Meta-heuristic approaches for reservoir optimisation operation and investigation of climate change impact at Klang gate dam

Lai, Vivien Mei Yen (2023) Meta-heuristic approaches for reservoir optimisation operation and investigation of climate change impact at Klang gate dam. PhD thesis, UTAR.

Preview

PDF Download (3792Kb) | Preview

Abstract

Due to the extraordinarily rapid growth in population and development, the demand for energy and water has increased to critical demanding levels, globally. Thus, the reservoir, the essential infrastructure for water storage during extreme events such as intense rainfall or drought periods, is indeed crucial to ensure availability of potable water. With the right reservoir functioning, society can achieve hydrological resilience, water sustainability, relief from and control of urban flooding, and sustainable energy. Over the years, dam operators, stakeholders, and scholars have shown their commitment to sustaining reservoir operations and doing their best to gain knowledge on how to manage reservoir operations, in order to maximise benefits while minimising the drawdowns in water supplies or overcoming poor performance. In 1998, a severe water crisis in the Klang Valley, Malaysia, due the El Niño phenomenon, had the water level at the Klang Gate Dam (KGD) dropped dramatically. This intricately added to the reservoir and dam issues in Malaysia, particularly the frequent intense rainfall within short periods of time, which made it difficult for the reservoir and dam operator to monitor and maintain the storage level of the reservoir and discharge water downstream to prevent overflow and flooding. Consequently, seeking managing of reservoir optimisation operations had always been at the forefront and to improve managing, algorithms have had been presented over the past few decades, beginning with conventional algorithms, followed by heuristic algorithms, and finally, the meta-heuristic algorithms (MHAs). However, due to the drawbacks of the conventional algorithm as well as the heuristic algorithms handling complicated and multi-objective reservoir optimisation, the advantages of the strategy of simultaneous exploration and exploitation led to the decision to utilise meta-heuristic algorithms in this study. The original idea of this study was to investigate the climate impact onto the KGD current and future operations. By preserving the equilibrium between the proposed MHAs and reservoir risk analysis indices, the stakeholder can select or control the optimal KGD operation by referring to the summary of findings for the observed period assessments. The Whale Optimisation Algorithm (WOA), Harris Hawks Optimisation (HHO) Algorithm, Lévy Flight WOA (LFWOA) and the Opposition-Based Learning of HHO (OBL-HHO) were proposed to simulate the initial model’s response and optimise the Klang Gate Dam (KGD) release operation with observed inflow, water level (storage), release, and evaporation rate (loss). There were two observed periods of timeline: (a) for year 2001-2019 and (b) for year 1987-2008 (compared with past studies). The results obtained from the proposed meta-heuristic algorithms of this study were then evaluated for reservoir risk analysis, for the observed period assessment and the climate assessment. In addition, extreme climate change occurrences have impacted the future reservoir operation, and this is something that previous KGD studies have yet to investigate. Thus, the continuing investigation of the optimisation of the future KGD operation under various climatic scenarios by leveraging on the proposed MHAs, was conducted for the climate assessment for year 2020-2099. The comparison between the reservoir simulation (ANN) and reservoir simulation-optimisation (MHAs) were carried out in terms of examining the reservoir climate assessments, as well as the monthly storage capacity. In addition, a few scenarios of the future water demand were developed and estimated based on a close proximity of real condition: (i) Temperature Scenarios and (ii) Forecasted Population Growth. Scenario 1 was developed for the base period and the water demand was identical to the observed period assessment. Scenario 2: Maximum Temperature, Scenario 3: Mean Temperature and Scenario 4: Minimum Temperature; were developed for the water demand conditions. The results obtained for year 1987 - 2008 assessments showed the proposed MHAs as an optimistic conclusion for the dam operator to consider based on the trade-off between reliability and resilience or other reservoir risk indices. The proposed MHAs were next compared to past studies and it was shown that the GA binary had the lowest reliability and the Artificial Bee Colony (ABC) in the past studies, had the most vulnerability and sensitivity in data interpretation, especially with limited observed datasets. The LFWOA showed the highest level of periodic reliability, with 69.70%, while the HHO exhibited a slightly lower percentage of 63.26%. The ABC and PSO algorithms exhibited lower periodic reliability percentages of 61.36% and 59.47%, respectively. The OBL-HHO and WOA algorithms showed periodic reliability with percentages of 56.44% and 56.06%, correspondingly. The GA�RC algorithm showed a periodic reliability percentage of 55.65%, whereas the GA algorithm exhibited the lowest periodic reliability percentage of 23.5%. For the year 2001-2019 assessments, the algorithms varied in the ranking of reservoir risk assessments for all the three inflow magnitude types (low, medium and high). For the high inflow category, the LFWOA exhibited the highest periodic reliability in terms of meeting exact demand with a value of 15.35% whilst the WOA had achieved a reliability of 14.47%. At the same time, the HHO and OBL-HHO algorithms resulted in lower levels of periodic reliability, with values of 13.16% and 9.21% respectively. The HHO model was still inspired to be the model to perform the reservoir optimisation operation even though it had obtained the highest sequence for the vulnerability in the high inflow category. Within the medium inflow category in terms of meeting precise demand, the LFWOA exhibited the highest level of periodic reliability with a percentage of 42.54%. This was closely followed by the WOA with 39.91%, the HHO with 38.60%, and then, the OBL-HHO with 20.54%. The resilience metric associated with the medium inflow category exhibited performances that align with the periodic reliability in a similar sequence. Regarding the medium inflow category for the vulnerability metric, it has been observed that the algorithms of OBL-HHO, HHO, WOA, and LFWOA exhibited significant robustness. For the 2020-2099 climate assessments, the sequence of the respective algorithms in terms of individual reservoir risk analysis assessment in accordance with RCP 2.6 of Scenario 2, Scenario 3, and the forecasted population growth of future water demand showed that the WOA was extremely vulnerable and sensitive. The monthly storage capacity fails in 2077, substantially earlier than the other three algorithms in Scenario 2. The LFWOA was used in this study to improve the efficacy of the algorithms by delivering a more accurate monthly storage capacity and reservoir risk assessment for Scenario 2 of RCP 2.6. In Scenario 4, the lowest ranking of vulnerability showed that the LFWOA was the most vulnerable and sensitive whereby the month storage capacity failed in 2077, but it was able to recover, as LFWOA gained the second-highest resilience index ranking. In RCP 4.5, Scenarios 3, Scenario 4, and forecasted population growth had no monthly storage failures. However, the LFWOA had the lowest vulnerability sequence for Scenario 2, which occurred in 2062 and after. On the other hand, the monthly storage failure occurred too rapidly in the near future for Scenario 3, Scenario 4, and future population growth since RCP 8.5 is called "High emissions". To retain the balance of trade-offs at the KGD operation, the best reservoir operation decision must be aware of the storage failure event when an extreme event occurs. From the above-mentioned, the major findings of this study were on the investigations of the climate assessments during KGD operations under the scope of the different climate change scenarios, using an ensemble of GCMs with the purpose of equally distributing the uncertainty accuracy of the downscaled as compared to previous studies which were then using the single GCM approach. Few recommendations of future work direction such as by hybridisation or utilising other algorithms (with a similar strategy of exploitation and exploration) to further examine the critical events obtained in this study, were suggested further improvements. Aside from that, it is also recommended that with the implementing of the most recent GCMs (ensemble) of CMIP 6 to analyse and compare with the current studies conducted at KGD using ensembles GCMs of CMIP 5, should pave the way for more intense forward research.

Actions (login required)