Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. International Journal of Engineering & Technology, 7 (3.13) (2018) 44-50 International Journal of Engineering & Technology Website: www.sciencepubco.com/index.php/IJET Research paper New Strategy Based on Combined Use of Genetic Algorithm and Gradient to Solve the UC Problem: Theoretical Investigation and Comparative Study Sahbi. Marrouchi 1 *, Moez Ben Hessine 1 , Souad Chebbi 1 1 Laboratory Technologies of Information and Communication and Electrical Engineering (LaTICE), National Superior School of Engineers of Tunis (ENSIT), University of Tunis, 5 Street Taha Hussein Montfleurie, 1008 Tunis, Tunisia. *Corresponding author E-mail: sahbimarrouchi@yahoo.fr Abstract This paper presents a comparative study between a new strategy based on hybrid Gradient-Genetic Algorithm method and metaheuristic methods for solving Unit Commitment problem. Strategies have been applied on the IEEE electrical network 14 bus test system for a variable load profile during a discretized margin of time (24-hour time requirement). The right choice of the initial population and the best knowledge of the technical constraints specific to each generator (power balance constraints, Spinning reserve constraints, minimum up time, minimum down time ) suggests the possibility of obtaining improvements in the time execution. The adopted strategy has pre- sented high performance both for minimizing the production cost and for the rapidity of convergence to optimal solutions and is promis- ing compared to Genetic algorithm. Keywords: Unit commitment ; Optimization ; Scheduling ; Genetic Algorithm ; Gradient method 1. Introduction In Unit Commitment Problem, each unit has its own production limits and minimum time to reboot and shutdown. It is, therefore, a mixed complex optimization problem [1,2], combinatorial and nonlinear [3,4,5,6]. It is difficult to determine a planning econom- ic operation for this reason; the researchers observed that the sto- chastic models are more efficient than deterministic models in uncertainty. However, stochastic search algorithms are able to overcome the shortcomings of conventional optimization tech- niques [7, 8, 9, 10]. These methods can handle complex nonlinear constraints and provide high quality solutions. However, PG Lowery et al [11] established a strategy based on dynamic programming (DP) to solve the UC problem. The strate- gy was effective in reducing the production cost and computation time but it presents a combinatorial complexity while the number of generators increases which makes the time to slower calcula- tion. For this purpose, Pang et al. [12] proposed two algorithms: combinatorial sequential dynamic programming (MHPD) and combinatorial truncation dynamic programming (DPTC) to solve the combinatorial complexity and calibrate the resolution of the Unit Commitment problem by dynamic programming. The results show that MHPD offers better solutions compared to DP and DPTC and is effective at reducing the total cost of production and in execution time. However, Merlin et al. [13] proposed a new approach to develop a flexible algorithm for simultaneous man- agement of pumping units and probabilistic determination of the reserve to guarantee through the Lagrangian relaxation method. The method was validated with the power company de France (EDF). The method was promising except that the running time of the program increases only linearly with the size of the system and that it is applied only for electrical networks including power plants. As for Ongsakul et al. [14], they proposed an adaptive improvement on Lagrangian relaxation which aims to eliminate the drawbacks mentioned by LR Aoki and merlin. This improve- ment is to adapt a heuristic search to adjust the program if any of the consumption of forecast errors and employ adaptive Lagrangi- an relaxation to find the best possible planning. The results show that this improvement provides a final solution at a lower cost compared to conventional LR and DP, but the degree of optimality (percentage of optimal solution) decreases when the units of the numbers increase and classical LR remains fast computing time that this technique. Furthermore, Kavatza et al. [15] suggested a new simulated annealing approach combined with the method of dynamic load to ensure the scheduling of production units, while load balancing procedure was applied by integrating the con- straints of the ramp rate in solving the Unit Commitment problem in order to determine the output power of each generator. Com- pared to methods LR, GA and SA, the simulated annealing meth- od is effective in minimizing the cost of production but it has quite a long time. Sheble et al. [16] presented a new approach based on genetic algorithm (GA) for the planning of production units. The results show that this method provides good quality solutions at lower total cost of production but requires a lot of computation because it manipulates several solutions simultaneously. The suc- cess of the development requires several tries so it takes a huge time. However, Rudolf et al. [17] proposed an approach to two levels of programming to solve the problem of engagement of the units. The first level uses the algorithm that was developed by the states to decide Sheble units on / off. The second uses a formula- tion of the nonlinear programming by the LR method to perform load balancing within the constraints of the system. This technique has been applied to a hydrothermal system at real scale. The simu- lation results show that the implementation of this strategy is easy