Extracting Optimal Policies of Hydropower Multi-Reservoir Systems Utilizing Enhanced Differential Evolution Algorithm

Deriving the optimal policies of hydropower multi-reservoir systems is a nonlinear and high-dimensional problem which makes it difficult to achieve the global or near global optimal solution. In order to optimally solve the problem effectively, development of optimization methods with the purpose of optimizing reservoir operation is indispensable as well as inevitable. This paper introduces an enhanced differential evolution (EDE) algorithm to enhance the exploration and exploitation abilities of the original differential evolution (DE) algorithm. The EDE algorithm is first applied to minimize two benchmark functions (Ackley and Shifted Schwefel). In addition, a real world two-reservoir hydropower optimization problem and a large scale benchmark problem, namely ten-reservoir problem, were considered to indicate the effectiveness of the EDE. The performance of the EDE was compared with the original DE to solve the three optimization problems. The results demonstrate that the EDE would have a powerful global ability and faster convergence than the original DE to solve the two benchmark functions. In the 10-reservoir optimization problem, the EDE proved to be much more functional to reach optimal or near optimal solution and to be effective in terms of convergence rate, standard deviation, the best, average and worst values of objective function than the original DE. Also, In the case of two-reservoir system, the best values of the objective function obtained 93.86 and 101.09 for EDE and DE respectively. Based on the results, it can be stated that the most important reason to improve the performance of the EDE algorithm is the promotion of local and global search abilities of the DE algorithm using the number of novel operators. Also, the results of these three problems corroborated the superior performance, the high efficiency and robustness of the EDE to optimize complex and large scale multi-reservoir operation problems.     

Robust Diversity-based Sine-Cosine Algorithm for Optimizing Hydropower Multi-reservoir Systems

Hydropower energy generation depends on the available water resources. Therefore, planning and operation of the water resource systems are paramount tasks for energy management. Since reservoirs are one of the important components of water resources systems, extracting optimal operating policies for proper management of energy generated from these systems is an imperative step. Optimizing reservoir system operation (ORSO) is a non-linear, large-scale, and non-convex problem with a large number of constraints and decision variables. To solve ORSO problem effectively, a robust diversity-based, sine-cosine algorithm (RDB-SCA) is developed in the present study by introducing several strategies to balance the global exploration and local exploitation ability and to achieve accurate and reliable solutions. An efficient linear operation rule is coupled with the RDB-SCA to maximize the energy generation. The proposed method is then applied to a real-world, multi-reservoir system to extract optimal operational policies and, consequently, maximize the energy production. It is shown that the RDB-SCA is able to generate 24, 14, and 6% more energy than the original SCA, respectively for 2-, 3-, and 4-reservoir systems. The present findings are useful to suggest guidelines for efficient operation of hydropower multi-reservoir systems. This paper is supported by https://imanahmadianfar.com/codes.

     

Nanofluids thermal conductivity prediction applying a novel hybrid data-driven model validated using Monte Carlo-based sensitivity analysis

Accurate estimation of the thermal conductivity of nanofluids plays a key role in industrial heat transfer applications. Currently available experimental and empirical relationships can be used to estimate thermal conductivity. However, since the environmental conditions and properties of the nanofluids constituents are not considered these models cannot provide the expected accuracy and reliability for researchers. In this research, a robust hybrid artificial intelligence model was developed to accurately predict wide variety of relative thermal conductivity of nanofluids. In the new approach, the improved simulated annealing (ISA) was used to optimize the parameters of the least-squares support vector machine (LSSVM-ISA). The predictive model was developed using a data bank, consist of 1800 experimental data points for nanofluids from 32 references. The volume fraction, average size and thermal conductivity of nanoparticles, temperature and thermal conductivity of base fluid were selected as influent parameters and relative thermal conductivity was chosen as the output variable. In addition, the obtained results from the LSSVM-ISA were compared with the results of the radial basis function neural network (RBF-NN), K-nearest neighbors (KNN), and various existing experimental correlations models. The statistical analysis shows that the performance of the proposed hybrid predictor model for testing stage (R = 0.993, RMSE = 0.0207) is more reliable and efficient than those of the RBF-NN (R = 0.970, RMSE = 0.0416 W/m K), KNN (R = 0.931, RMSE = 0.068 W/m K) and all of the existing empirical correlations for estimating thermal conductivity of wide variety types of nanofluids. Finally, robustness and convergence analysis were conducted to evaluate the model reliability. A comprehensive sensitivity analysis using Monte Carlo simulation was carried out to identify the most significant variables of the developed models affecting the thermal conductivity predictions of nanofluids.

     

Maximizing the Firm Energy Yield Preserving Total Energy Generation Via an Optimal Reservoir Operation

As specified from the marketing standpoint, firm hydro-electrical energy must be available on an assured basis. To achieve the firm energy yield at maximum level, simultaneous optimization of operational variables is utilized. The variables include reservoir releases over historical stream-inflow records as well as the plant factor. Here, the plant factor denotes the percentage of day for operating a peak-time power plant at full capacity. To this end, a differential evaluation (DE) algorithm equipped with an indirect constraint handling is employed, in which an adaptive penalty system imposes the desirable reliability and preserves the total energy generation. To implement the model for a real example, a reservoir dam named Karun4, in southwest of Iran, is applied. The experimental results show the firm energy produced by the current and developed method are 49.2 and 127.5 (10⁹ watt hours), respectively. Also, the results reveal the maximum capability of generating the firm energy yield with target reliability whereas the total energy generation is preserving.     

Optimizing Multiple Linear Rules for Multi-Reservoir Hydropower Systems Using an Optimization Method with an Adaptation Strategy

Water Resources Management - Water resources crisis has a significant impact on hydropower energy production, which highlights the importance of water resources management. Reservoirs are effective...     

A novel boosting ensemble committee-based model for local scour depth around non-uniformly spaced pile groups

Engineering with Computers - Prediction of the scour depth around non-uniformly spaced pile groups (PGs) is one of the most complex problems is hydraulic engineering. Different types of empirical...