| dc.description.abstract | The main goal of the thesis is to study the value of snow measurements in the scheduling of hydropower plants. This is done by implementing an approximative Least-Square Monte Carlo(LSMC) method to calculate the scheduling. The optimization algorithm uses dynamic programming to find the optimal strategy. The spot price is deterministic, while the inflows are stochastic. Snow measurements are then included in the model, to calculate the Value of Information. Realistic data from a large-size Norwegian power plant is used to fit a normal distribution to simulate different inflow scenarios used in the model. The correlation between the snow measurements and the inflows are also studied to see how different snow levels influence the inflows in the flood period.
The analysis of the data showed that the maximum amount of snow and the total inflow in the flood period, week 16-32, is highest correlated with a correlation coefficient equal 0.85. The numerical testing of the LSMC-method showed that the spot price and the ratio between the total inflow and the upper reservoir limit have the biggest impact on the optimal strategy. The value of the snow measurements vary for different parameters in the model. When the reservoir is big compared to the total inflow, the snow has no value. When the reservoir is smaller, the probability for overflow is bigger and the snow measurements are valuable. The increase in value by using the snow measurements variates between 0 and 10 %. The annual production in Norway is 130 TWh, worth more than 4 billions NOK. This means that even a small improvement results in a big revenue.
If 40 % of the snow measurements are uncertain and in average 10% higher or lower than the real snow reservoir, the value of the measurement decreases with 25%. The value was also calculated for different correlation coefficients 0.55 and 0.25, and the value of the measurements decreased with 24 and 60%. This shows that even though there are some uncertainty in the measurements, the additional information is very valuable if the reservoir is small compared to the total inflow in the flood period. | |
