Although the current system has multiple capabilities and the future technology has a bright future, there are still many basic scientific problems in the development process of using satellite information for solar energy prediction. Some problems can indeed be solved by introducing more advanced observation systems, improving the details of radiation transmission calculations, and inputting more information to the NWP model. But there are still some difficulties that force us to admit that there are limitations in observation and predictability, which make us focus our research on the continuous development of more observation and prediction technologies. We only selectively introduce some research needs in short-term to long-term solar forecasting.
1. Three-dimensional effects of short-term forecasts
Ultra-short-term solar energy prediction needs to predict the small and medium-scale details of the downward radiation field, including capturing high-frequency sudden changes due to cloud shadow occlusion; cloud shadows will also cause local cloud areas to affect the scattered radiation in the sky, which is obviously a problem To solve this problem, it is necessary to understand the precise cloud distribution (horizontal and vertical) and movement trends, as well as a complete description of the observation system and the geometric structure of the sun.
Figures 1 and 2 illustrate the importance of cloud height in determining the current and future effects of clouds on the surface radiation. Without specifying the satellite parallax and the geometric position of the sun, it is not enough to indicate the radiation influence of clouds on the ground on the original satellite image. In the same way, a single wind direction indicator cannot explain the speed and direction of changes in all heights of clouds. At this time, the satellite information about the cloud height helps to correct the cloud shadow distribution and short-term horizontal movement.
Even after considering all geometric corrections, the radiation field at a certain location is not just a function of a single cloud pixel, but a function of the neighboring area of cloudy and clear sky pixels, which make up the field of view of the ground station. Clouds with partial gaps will affect the scattered radiation in the sky. For example, such clouds often cause the downward radiation illuminance to exceed the predicted value. Explain clearly that these multiphase cloud fields need to be parameterized on the basis of the three-dimensional radiative transmission model.
2. Improved use of satellite retrieval cloud products in NWP analysis
Figure 3 shows the true level of the current NWP model in representing the observed cloud field. The weather scales of the two are similar, but the differences between the macro and micro scales are getting bigger and bigger. Although the short-term prediction of 1~3h is inherently problematic (too few constraints), in order to better represent the cloud layer on the micro-scale, using satellite observations to initialize the NWP-model cloud field is the key first step to solve this problem. Simply put, multiple system states can affect the reflectance and brightness temperature measured by the satellite. If the model assumes one of these states without proper constraints, it will cause serious distortion. At the other extreme, when analyzing and observing the moisture content of the cloud layer at a certain location, if the cloud layer state is not changed to the state of the current cloud layer, invalid predictions such as improper cloud evolution or rapid disappearance will occur.
Multi-cloud data assimilation generally encounters a difficult problem, that is, how to make reasonable assumptions in the process of modifying the environmental state of the model, so as to support the current cloud cover and avoid the model’s serious error loop within the target prediction window. From satellite cloud inversion, the height of the top of the cloud and the integrated water path can be understood. For example, satellite cloud inversion can modify the specified atmosphere-humidity profile, so that the environmental state of the model is consistent with the observed value, and supports the subsequent evolution of clouds to improve short-term prediction results.
3. Fusion of simulation and observation
Although the cloud has a complex visual performance, it is essentially a kind of atmospheric circulation, which involves temperature, humidity, dynamics and other characteristics of the air mass. At present, although people know very little about cloud cover, it is still one of the most important NWP elements. The NWP model can predict the actual state of the environment, but it will still distort some details of the cloud field. On a time scale of more than a few days, the ability of the NWP model to represent cloud cover limits its own forecasting capabilities. At this time, it may be more effective to adopt a hybrid method, that is, under the guidance of conventional NWP forecasting flow patterns, use the cloud amount statistics data observed by satellites to make predictions. Figure 4 illustrates how the satellite’s local cloud climate changes drastically with different atmospheric flow conditions.