Forecasting solar power generation using real-time power data

1. A computer-implemented method of forecasting power generation, the method comprising:

receiving, using a processor, a first power generation forecast;

creating, using the processor, a second power generation forecast, wherein the second power generation forecast is created by:

analyzing, using the processor, a set of historical power generation data;

determining, using the processor, a set of clusters, wherein each cluster of the set of clusters represents power generated during a time period;

receiving, using the processor, power generation data for a portion of the time period;

identifying, using the processor, a cluster of the set of clusters that contains historical power generation data that is most similar to the power generation data; and

forecasting, using the processor, power generation for a remainder of the time period, using the historical power generation data of the cluster of the set of clusters to determine the second power generation forecast;

receiving, using the processor, an observed power generation;

creating, using the processor, a model for the observed power generation based on the second power generation forecast;

calculating a first error between the first power generation forecast and the observed power generation;

calculating a second error between the model for the observed power generation and the observed power generation; and

creating, using the processor, a third power generation forecast based on a determination that the first error is lower than the second error, the third power generation forecast comprises a weighted average of the second power generation forecast and the model for the observed power generation.

2. The computer-implemented method of claim 1 , wherein creating a weighted average comprises:

calculating a first weight and a second weight using the second power generation forecast and the model for the observed power generation; and

calculating the weighted average by summing the product of the first weight and the second power generation forecast with the product of the second weight and the model for the observed power generation.

3. A computer system for forecasting power generation, the system comprising:

a memory; and

a processor system communicatively coupled to the memory;

the processor system configured to perform a method comprising:

receiving a first power generation forecast;

creating a second power generation forecast by:

analyzing a set of historical power generation data;

determining a set of clusters, wherein each cluster of the set of clusters represents power generated during a time period;

receiving power generation data for a portion of the time period;

identifying a cluster of the set of clusters that contains historical power generation data that is most similar to the power generation data; and

forecasting power generation for a remainder of the time period, using the historical power generation data of the cluster of the set of clusters to determine the second power generation forecast;

receiving observed power generation;

creating a model for the observed power generation based on the second power generation forecast;

calculating a first error between the first power generation forecast and the observed power generation;

calculating a second error between the model for the observed power generation and the observed power generation; and

creating a third power generation forecast based on a determination that the first error is lower than the second error, the third power generation forecast comprises a weighted average of the second power generation forecast and the model for the observed power generation.

4. The computer system of claim 3 , wherein creating a weighted average comprises:

calculating a first weight and a second weight using the second power generation forecast and the model for the observed power generation; and

calculating the weighted average by summing the product of the first weight and the second power generation forecast with the product of the second weight and the model for the observed power generation.

5. A computer program product for forecasting power generation comprising:

a computer-readable storage medium having program instructions embodied therewith, the program instructions readable by a processor system to cause the processor system to perform a method comprising:

receiving a first power generation forecast;

creating a second power generation forecast by:

analyzing a set of historical power generation data;

determining a set of clusters, wherein each cluster of the set of clusters represents power generated during a time period;

receiving power generation data for a portion of the time period;

identifying a cluster of the set of clusters that contains historical power generation data that is most similar to the power generation data; and

forecasting power generation for a remainder of the time period, using the historical power generation data of the cluster of the set of clusters to determine the second power generation forecast;

receiving observed power generation;

creating a model for the observed power generation based on the second power generation forecast;

calculating a first error between the first power generation forecast and the observed power generation;

calculating a second error between the model for the observed power generation and the observed power generation; and

creating a third power generation forecast based on a determination that the first error is lower than the second error, the third power generation forecast comprises a weighted average of the second power generation forecast and the model for the observed power generation.

6. The computer program product of claim 5 , wherein:

creating a third power generation forecast comprises creating a weighted average of the second power generation forecast and the model for the observed power generation; and

creating a weighted average comprises:

calculating a first weight and a second weight using the second power generation forecast and the model for the observed power generation; and

calculating the weighted average by summing the product of the first weight and the second power generation forecast with the product of the second weight and the model for the observed power generation.