# Fuzzy

• What is fuzzy logic (FL)?

class of AI, Based on the nature of fuzzy human thinking, deals with problems that have fuzziness or vagueness, defined as multi-valued logic (0 to I), deals with imprecise information

• What's a fuzzy set?

fuzzy set (fuzzy value) theory based on FL, a particular object has a degree of membership in a given set that may be anywhere in the range of O (completely not in the set) to 1 (completely in the set)

• What's a Fuzzy problem ?

an input/output, static, nonlinear mapping problem through a "black box," as shown, all the input information is defined in the input space, it is processed in the black box, and the solution appears in the output space.

• mapping can be static or dynamic, and the mapping characteristics are determined by the black box's characteristics.
• The black box cannot only be a fuzzy system, but also an ES, neural network, general mathematical system, such as differential equations, algebraic equations, etc., or anything else.
• What's Fuzzy Control (FC)?

Fuzzy control is basically an adaptive and nonlinear control, which gives robust performance for a linear or nonlinear plant with parameter variation

• What are the available control implementation methods for fuzzy control ?
1. Using real time computation (C program, FL Toolbox or DSP ASIC chips)
1. Using a look up table where all computation are done a head in hierarchical tables (coarse, medium and fine)

• What is the fuzzy logic control design methodology?
1. Analyze if the problem requires FC ?
1. if plant model available simulate it and study its characteristics.
1. identify functional elements where to apply FL.
1. identify the inputs and outputs of each fuzzy sys.
1. define universe of discourse and convert to pu values.
1. formulate fuzzy sets, choose MF shapes
• sensitive : need more fuzzy sets
• more precision near S.S : crowding of MFs near the origin
1. formulate rule table.
1. if there is a plant model , test and iterate fuzzy sets and rules

if not available, design a conservative fuzzy control and test on plant operation.

1. implement in real time, and iterate to improve performance

• Mention some of fuzzy logic control applications in power electronic systems.

Applications include 1. speed control of dc and ac drives

2. feedback control of converter 3. off-line P-I and P-I-D tuning. 4. nonlinearity compensation 5. on-line and off-line diagnostics 6. modeling 7. parameter estimation. 8. performance optimization of drive systems based on on-line search 9. estimation for distorted waves

• Discuss efficiency improvement under FLC1 and FLC2

• The gain due to FLC- I by tuning the generator speed to optimal value can be very high at low wind velocity although P0 may be small
• It decreases to zero near 0.7(pu) wind velocity
• and then increases again
• the efficiency gain by FLC-2 decreases because of higher generator loading.

### Paper Microgrid PV MPPT (hill climbing)

• define Microgrids.

Microgrids are defined as systems that have at least one distributed energy source and associated loads and can form intentional islands in the electrical distribution systems

• What are PV systems main problems?
1. High fabrication cost
1. Low conversion efficiency especially under variable weather conditions
1. The nonlinearity between the PV array output power and current
• what's Ripple correlation control (RCC)

is an optimization technique that takes advantage of the converter signal ripple to track the MPP

• what are the drawback of old MPPT tracking methods (fuzzy or not)

like current-power change method is that the operating point is mover away from the maximum point when irradiance changes

1. Simplicity
1. Easy implementation
1. low cost

1. Slow converge to optimum point
1. at SS power oscillates around MPP, causing losses
1. during cloudy when irradiance varies, operating point may move away from MPP.
• How to ensure PV output doesn't diverge from mpp with varying weather (3rd drawback solution)

To ensure that the PV output power does not diverge from the optimum point during varying weather conditions, ΔP passes through a gain controller to reverse its direction.

• What advantages does the FLC hill-climbing MPP has?

The FLC computes variable step sizes to increment or decrement the duty cycle, therefore the tracking time is short and the system performance during steady-state conditions is much better than with conventional hill-climbing algorithm. Moreover, divergence problem no longer exist since the controller input, change of power (dP), reverses its direction in response to atmosphere condition variation

# Artificial Neural Networks

• What are the general applications of Neural Networks? What are neural networks used for ?
• Classification
• Regression
• Combinatorial optimization
• What's the objectives of neural network ?

compute the weights to obtain the desired decision for the given input patterns

• What are the NN learning or training techniques ?
• steepest descent (error backpropagation)
• Random search
• Genetic algorithm
• Particle swarms

# Genetic Algorithm

• What's Genetic Algorithm ?
• a search technique used in computing to find true or approximate solutions to optimization and search problem.
• global search heuristics
• class of evolutionary algorithms that use techniques inspired by evolutionary biology such as inheritance, mutation, selection, and crossover (recombination)
• probabilistic intelligent search algorithm, which searches a population of points in parallel.
• In what does GA differ from other optimization techniques ?

• It searches a population of points in parallel • It uses probabilistic rules rather than deterministic ones • It can process an encoding set of parameters

• Give an overview of the evolution process in GA.
1. starting from a population of randomly generated individuals
1. then evolution happens in generations
1. In each generation, the fitness of every individual in the population is evaluated
1. multiple individuals are selected from the current population (based on their fitness)
1. the selected individual are modified to form a new population.
1. new population is used in the next iteration of the algorithm
1. To terminate algorithm either 1) a max. number of generations 2) satisfactory fitness level acquired.

• Start with a large “population” of randomly generated “attempted solutions” to a problem • Repeatedly do the following:  Evaluate each of the attempted solutions  (probabilistically) keep a subset of the best solutions  Use these solutions to generate a new population • Quit when you have a satisfactory solution (or you run out of time)

• Define, individual, Population and Fitness.

Population : Group of all individuals Target (search space)

Individual : Any possible solution also called chromosome

Fitness : function that we optimize (each individual has a fitness)

• What's penalty cost ?

To initiate the population, a number of individuals are randomly formulated in the range of variables Each individual in the randomly-created population is tested for violating the constraints If the solution is infeasible, a suitable additional penalty cost is assigned to the individual The performance of each individual is evaluated by calculating the value of objective function and adding the corresponding penalty term if exists

• What's individuals fitness value ?
• The individuals are ranked depending on their corresponding costs and a suitable fitness value is assigned to each one
• The fitness values could be calculated depending on the position of the individuals within the population rather than their distinct performance
• Fitness values between maximum and minimum limits are calculated with fixed incremental steps and assigned to the ranked individuals
• What's the Constraints representation problem in GA ? and what are the approaches to solve this problem ?
• To employ GA with a tightly-constrained problem, it is possible to generate only feasible solutions by avoiding individuals which violate the given constraints
• but, the infeasible solutions mostly cover the search space at the initial generation, so the avoidance will cause missing the area of global minimum

Approaches,

1. Move the infeasible individuals to the nearest feasible area,
• but this would be too complex and a very time consuming
1. Penalty function convert to an unconstrained problem by augmenting additional cost terms with the main objective function
• nonlinear costs for solutions that violate the constraints depending on their relative locations with respect to the feasibility boundaries
• ensure adequate choice of the penalty functions and their parameters for rapid rejection
• differentiate in performance between the infeasible individuals themselves, which will help in the evolution process
• higher cost value has to be assigned to any infeasible solution than the feasible members, for fast rejection

• How to do recombination in non-binary chromosomes ?

### Crossover

Crossover is a genetic process to exchange information between members of the population.

For the real-valued encoding, the max-min arithmetical crossover operator can be used as follows

### Mutation

It is the second process in the recombination, which is used to escape from possible local minima. The mutation in the real-coded representations is accomplished by disturbing the gene values with low probability.

• What's Elitism?
• to ensure that the best solutions are not lost in moving from one generation to the next.
• some of the fittest members of each generation are saved and copied into the next generation.
• the average fitness of the new generation is improved.
• Give a numerical example of a GA problem parameters

### Number of individuals per subpopulation (20)

for migration purposes.

### Total population size (200)

the number of individuals

### Generation gap (0.8)

number of individuals recombination are made on

### Insertion rate (0.9)

that 0.9 of the new generation will replace the worst same number of individuals in the old generation

### Probability of mutation (0.01)

how much of the genes will be mutated

### Maximum generations (1000)

the number of generation at which the algorithm terminates