Swarm Intelligence in Power System Planning

Power system planning is one of the essential tasks in the power system operation management, which requires in-depth knowledge of the system under consideration. It can be regarded as a nonlinear, discontinuous, constrained multi-objective optimization problem. Although the traditional optimization tools can be used, the modern planning problem requires more advanced optimization tools. In this paper, a survey of state-of-the-art mathematical optimization methods that facilitates power system planning is provided, and the needs of introducing swarm intelligence approaches into power system planning are discussed.


Introduction
Power system planning plays a significant role in maintaining power system stability and reliability.It determines the right schedule of introducing additional generation facilities, transmission lines, substations, transformers, reactive compensations, and control devices; also it covers the direction of replacement needs with respect to aging existing power system devices, where all of these are directed at increased stability and reliability through market-driven augmentations [1].
Generally speaking, in terms of objectives, the planning issues may be categorized as generation planning and transmission planning; while in term of periods, the planning problems can be classified into short-term planning, mid-term planning, and long-term planning [2].Generation planning is intended to determine the optimal timing, locations, generation equipments and associated capacities required to satisfy various system constraints and operation conditions, which can maximize profit and minimize risk [1].Similarly, transmission planning aims to select the best time, siting, and transmission facilities to meet the rising customer demand, which minimizes investment and maintains stability [2].From planning period point of view, the major purpose of short-term planning is to devise operation plans for power plants or single generating unit so as to ensure the balance of supply and demand [3].Med-term planning provides the guidance for making market decisions and system operations.Long-term planning ensures adequate generation capacities and delivery capabilities will be available to meet the expected future demand increases.
In a word, planning is one of the most important re-search areas in power system analysis, which needs careful and extensive studies and it should be carried out in a timely and effective manner.

New Challenges
Nowadays, power industries worldwide have been undergoing profound changes with system deregulations and reconstructions.In particular, the traditional, vertically monopolistic structures have been reformed into competitive markets in pursuit of increased efficiency in electricity productions and utilizations.Along with the introduction of competitive and deregulated electricity markets, many power system problems have become difficult to be analyzed with traditional methods, especially when power system planning issues are involved.Since the change in the competition environment, the perspectives of power system planning are also change correspondingly.In the traditional vertically-integrated structure, the whole power system is operated by single system and service provider, who owns all the generation and transmission assets.Therefore, when conducting power system planning, the generation and transmission planning can be carried out simultaneously.However, after deregulations, the conventional monopolistic structure has been separated into three independent parts: generation, transmission, and distribution.This separation results in a situation that transmission companies have no direct role in deciding the patterns of power generation and distribution.Furthermore, some investment and operation information about generation and distribution companies becomes business confidential and cannot be obtained by the planner [4].Moreover, the modern power system planning process requires a wide array of input information, such as weather forecasts, load forecasts, market forecasts, expected water supply and fuel price variations.In addition, system constraints, single unit constraints, and various chance constraints, along with other environmental and physical influence, are taken into account.Besides technical information, there are social and governmental organizations to be consulted in the process of planning so as to ensure that every decision is well rounded and completed.All these have made existing problems even more complex.As a consequence, more advanced and effective techniques need to be introduced into planning issues.

Swarm Intelligence
Swarm intelligence is an artificial intelligence technique involving the study of collective behavior in decentralized system, which is made up by populations of simple individuals interacting locally with each other and with external environment [5][6][7][8].Several examples of these systems can be found in the nature, for example, colonies of ants, flocks of birds, schools of fish, groups of bees, packs of wolves, and so on.An interesting phenomenon of swarms is that collective swarm behavior can emerge on a global scale even when all individuals have only a restricted view of the system and interactions between individuals and their environment occur only on a local scale [9].Owning to these outstanding characteristics, the principles of swarm behavior have been studied extensively and been widely applied into many fields.Computational swarm intelligence is the algorithmic models that imitate the principles of large groups of simple swarm individuals working together to achieve a goal through self-learning, self-adjusting, and mutual cooperation manners.These algorithms have shown to be able to adapt well in changing environments, and are immensely flexible and robust [8,10].Two of the computational swarm intelligence techniques are ant colony optimization (ACO) [11] and particle swarm optimization (PSO) [6].In the next section, these two swarm intelligence algorithms will be discussed in detail.

Ant Colony Optimization (ACO)
ACO is a metaheuristic inspired by the foraging behavior of ants [12][13][14].In order to find the shortest path from the nest to food source, ant colonies exploit a positive feedback mechanism: they use a form of indirect communication called stigmergy, which is based on the laying and detection of pheromone trails [15,16].These ants deposit pheromone on the ground in order to mark some favorable path that should be followed by other members of the colony [13].ACO takes inspirations from the collec-tive behavior of ants and exploits a similar mechanism for solving optimization problems.In ACO, firstly colonies of artificial ants with given size are generated, and each ant denotes a potential solution, whose performance is measured based on a quality function.Many different paths can be constructed by ants walking on the graph, and these paths encode the target problem.The cost of the generated paths is used to modify the pheromone left by ants, and therefore to bias the generation of further paths towards promising regions of the search space [17,18].Among these feasible paths, ACO attempts to find the one with minimum cost.

Particle Swarm Optimization (PSO)
PSO is a heuristic algorithm developed in [19][20][21].The algorithm is inspired by the social behavior of a bird flock.It has been found to be successful in a wide variety of optimization tasks.In PSO, each solution is a bird in the flock and is referred to as a particle, which denotes a candidate solution to the optimization problem.The birds in the population evolve their social behavior and accordingly their movement towards a destination.In a PSO system, each particle flies through the multidimensional search space, adjusts its position in search space according to its own experience and that of neighbor particles.A particle makes use of the global best position which the current particle has visited so far, as well as the population best position which the entire population has found so far and the process repeats until the swarm reaches a desired destination.The effect is that particles fly toward a minimum, while still searching a wide area around the best solution.The performance of each particle is measured by using a predefined fitness function, which encapsulates the characteristics of the optimization problem.Two parameters inertia weight and constriction factor are used to control over the previous velocity of the particles [22].In short, PSO is characterized as a simple heuristic of well balanced mechanism with flexibility to enhance and adapt to both global and local exploration abilities, which gains lots of attention in power system applications [23].

Power System Planning
In this section, a survey of state-of-the-art mathematical optimization approaches that facilitates power system planning is provided and the merits and needs of introducing swarm intelligence methods into power system planning are discussed.

Generation Expansion Planning
Generation expansion planning is intended to determine the locations, capacities, and expected operations of gen-eration plants required to satisfy various requirements and constraints imposed by future expectations and forecasting conditions, which is to be done in a manner that maximizes profits and minimizes risks [1], [24].Mathematically, the consequent typical generation planning challenge can be expressed as a large-scale, non-linear optimization problem with the objectives of maximizing profits and minimizing risks subject to a set of complicated constraints.
To solve the complicated issues of generation expansion planning, different mathematical methods have been suggested and reported.The initial work started in [25], where a linear programming method was applied to necessitate the linear approximation of objective function and various constraints.Then linear programming model was further enhanced to address the increasingly complex planning issues with multi-objectives [26].An extensive study of the applications of linear programming methods in power system was given in [27].Linear programming methods are fast and reliable, but the main drawback is that they are associated with the piecewise linear cost approximation.Another great alternative for generation planning is nonlinear programming methods [28].However, nonlinear programming methods have a problem of algorithm convergence and complexity.Along with the ever expanding large-scale interconnection of modern power network, both linear programming and nonlinear programming techniques were not adequate for most applications until dynamic programming appeared, which overcame some limitations and received wide acceptance [29].In general dynamic programming based methods have the advantage over the other techniques, in that, nonlinearity in project capital costs and engineering constraints, sophisticated techniques of production costing such as probabilistic simulation, and an adequate modelling of the reliability of the system during its future stages, can all be more adequately accounted for [30,31].However, the curse of dimensionality problem of generation expansion planning afflicts the method of dynamic programming particularly severely [32,33].In many cases, the mathematical equations involved have to be simplified or decomposed in order to obtain possible solutions because of the limited capability of existing mathematical approaches [34,35].
Recently, the advent of global optimization techniques provides another tool for solving power system generation expansion planning problems and satisfactory performance has been reported in a number of references.Typical modern heuristic methods include evolutionary programming (EP) [36], simulated annealing (SA) [37], genetic algorithm (GA) [33,[38][39][40][41][42], immune algorithm (IA) [43], PSO [44,45], and composite method [46].Although the heuristic methods do not always guarantee discovering globally optimal solutions in finite time, they often provide a fast and reasonable solution.Generally speaking, each method has its own merits and drawbacks.Many attempts try to merge some of the individual implementations together into a new algorithm, so that it can overcome individual disadvantages and benefit from each others' advantages.Extensive reviews and comparisons of these techniques in power system generation expansion planning are given in [47][48][49].Based on the experience, when compared with other methods, the PSO is computationally inexpensive in terms of memory and speed.The most attractive features of PSO could be summarized as: simple concept, easy implementation, fast computation, and robust search ability [50].

Transmission Expansion Planning
Transmission lines are key components in a power system, especially where system stability and reliability analysis is concerned.In a deregulated electricity market, transmission network service providers make possible the required competitive environment for the market participants.Therefore, as the market grows, transmission planning should be carried out in a timely and effective manner.In a competitive market, such planning is mainly driven by market needs, with the proviso that certain constraints, such as reliability, security, economic considerations, and regulatory rules, are satisfied [51,52].However, restructuring and deregulation of the power industry have given rise to more and more system uncertainties and have changed the objectives of transmission planning.As a consequence, the process of transmission planning requires an evaluation of the annual load shape and of the cost effectiveness and financial performance of programs and plants, together with an analysis of product attributes, profitability niches, delivery preferences, and investment risks [53].The intention of such planning is to minimize revenue requirements, meet customer needs, as well as maximize network profits [53,54].Following these changes, new approaches and action criteria are demanded.These should not only consider the traditional constraints, but they should also promote fair competition in the electricity market as well as ensuring certain levels of reliability.
Transmission expansion planning is a complex multiperiod, multi-objective optimization problem.In previous research, as far as optimization approaches used for transmission expansion planning, the linear programming was most frequently applied [55][56][57][58].Since the transmission expansion planning problem is essentially of a discrete nature, it can be defined via a mixed-integer programming formulation.The applications of mixed-integer programming methods can be found in [59][60][61][62][63][64].Dynamic programming method [65] can also been applied to this problem.Similar as generation expansion planning, the artificial intelligence based methods also pro-vide perfect options for transmission expansion planning.Technical references can be classified according to the methodologies used to solve the problem, which includes expert systems [66] and fuzzy theory [67].Recently, different heuristic methods have been proved to be effective with promising performance, such as SA [68,69], tabu search (TS) [70], GA [71][72][73], differential evolution (DE) differential evolution (DE) [4,74], and PSO [75,76].

Planning with Distributed Generation and Renewable Energy resources
Today, more and more renewable energy resources are being built up and connected to the power grids at transmission as well as distribution levels.Large scale wind farms are connected to transmission networks and are so far the largest renewable generation source except hydro power generations.Planning of wind farms requires significant amount of studies including the conventional generation connection studies as well as very expensive wind farm planning itself.Normally to connect a wind farm, significant historical wind resource data are required to evaluate the suitability of the wind farm site.
Once the site is selected, further optimization planning is required to design the exact scheme of the wind farm so as to maximize the energy output of wind power.This is normally a multi-objective, constrained, highly nonlinear problem.Computational intelligence such as PSO and DE can be used to solve the optimization problem in designing the wind farms, [77,78].At distribution level, the increasing penetrations of distributed renewable generation can potentially cause problems with some feeders.The common problems are protection system design and reactive power support issues with such heavily DG connected feeders.Planning of DGs and power quality control devices such as STATCOM is another complex optimization problem, and evolutionary computation can be used in the planning as well, [79,80].

Conclusions
Owning to these outstanding characteristics, the swarm intelligence techniques have been studied extensively and have been widely applied in many fields.In general, the swarm intelligence techniques can be used to solve the nonlinear, discontinuous, constrained, optimization problem.In this paper, a comprehensive survey of stateof-the-art mathematical optimization methods that facilitates power system planning is provided; the importance of introducing swarm intelligence methods into power system planning is discussed.