BOATS: Bayesian Optimization for Active Control of ThermoacousticS
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913411193896960 |
|---|---|
| author | Dharmaputra, Bayu Reckinger, Pit Schuermans, Bruno Noiray, Nicolas |
| author_facet | Dharmaputra, Bayu Reckinger, Pit Schuermans, Bruno Noiray, Nicolas |
| contents | This investigation presents novel adaptive control algorithms specifically designed to address and mitigate thermoacoustic instabilities. Two control strategies are available to alleviate this issue: active and passive. Active control strategies have a wider flexibility than passive control strategies because they can adapt to the operating conditions of the gas turbine. However, optimizing the control parameters remains a challenge, especially if additional constraints have to be fulfilled, such as e.g. pollutant emission levels. To address this issue, we propose three adaptive control strategies based on Bayesian optimization. The first and foundational algorithm is the safeOpt algorithm, and the two adaptations that have been made are stageOpt and shrinkAlgo. The Gaussian Process Regressor (GPR) is employed to approximate both the objective and constraint functions, with continuous updates occurring during iterations. The algorithms also enable the transfer of knowledge obtained from one operating point to another, thereby reducing the number of iterations needed to reach the optimal point. We demonstrate the effectiveness of the algorithms both numerically and through two distinct experimental validations. In the numerical demonstration, we employ a low-order thermoacoustic network model to simulate a single-stage combustor setup equipped with loudspeaker actuation and a gain-delay ($n-τ$) controller for active stabilization. The first experimental demonstration has the same structure as the numerical case. For the second experimental validation, we apply the framework to a sequential combustor configuration utilizing nanosecond repetitively pulsed discharges (NRPD) as the control actuator. This demonstrates the framework's adaptability to various control actuation methods in turbulent combustors where control parameter optimization is required. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_07865 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | BOATS: Bayesian Optimization for Active Control of ThermoacousticS Dharmaputra, Bayu Reckinger, Pit Schuermans, Bruno Noiray, Nicolas Optimization and Control Fluid Dynamics This investigation presents novel adaptive control algorithms specifically designed to address and mitigate thermoacoustic instabilities. Two control strategies are available to alleviate this issue: active and passive. Active control strategies have a wider flexibility than passive control strategies because they can adapt to the operating conditions of the gas turbine. However, optimizing the control parameters remains a challenge, especially if additional constraints have to be fulfilled, such as e.g. pollutant emission levels. To address this issue, we propose three adaptive control strategies based on Bayesian optimization. The first and foundational algorithm is the safeOpt algorithm, and the two adaptations that have been made are stageOpt and shrinkAlgo. The Gaussian Process Regressor (GPR) is employed to approximate both the objective and constraint functions, with continuous updates occurring during iterations. The algorithms also enable the transfer of knowledge obtained from one operating point to another, thereby reducing the number of iterations needed to reach the optimal point. We demonstrate the effectiveness of the algorithms both numerically and through two distinct experimental validations. In the numerical demonstration, we employ a low-order thermoacoustic network model to simulate a single-stage combustor setup equipped with loudspeaker actuation and a gain-delay ($n-τ$) controller for active stabilization. The first experimental demonstration has the same structure as the numerical case. For the second experimental validation, we apply the framework to a sequential combustor configuration utilizing nanosecond repetitively pulsed discharges (NRPD) as the control actuator. This demonstrates the framework's adaptability to various control actuation methods in turbulent combustors where control parameter optimization is required. |
| title | BOATS: Bayesian Optimization for Active Control of ThermoacousticS |
| topic | Optimization and Control Fluid Dynamics |
| url | https://arxiv.org/abs/2401.07865 |