Image-based solar estimates
An example device is configured to determine, based on a sky image of a portion of sky over a power distribution network and using a convolutional neural network (CNN)-based image regression model, an estimated global horizontal irradiance (GHI) value and manage or control the power distribution network using the estimated GHI value. The device may also be configured to determine, based on GHI values and aggregate load values for at least a portion of the power distribution network, using a Bayesian Structural Time Series model, an estimated photovoltaic power output value for the at least a portion of the power distribution network. The device may manage or control the power distribution network using the estimated photovoltaic power output value.
1. A device comprising:
at least one processor configured to:
determine, based on a sky image of a portion of sky over a power distribution network, an estimated global horizontal irradiance (GHI) value;
determine, based on the estimated GHI value and an aggregate load value for at least a portion of the power distribution network, using a Bayesian Structural Time Series (BSTS) model that estimates photovoltaic (PV) generation based on GHI values, an estimated PV power output value for the at least a portion of the power distribution network; and
cause at least one device in the power distribution network to modify operation based on the estimated PV power output value for the at least a portion of the power distribution network.
2. The device of claim 1 , further comprising a sky imaging device configured to capture the sky image.
3. The device of claim 1 , wherein the BSTS model comprises a state space model that captures time evolution of state variables.
4. The device of claim 1 , wherein the BSTS model estimates the PV generation based additionally on temperature values.
5. The device of claim 1 , wherein causing the at least one device to modify operation based on the estimated PV power output value for the at least a portion of the power distribution network comprises outputting the estimated PV power output value for the at least a portion of the power distribution network.
6. The device of claim 1 , further comprising determining the estimated GHI value using a convolutional neural network (CNN)-based image regression model.
7. The device of claim 6 , wherein the CNN-based image regression model comprises five convolutional blocks and a flattening/densifying block.
8. The device of claim 7 , wherein each of the five convolutional blocks is configured to perform a respective convolution, a respective batch normalization, respective operation by a rectified linear unit (ReLU), and a respective max pooling.
9. A method comprising:
determining, by a computing device comprising at least one processor, based on a sky image of a portion of sky over a power distribution network, an estimated global horizontal irradiance (GHI) value;
determine, by the computing device, based on the estimated GHI value and an aggregate load value for at least a portion of the power distribution network, using a Bayesian Structural Time Series (BSTS) model that estimates photovoltaic (PV) generation based on GHI values, an estimated PV power output value for the at least a portion of the power distribution network; and
causing at least one device in the power distribution network to modify operation based on the estimated PV power output value for the at least a portion of the power distribution network.
10. The method of claim 9 , wherein the BSTS model comprises a state space model that captures the time evolution of state variables.
11. The method of claim 9 , wherein the BSTS model estimates the PV generation based additionally on temperature values.
12. The method of claim 9 , wherein determining the estimated GHI value comprises determining the estimated GHI value using a convolutional neural network (CNN)-based image regression model.
13. The method of claim 12 , wherein the CNN-based image regression model comprises five convolutional blocks and a flattening/densifying block.
14. The method of claim 13 , wherein each of the five convolutional blocks is configured to perform a respective convolution, a respective batch normalization, respective operation by a rectified linear unit (ReLU), and a respective max pooling.
15. The method of claim 12 , further comprising generating, by the computing device, based on a plurality of sky images of the portion of sky and a respective plurality of GHI values, the CNN-based image regression model.
16. The method of claim 9 , wherein causing the at least one device to modify operation based on the estimated PV power output value for the at least a portion of the power distribution network comprises outputting the estimated PV power output value for the at least a portion of the power distribution network.