Demand Shifting Bidding in a Hybrid System with Volatile Wind Power Generation

Price responsive demand shifting bidding mechanism is discussed as an alternative solution to deal with intermittency in wind generation. This paper proposes a formulation of social welfare equation with price responsive demand shifting bidding and economic emission dispatch with emphasis on integration of wind power. The analysis is based on the IEEE 30 bus test system generation data, with conventional and wind generation plant over a period of 24 hours. It has been demonstrated that the proposed approach leads to reduction in emission as well deal with intermittency in wind generation.


Introduction
Due to environmental and energy security benefits there is a positive shift towards the production of electrical energy from renewable sources of energy especially from wind which are clean and abundantly available in nature.On regulatory side in India and many other countries, there are necessities to generate a certain amount of electrical energy from renewable sources.China is the country with the largest installed wind power capacity in the world at the end of year 2010 whereas India's total installed wind power capacity is fifth in the world.It is reported by the Global Wind Energy Council (GWEC) that global installed wind power capacity increased by 24.1% during the year and stands at 197.0 GW in 2010 [1].Large capacity wind power generators are connected to transmission or sub-transmission systems.At the end of 2010, India had 13.1 GW of installed wind capacity, with 40% operating in the southern state of Tamil Nadu and wind power potential estimated by the Centre for Wind Energy Technology (C-WET) is 49.13 GW [2].
The generation of electric power by conventional sources produces mainly sulfur dioxide, carbon, NOx, and mercury emissions causing acid rain, urban smog, and eventually global climate change in addition to posing significant health risks.Renewable electricity generations mainly from Wind farms help to prevent release of emissions into the atmosphere preventing environment damage.On the other hand, unpredictable, intermittent and volatile nature of wind energy may threaten power system characteristics such as voltages, frequency and generation adequacy which can potentially enlarge the weakness of power systems.
Demand side management (DSM) incorporates energy efficiency (EE), Energy Conservation (EC) and Demand Response (DR).In the most electricity markets; the consumers play a much more limited role than producers.It is widely acknowledged that a more active participation in the market by the demand side could have significant benefits [3].A good deal of research has been reported on measurement of load elasticity, presuming that industrial, residential and commercial consumers will respond to price signals [3]- [8].DR at end user's premises can reduce generation, transmission and distribution capacity of utility at relatively fractional cost as compared to investment required to provide new capacity.The appropriate demand management in modern interconnected power system with dispersed generation may also result in reduced associated CO2 emissions in day-ahead electricity markets through price responsive demand shifting bidding (PRDS).Su and Kirschen [9] proposed the PRDS bidding for market clearing mechanism of day-ahead markets.PRDS bidding quantify the demand response in day-ahead market, and some responsive customers are able to shift the demand from periods of high locational marginal price (LMP) to the periods of low LMPs.However, the market clearing mechanisms developed in [9] do not take into account the operational and security constraints of transmission networks.Kanwardeep Singh et al. [10] discussed the influence of PRDS bidding on congestion and LMP in Pool-Based Electricity Markets.Impacts of availability based tariff on wind power trading option were analyzed in [11].
This paper investigates these unmatched challenges caused by wind power plants to the optimization problem.PRDS bid, emissions constraints, and fuel costs are considered in the realization of most favorable generation mix for a system with wind power generation.Fuel costs, environmental costs and emissions are considered in the implementation of optimal generation mix for a system with wind generation along with PRDS bid to maximize social welfare.The rest of the paper is organized as follows: Next section 2 describes wind power scenario reduction and PRDS biding mechanism.Section 3 describes research method.Results are presented and discussed in section 4. Finally, section 5 concludes the paper.

The Proposed Method
The wind power participation into total production of electrical energy depends upon the forecast of wind momentum.A principal difficulty with modeling wind power production is that the relationship of wind speed to wind power production is extremely nonlinear.The wind power generators require no fossil sources hence, the operational cost of wind units has been assumed to be zero.Different forecasting approaches available can be studied in [12]- [13].Result of wind power unpredictability can be studied by applying different scenarios into the model.Monte Carlo simulation was popularized by scientist in the 1950s.Monte Carlo simulation is a method that can model thousands of scenarios and helps to model uncertainties of wind power output.It provides a range of possible outcomes together with there probability.Modeling all main and possible scenarios determined by the uncertain variables can be done by Monte Carlo simulation.These scenarios are defined by the probability distributions and their simulation parameters.Many types of probability distributions are used in different situations such as normal, uniform and triangular distributions.
There are many sampling techniques such as Importance sampling, Sobol numbers sampling, Midpoint sampling, Latin hypercube sampling (LHS), and LHS Monte Carlo sampling to eliminate scenarios with very low probability.These techniques are engaged to reduce the computational requirement to simulate large number of scenarios.LHS has the benefit of generating a set of stratified samples that more precisely reflect the shape of a sampled distribution and reduces the number of runs.The general effect is that the mean of a set of simulation results more quickly approaches the 'true' value, particularly for models that are simply adding or subtracting a number of variables.The tradeoff between the number of reduced scenarios and the simulation precision is possible by choosing the number of reduced scenarios so that the objective function would not change much or the relative distance between original scenarios and reduced scenarios is within an acceptable level [14], [15].In this model the number of reduced scenarios is chosen to be ten since the value of objective function at this number does not change much.The scenarios taken were having the higher probabilities.The deviations in wind power have been taken in to account by considering different scenarios.The forecasted wind power generated and reduced scenarios data are taken from [16].
Not all consumers have the facility or the incentive to adjust their demand when prices change.Large part of the power requirement will therefore totally inelastic.In price taking bids, the demand aggregator is ready to accept a specified amount of power at prevailing market price, and its power consumption remains constant irrespective of variations in market price.This kind of bid is required to meet necessary daily services to industrial, residential and domestic loads.In price responsive bids, the price to be paid by a bidder decreases consistently decreasing with respect to increase in power use.Details of PRDS bidding scheme has been formed in [9].In PRDS bids, an aggregator on behalf of consumers is able to increase or decrease its certain percentage of demand in response to market price.Key factor in PRDS bidding scheme is price responsive consumer who can transfer its demand from peak demand periods of high market price to off peak demand periods in which market price is somewhat low.
The scope of present paper is to study the influence of PRDS bidding in a hybrid system with volatile Wind Power Generation in day-ahead markets.A usual PRDS bid is shown in [10].Main dissimilarity between PRDS bidding scheme and price responsive demand bidding is that in PRDS bidding during a particular th t period, maximum demand limit can be greater than earlier, to include the fall of load occurred during peak periods due to soaring electricity rate, whereas in price responsive bidding scheme loss of load during peak hours can not recovered in off peak hours.In PRDS bidding aggregator's on behalf of responsive consumers specifies for particular th t period, its maximum, and minimum price bids and maximum power demand which in simplest form can be sum of entire energy need of responsive part.Due to negative slope of PRDS bidding responsive part of demand can be less than its maximum value because aggregator would accept only that part of demand for which its willing price is less than or equal to market declared price.In PRDS bidding aggregator's price responsive part of energy of scheduling period can be consumed in few sub-intervals.In simplest form it can be consumed in a single period.Mathematically, it can be represented as: , , Price taking consumers have infinite marginal value due to vertical curve with respect to power so it can not be included in consumer gross surplus.In optimization problem gross surplus of this type consumers are not included and assumed constant.Thus equations ( 4) and (5) shows how the consumer gross surplus is calculated based on the accepted demand-side bids and the marginal value that consumers attach to these bids: where: In this model certain percentage of power for each hour which is price taking can be procured by consumers regardless of market-clearing prices.For a complete and compact formulation of the PRDS bidding mechanism the reader is referred to [9], [10].In the optimization program used for market-clearing, the bid specifications are translated into constraints: (i) on the demand during each period and (ii) on the total demand over the optimization horizon.This last specification is implemented as an inequality rather than an equality constraint because a demand-side bid below the lowest price at which generators are willing to produce would otherwise prevent the market from clearing.The price responsive bids are converted into a form suitable for mixed-integer linear programming such as in [9].

Research Method
The objective is to maximize the social welfare, i.e., the difference between the value that consumers attach to the electrical energy that they buy and the cost of generation that has been formulated based on classical ELD with emission.Equation ( 6) as proposed in [9] is used to consider economic load dispatch with emission.The proposed social welfare equation after modification is represented as: where: The th i generator startup cost is described in term of the number of hours the generator has been down: where: The fuel cost curves and emission curves of the power plants are modeled as step-wise linear function, in order to approximate the typical quadratic shaped cost curve of a power plant.The dual-objective combined economic emission dispatch problem is converted into single optimization problem by introducing price penalty factor h which convert emission output into equivalent emission cost as follows: The price penalty factor h merge the emission with fuel cost and sum of merged emission cost, fuel cost and start up cost gives total production cost in $/hr [17].Modified price penalty factor approach was proposed in [18] to give exact total operating cost.The price penalty factor i h is the ratio between maximum fuel cost and maximum emission of th i generator: ,max ,max $/lb 1,2, 3 ........ where: ,max i FC Fuel cost of an th i unit at maximum power output; ,max i EC Emission of an th i unit at maximum power output; To find out the modified price penalty factor for a particular load demands following steps are proposed in [18].• The price penalty factor i h is calculated.
• Price penalty factors i h are arranged in ascending order.
• Add the maximum capacity of each unit If a unit must be "on" for a certain number of hours before it can be shut down, then a minimum up-time ( i UP T ) is imposed.On the contrary, the minimum down-time ( i DN T ) is the number of hour(s) a unit must stay off-line before it can be brought on-line again.
Mathematically, the minimum up/down time constraints for th i unit can be expressed as: where: Amount of time th i unit has been running; Equations ( 13), ( 14) are nonlinear.It can be linearized using the method presented by Chang et al. [20].The UC schedule should provide the exact amount of power to meet the consumer's demand.Therefore: where:

Results and Analysis
Mathematically social welfare equation is a decision problem with an objective to be maximized with respect to a series of prevailing equality and inequality constraints.The equation is a mixed-integer non linear problem and includes a large number of integers and continuous variables.Non linear part is converted into piecewise linear function using the technique given in [9].The market-clearing algorithm described in Section-3 has been applied to several scenarios to assess economic viability of demand shifting and evaluate its impact on emission dispatch and on wind scenarios.The platform used for the implementation of this proposed approach is on INTEL[R], Pentium [R] 4 CPU 3.06 GHz, 512 MB of RAM.Many commercial packages such as CPLEX, LINDO, OSL and XPRESS-MP exist in the market place have been successfully applied to UC problems.In this paper, we use XPRESS-MP to solve the problem [19].
The test system used in the studies consists of IEEE 30 bus system with a total capacity of 435 MW.The IEEE 30 bus system has six generating units.The characteristics of generators, unit constraints and the emission coefficients are given in Table I.The maximum and minimum loads are 396.76MW and 183.4 MW, respectively, while the total system forecasted energy demand is 6934.76MWh.The study period is 24-hours.The 24-hour system load and forecasted wind power are presented in Table 2.The proportion of the demand that responds to prices affects the shape of the demand curve.Load participation factor (LPF) is defined as the ratio of the price responsive demand to the total possible demand [9].The simulated wind power scenarios are assumed to follow a normal distribution with a standard deviation (volatility) of 10% of expected values which is the forecasted value and in the following studies, the number of reduced scenarios is chosen to be ten, taken from [16].The following three cases are discussed in this paper: Case 1: Case 1 is the base case without PRDS bidding which applies 11 commitment and dispatch for the forecasted wind power and the other ten simulated wind power scenarios, without considering the correlation between scenarios.Case 2: In this case, we observe the impact of PRDS on system operation and compare generation dispatch and total system operating cost with and without PRDS.
Case 3: Demonstrate the relationship between LPF and emission at different value of load participation factor from 0 to 0.1.

Social Welfare without PRDS bidding
With forecasted wind power given in Table 2, we solve the social welfare equation without inclusion of demand shifting and determine the dispatch of non-wind units given in Table 3(a).The cheapest Unit 1 and 2 are always committed.The more expensive Unit 3 and 4 are committed between Hours 1 and 22. Unit 5 is committed between Hours 16 and 21.Unit 6 is committed between Hours 2-7 and 15-21.The system generation cost is $16080.296.To observe the impact of wind power scenarios, we solve ten social welfare equations.The operation costs are shown in Figure 1(a) which range from $16185.804 for Scenario 5 to $15837.475 for Scenario 9. Table 4(a).shows the social welfare equation solutions for the 10 scenarios.Each value in the table shows the number of times certain units are ON in the 10 scenarios.Here, Unit 3 and 4 are ON mainly between Hours 1-22, while Unit 5 is ON mainly between Hours 16-21, while Unit 6 is ON mainly between Hours 3-7 and Hours 15-21.

Social Welfare with PRDS bidding
To observe the impact of demand shifting we take LPF 0.1.The load and wind profile are same as Case 1. Table 3(b).presents the dispatch of non-wind units for 24 hours.The system generation cost reduces to $15657.407.In all 10 scenarios operation costs are shown in Figure 1(b) which range from $15773.616 for Scenario 5 to $15330.454 for Scenario 10.In this case, for forecasted and all ten scenarios the expensive unit 5 is not dispatched as shown in Table 4(b).These results show the lower cost of using PRDS bidding for supplying the load in the system.In this section, we study the sensitivity of social welfare results to the size of PRDS bidding with and without wind power.For the same system we use different PRDS with several LPF options as shown in fig.1(C).It can be seen that with wind power total emission reduces from 9275.748 lb to 6628.245 lb when we do not consider PRDS.By increasing the size of PRDS, the total emission further reduces to 6449.473 lb.

Conclusion
A MIP based social welfare problem including wind and PRDS bidding is described in this paper.The case studies based on an IEEE 30 bus system generator data indicate that the application of PRDS can impact the peak load reduction, system operating cost, emission reduction, commitment and dispatch of the units.Much of the benefits listed here will depend on the load participation factor of PRDS bidding.The example on IEEE 30 bus system generator data showed the effectiveness of the proposed model.The proposed algorithm can be used for the operation planning in the day-ahead as well as the long term planning of wind units in a constrained thermal power system.

Figure 1 .
Figure 1.System operation cost ISSN: 1693-6930 Demand Shifting Bidding in a Hybrid System with Volatile Wind Power …. (D.K. Agrawal) 41 Table 4. Statistics of Unit Commitments in Scenarios 37In this optimization problem, for the sake of simplicity, the start-up costs are taken constant, which can be given as: i κ th i unit down at th t hour; i τ Demand Shifting Bidding in a Hybrid System with Volatile Wind Power …. (D.K. Agrawal)

Table 2 .
Hourly load and forecast wind power

Table 3 .
Generation Dispatch (MW) with Forecasted Wind Power