Netload-constrained Unit Commitment Considering Increasing Renewable Energy Penetration Levels: Impact of Generation Schedules and Operational Cost

Received Jan 8, 2018 Revised Feb 16, 2018 Accepted Mar 25, 2018 In the context of low carbon power systems, the penetration levels of renewable energy sources (RES) are expected to increase dramatically. In this regard, this paper investigates the maximum RES penetration level constrained by net load while considering an inflexible unit commitment (UC) model. To solve the UC problem, an enhanced priority list (EPL) based method is developed. In the proposed method, the plants were activated sequentially based on the operational price. The system constraint violations were repeatedly corrected until all system constraints (such as net load and spinning reserves) were satisfied. The proposed EPL method was efficient to achieve a near optimal solution under high shares of RES. Furthermore, the research work investigates three different scenarios representing penetration levels of 10% solar-only, 14.5% wind-only and 27.5% mixture of both solar and wind. The impact of each penetration level on the system scheduling and operational cost were analyzed in detail. The analysis presented shows that a potential operational cost savings of 21.6 $/MW, 20 $/MW and 11.1 $/MW is feasible under each of the represented scenarios, respectively. Keyword:


INTRODUCTION
The climate change mitigation and security of energy supply has stimulated national energy policies towards higher integration of renewable energy sources (RES) [1]. Unlike conventional power sources, variable RES has a maximum available generation limit that changes with time (variability), and this limit is not known to be in perfect accuracy (uncertainty) [2]. The adoption of high penetration levels has significantly intensified the degree of variability and uncertainty involved in the short-term operation and long term planning. Furthermore, RES integration has become a challenging task due to the nonlinear and random nature of RES which involves definite constraints and nonlinear objective functions [3]. Different levels of RES generation penetration have different impacts on system generation scheduling. The uncertain and volatile nature of RES may pose many challenges to the power system such as imbalance between load and generated power, transient and voltage stability [4].
The wind and solar power generation sources are the most popular RES that can be considered as the most essential and sustainable energy resources [5][6]. As a must-taken energy, RES generation acts purely as load-shaving, leading to the concept of net load (or sometimes known as residual load). It is defined as the difference between the load and output of solar and/or wind output generation [7]. Therefore, in the prospective of the power system, the UC is an important optimization problem for daily operation and economic planning. It is the process of determining the optimal schedule of generating units over a set of a. The maximum penetration levels constrained by net load b. The impact of high RES penetrations on system schedules and c. The overall achievable cost saving under high RES penetrations.

UNIT COMMITMENT PROBLEM FORMULATION
In electric power systems, the UC problem formulation is mainly divided into two sub-solutions, namely unit commitment decision and economic load dispatch by determining the optimal generated power for each committed unit. Furthermore, this would satisfy generating units and system constraints over the scheduling period.

Unit commitment objective function
The on-off states of the generation units or the "commitment decision" contribute to the first step towards the optimal solution. It is the discrete variables that determine the state of on or off of a particular unit at any particular time. U n t , the unit n at hour t, is 1, if the unit is "on line" and 0 if the unit is "off line" and is represented by (1) [22].
Thus, the principal objective in UC is to prepare the on/off schedule of the generating units in every sub-period (typically 1 hour) of the given planning period (typically 1 day or 1 week) in order to serve the load demand and spinning reserve at minimum total production cost (fuel cost, start-up cost, shut down cost), while meeting all unit and system constraints. In this study, the main objective is to efficiently minimize the total operation cost (TOC) over the scheduling period. The TOC is subject to the fuel cost, start-up cost and shut down cost. The UC problem can be formulated as a mixed integer constrained, in which the overall objective function of the UC problem is described by (2

89
Where TOC is the total operating cost, N is the total generating units, M is the wind farm, S is the PV power plant, T is the time horizon which is 24 hours in this case study, ( ) oc w nt is the operating cost of the wind farm and ( ) oc pv nt is the operating cost of the PV power plant. The fuel cost of n-th thermal unit with the generating output p-th power at t-th hour   t t FP n n is expressed as a second order (parabolic) function of every unit output as follows.
The ,, a b c nnn are the fuel cost coefficient of n-th unit. The generator start-up cost t S n for restarting a de-committed thermal unit, which is related to the temperature of the boiler is included in this model. The t S n depends on the duration the unit has been turned off prior to start-up. By changing the on/off status of the units, the number of start-up and shut down and the type of units (hot or cold) will also change accordingly [23]. The start-up cost can vary from a maximum "cold-start" value to a much smaller value if the unit is only turned off recently and remains relatively close to the operating temperature [19], as presented by (4).

System constrains
In minimizing the objective function, the problem solution must respect both generator physical constraints and system operational constraints. These constraints include one or more of the following:

Generating unit limit constraint
The active power output of the generating unit must satisfy the minimum and maximum limits of the unit as given in (6).
are the minimum and maximum real power output, respectively.

Power balance constraint
The actual power output of the online committed units must be sufficient to satisfy the customers load demand for each hour and is given by (7).
where t D is the total demand at t-th hour.

Minimum up-time constraint
Minimum up-time is the minimum number of hours of operation at or above the minimum generation capacity. In other words, once the unit is committed to be online, it cannot be turned off for a specific number of hours.

Minimum down-time constraint
Minimum down time is the minimum number of hours once the generator is shut down before it is re-committed again to generate power.

Ramp rate up/down constraint
The change of the generating unit power output does not increase or decrease instantaneously. The change of this power output is restricted by ramp rate limits. These constrains are formulated based as on the following conditions.
where n UR and n DR are the ramping up and ramping down of n-th unit, respectively.

Spinning reserves constraint
Spinning reserve (SR) is an indicator of the percentage or amount of power that is required to fulfill the percentage of forecasted peak demand. It is also capable to make up the loss of the most heavily loaded unit in a given period of time. D t is the power demand at hour t. The formulation for SR can be seen in (12).
where t SR is the spinning reserve at t-th hour.

Must run unit
The must run unit is specified by the system operator to be committed on line for a particular interval of operation to address operating reliability considerations and voltage support on the transmission network or commercial considerations. The solar and wind are considered as must run units for better economic system operation.

Transmission constraint
Transmission constraint is the limitation of the transfer capability of transmission systems, such as thermal rating of individual transmission line and contingency constraint.

Capacity limit
The capacity limits of thermal units may change frequently due to maintenance or unscheduled outages of different types of equipment in the plant. This must also be taken into account in unit commitment.

PROPOSED IMPROVED PRIORITY LIST METHOD
The proposed improved PL method is composed of several processes and sub-problems which jointly lead to a feasible and cost-effective solution to the UC problem [23], [25]. The sub-problems involved are UC decision, minimum up/down time repairing, spinning reserve repairing and shut-down excess of power generation.
In this approach, the commitment order of each generator is based on its maximum production cost and heat rate. The solution obtained must satisfy the minimum up/down time constraints. The correction for this constraint satisfaction may lead to reduced total generated capacity. As a consequence, an extra generation capacity could be needed. The process of minimum up/down time repairing and spinning reserve repairing may lead to extra power generated at certain hours which will lead to the increase of generation cost. In order to obtain optimized generation and cost effective scheduling, shutdown of excess generated power must be carried out. Figure 1 represents the flowchart for the proposed improved PL method to clarify the system constraint satisfaction and optimization process.

Data set specification/data description
A ten unit system used in this research is to investigate the effect of solar and wind power output into the power system generation scheduling. The data from the generators and the 24-hour load demand profile are presented in Table 1 and 2 respectively [20]. The load profile is assumed to be similar throughout the year. The load consumption is observed to be high during day time. The initial status of the generator prior to the time frame is considered in the optimization. The negative number indicates down-time while the positive number indicates up-time. The start-up cost is modeled as a stepwise cost function with two steps. The first step is the hot start-up cost and the other step is the cold start-up cost. The reserve requirement is equal to 10% of the hourly load demand.

Solar radiation and wind speed data
A multiplying scale factor is used to model the wind and solar farms. By using this scale, the power production level or the penetration level of a single PV module and wind turbine is received. The power output of each PV cell and turbine is proportional to the resource potential in terms of solar radiation and wind speed respectively. The scale factors are generated separately for wind and PV for each penetration level/ scenario.
In this research, the actual solar radiation and wind speed data are collected at Kuala Terengganu, Malaysia [26]. The wind power output, P WTG generated from the wind turbine can be presented as: where P r is the wind turbines rated power; V ci is the cut-in wind speed; V r is the rated wind speed; V co is the cut-out wind speed. a and b are defined as: 3 3 On the other hand, solar radiation in Malaysia is relatively high in comparison with the world standards. It is estimated that the solar power in Malaysia is four times higher than the world fossil fuel resources [27]. The highest amount of solar radiation with an average of more than 3 kWh/m 2 throughout the year can be found in Malaysia. From this solar irradiation profile, the output power, P pv generated from the PV array can be calculated using (16).  (16) where G is the solar radiation ( KW/m 2 ); A pv is the PV area ( m 2 ); P pv is the PV module efficiency and is equal to 16% in this case study. The penetration level is defined as: Penetration Level 24 24

Scenario descriptions
In this study, four different scenarios are created to achieve the three main objectives of the work which are: a. Scenario 1: 0% RES penetration. b. Scenario 2: 14.5% solar-power penetration. c. Scenario 3: 27.5% wind-power penetration. d. Scenario 4: mixture of 10% solar-power and 10% wind-power penetration.
The maximum feasible penetration level in scenario II and III were determined as 14.5% and 27.5% for this case study before the total generation exceeds the amount of load at certain hours.

Scenario 2: solar-only
Based on the comparison to full load thermal generation cost, 14.5 % of solar penetration, which is equal to 3929.5 MW can save $85009. The shaded cells in Table 4 represent the solar penetration effect on the generation scheduling where the value of each unit output at a certain hour has been affected. Throughout the time horizon schedule, the scheduling hours at 13 and 14 hours are most affected by the solar power integration. At these hours, the generation scheduling is affected from the cheapest units (U1 and U2) to the most expensive units. This effect is clear where the output of some units is reduced while some other units are kept off.  2 pm in the afternoon. During the same time, the load demand is 1300MW. A 375.3 MW is generated from 6 units, satisfying the load demand and power system constraints as shown in Table 4.

Scenario 3: wind-only
By comparison to the full load thermal generation cost, the total cost saving achieved is equal to $149050 for wind power generation of 7452.5 MW at 27.5% wind penetration. It can be observed from Table  5 that the cheaper units of U1 and U2 remain ON continuously to share the major portion of load demand. The expensive units of U6 to U10 are OFF for the total operation hours as compared to the base case. Furthermore, it is noticed that the minimum thermal power output occurs during 14 hours with 5 units ON due to the highest generation of wind power at that time. Some units are kept within their minimum generation capacity at peak load hours to fulfill reserve requirement and generation constraints. The shaded cells in the following Table represent the wind penetration effect on the generation scheduling where the value of each unit output at that certain hour has been affected.

Scenario 4: mixture of 10% solar and 10% wind penetration
In this scenario, a mixed generation of solar and wind power resources is presented. A feasible penetration level of 10% for each resource is hybridized. The optimal scheduling of this scenario is given in Table 6. The 3 units that are most expensive and satisfied the load demand and the system constraints are kept off. The shaded cells in the following table represent the effect of mixed generation on the generation scheduling compared to base case system. From the analysis, the total cost obtained in this case is $503830 for 5420 MW power generated from both solar and wind. This combination is able to reduce $670 in total SUC. Table 6 and Figure 4 show that the hourly load demand minus the summation of solar and wind power output for each hour is supported by the thermal units. The minimum mix power generated is about 0.61 MW at the 6 th hour. The thermal power output is considered to be the highest and generated from the first 5 cheapest units (U1 to U5). The highest RES output is observed at the 14 th hour with 852.34 MW. The net load is fulfilled through U1 to U5 satisfying the load, reserves and systems constraints with the lowest price.

The impact of high variable res on overall operation cost
The variability associated with the wind and solar power outputs has a significant effect on the power system generation scheduling. According to Table 7, the lowest TOC is obtained in the case of 27.5% of wind penetration. Among all the cases, the wind power implementation in the modern power system offers lower generation cost. Table 7 confirms that wind generation is cheaper because of its availability throughout 24 hours and requires minimum installation area compared to the same daily output of solar system. The minimum SUC is with a higher percentage of wind.

CONCLUSION
In this paper, an improved method based on priority list is presented to deal with net load specifically. The improved method satisfies all the system technical constraints for optimal economic power system operation. The most crucial task of generation scheduling with stochastic RES power generation is more challenging. Hence, an efficient EPL algorithm is executed to achieve the optimal solution. The combined effects of these two RES on UC significantly reduce the total cost and enhance the net load profile. Furthermore, the maximum variability results in the least penetration levels in the case of solar power penetration scenarios. On the other hand, compared to solar, wind penetration levels which show less difference cause a reduction in the number of committed thermal units which lead to reduction in start-up cost.