. To do this, power analyzers can implement several methods. In principle, it is always necessary to identify specific points of waveforms (e.g., using a zero-crossing detection technique) and get their time shift. In a similar way, frequency value must be evaluated in order to calculate angular frequency. Unfortunately, this kind of common methods exhibits some issues as the large sensitivity to noise. Moreover, inaccuracies in the evaluation of power factor have a big impact on the final estimation of electric power. On the contrary, the practical implementation of the proposed formulation in power analyzers guarantees several benefits without reducing accuracy.File | Dimensione | Formato | |
---|---|---|---|
Ester Vasta - tesi phd.pdf
Open Access dal 29/06/2024
Tipologia:
Tesi di dottorato
Licenza:
NON PUBBLICO - Accesso privato/ristretto
Dimensione
13.77 MB
Formato
Adobe PDF
|
13.77 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.