Gujarat: Managing Solar Microgrid

1. Analysing Solar Radiation Data Over the Region of Interest

PDPU (Pandit Deen Dayal Petroleum University) Gandhinagar has provided with the solar output data. The data included energy generation, irradiance and insolation data from the 1MW power plant installed in PDPU campus using poly and thin films. Also understanding the cloud cover and the solar radiation received over the region, the pattern for the state will be analysed. The amount of solar radiation received over the period will then help plan the capital cost for installing solar panels. These calculations are compared with the Asanapura feeder data to find the optimum mix to take the complete feeder to solar. The following image shows the description of Asnapura feeder.

2. Sizing of Solar Panels for Irrigation

For calculating the size of the required solar installation, electricity generation data from a 1 MW solar PV plant and electricity demand for irrigating the farms fed through the feeder, for the year 2012 were used. The generation data pertained to a 1 MW solar PV installation belonging to PDPU and the electricity demand is the actual consumption data as metered for the ladol feeder by the utility.

Heuristic Process To Find The Optimal Panel Size

Both data were normalized at their respective mean values and outliers were removed. The remaining data was used to obtain the size of solar panels to meet the electricity need in the worst case scenario, by taking the ratio of electricity demand and generation from 1 MW solar panels.

Fig 1: shows a normalized comparison of the demand and generation data. It can be seen that there are several spikes in the generation data which are probably an error or days corresponding to equipment failure or shutdown. For most part of the year the data seems to be uniformly correlated implying a balanced loading of the system throughout the year.

Fig 2: As per the analysis the required size of the solar PV installation came to 2.73 MW, by considering the demand corresponding to 1 standard deviation. The fig. shows a graph of generation matched with the demand for a 2.73 MW PV capacity. It can be seen that the solar generation is able to meet the irrigation needs for major part of the year except for few days in the month of July and August.

Fig 3: When the sizing was done considering all the days, without discriminating those beyond 1 standard deviation. The required capacity was found to be 4.23 MW. The fig. shows a graph of generation matched with the demand for a 4.23 MW PV capacity. It can be seen that as expected the solar generation is able to meet the irrigation needs for the year. However for most part of the year their is an excess of un-utilized capacity available.

Fig 4: In Fig the demand and generation have been compared, with generation normalized at mean and demand normalized at maximum. It is a representation of a system which has been designed to meet the maximum demand with the mean solar radiation. By comparing the area under the two curves it can be seen that such a system would be generating 143% of yearly electricity demand.

Fig 5: In Fig the demand and generation have been compared, with generation normalized at mean and demand normalized at minimum. It is a representation of a system which has been designed to meet the minimum demand with the mean solar radiation. By comparing the area under the two curves it can be seen that such a system would be generating only 74% of yearly electricity demand.

The following analysis shows that Asnapura feeder could do with 2.73 MW or even lesser amount of solar panels. During the rest of the time one could change the pump cycles and also follow better scheduling practices in the field to reduce the demands.

3. Demand Shifting for Optimal Sizing of Solar Panels

In this section, we calculate the required solar panel area if one has perfect information about the solar power generation (i.e there is no uncertainty about the solar radiation and how much solar radiation will be observed throughout the year is known in advance) and schedules activities in the field based on this information to reduce the demand.

Here, linear programming models are constructed to determine the optimal solar panel area and IBM ILOG Optimization Studio (CPLEX) is used to implement and solve them. In the first analysis, we determine the required solar power generation capacity given the potential output curve for 1MW system for the whole year, if the farmer could schedule activities at an earlier (shifting the demand to the left in the following figures) or later time (shifting the demand to the right) when the solar generation is possible.

LP Model <Altenative 1>

dexpr float Objective=M;

minimize Objective;

subject to{

///////////////////POSTPONE-PREPONE Both Ways. Theoretical Minimum

forall(t in time:1<t<=365)

D_t[t]-S_Before_t[t]-S_After_t[t]+S_Before_t[t+1]+S_After_t[t-1]<=S_t[t]*M;

D_t[1]-0-S_After_t[1]+S_Before_t[2]+0 ==S_t[1]*M;

D_t[366]-S_Before_t[366]-0+0+S_After_t[365]==S_t[366]*M;

RESULT= 2.1075 ------>>> THIS IS THE THEORETICAL MINIMUM!!!

LP Model <Altenative 2>

When we have additional constraint which says the shift at time t can not exceed the demand at time t. (i.e. in a time period t, we can not postpone more than the demand at t.)

minimize M

subject to:

Dt - Pt + Pt-1St*M t:t>1

D1 - P1 S1*M

D365 + P364S365*M

RESULT= 3.3208

LP Model <Altenative 3>

This model makes the most sense. we say we can shift to tomorrow as much as the surplus of tomorrow.

Although we put a constraint on the postponed demand, this constraint does not guarantee that we can not shift more than 1 day.

dexpr float Objective=M;

minimize Objective;

subject to{

///////////////////POSTPONE

forall(t in time:1<t<=365)

D_t[t]-S_After_t[t]+S_After_t[t-1]<=S_t[t]*M;

D_t[1]-0-S_After_t[1]+0 ==S_t[1]*M;

D_t[366]-0+0+S_After_t[365]==S_t[366]*M;

forall(t in time:1<t<=366)

S_After_t[t]<=DUMMY[t];

forall(t in time:1<t<=366)

DUMMY[t]>=S_t[t]*M-D_t[t];

RESULT= 3.30663