Electrical power generated by solar Photo-Voltaic (PV) is one of the best options for sustainable energy requirements but the variability and unpredictability inherent to solar create a threat to grid reliability due to balancing challenge in load and generation. Variability of power represents the change of generation output due to unscheduled fluctuations of solar radiation patterns. Large unscheduled changes in solar power generation are called ramp events which hamper the penetration of variable power in the existing grid. The variability in PV power is not a major issue with small-scale systems, but a large scale grid-connected system needs sufficient statistical analysis to model the power fluctuations to assess the reliability of the system. The variability can be quantified as a measure of dispersion in variable renewable power generation. Though there are different ways to quantify variability, the simplest way to measure variability is using the normalized squared deviation about mean of the actual generation for some time blocks.
To accommodate variability, the short-term and long-term forecasting became an important tool in asset management, operations and maintenance of PV solar generation. The recent developments of deep learning algorithms in ANN (Artificial Neural Network) based methodologies using NWP (Numerical Weather Prediction) models have created a huge scope in forecasting the solar power generations with acceptable error margins. Since there are computational issues like the uncertainty in the initial value vector of NWP, proper availability of learning vectors in DNN (Deep Neural Network) and choice of proper architecture of DNN, forecasting can be viewed as statistical prediction rather than a problem with deterministic solutions. Thus from an analytical viewpoint, forecasting can be regarded as the temporal evolution of probability distributions associated with variables required in predicting the solar power generation. Predictability can be viewed as the ability to forecast the solar power generation with sufficient accuracy such that penalty due to error-deviation is very small; and predictability can be measured using probability.
Variability of actual solar power generation is the characteristics of solar plant whereas the predictability refers the property of proper forecast models. A good forecast model is that which captures the genuine patterns in the historical data, but does not replicate past events that will not occur again. Though there is a common belief that the high variability of power generation reduces predictability, the predictability of good forecast is independent of variability. If predictability depends on the variability of the solar generation, then either there is some issues in the basic architecture of the forecast model or the forecast system is not running in its full potential.
-By Abhik Kumar Das, del2infinity Energy Consulting
- Font Size
- Reading Mode