# Forecasting Energy Generation

## Sep 22, 2018 00:00 · 659 words · 4 minute read

The purpose of this exercise is to examine policy impacts on the energy landscape in Hawaii. In this case, the forecast is used as a temperature check - we are interested in the trajectory of energy generation to the system.

## Background

2018 marks the 10th anniversary of the Hawaii Clean Energy Initiative, which has set a goal for Hawaii to achieve 100 percent clean energy by 2045. Reaching this goal has been helped by advancements in renewable energy. Rooftop solar systems, for example, have been widely adopted across residential and commercial sectors in recent years. Hawaii’s utilities are in the process of diversifying its energy portfolio by including more renewable sources. With that said, an interesting way to examine the changing landscape brought by this policy is to study energy generation and delivery to the grid.

## Data

Public available data was downloaded from the DBEDT Data Warehouse. The indicators follow the standard and non-technical understanding of calculating enery generation at a utility. For the most part, the total generated energy is the sum of steam, diesel, biodiesel and hydro/wind. A few observed months were not equivalent, but that could be due to rounding error, etc. The total energy delivered to the system is the sum of all generated energy by the utility and independent power producers (IPP) minus the energy loss.

## Declining Dependency Towards Utilities

### Consumers rely less on utility companies for energy generation.

Overall, it appears that consumers are relying less on the utilities for their energy. This makes sense as more consumers have invested in distributed energy resources (DER) such as rooftop solar to curb their dependency on utilities.

#### Augmented Dickey-Fuller test

Test Statistic                -0.916453
p-value                        0.782506
#Lags Used                           12
Number of Observations Used         137
Critical Value (1%)            -3.47901
Critical Value (5%)            -2.88288
Critical Value (10%)           -2.57815
Stationary                        False
dtype: object


#### Forecast of the Total Energy Delivered to the System

The forecast suggests that energy delivered to the system will continue on a declining trajectory. The data was fit using the SARIMAX model. The parameters were tuned using Bayesian optimization after gathering parameter ranges using convention EDA methods. Prediction accuracy was evaluated using the MAPE, which was at 3.29.

#### Example of Bayesian optmization output

Bayesian Optimization
-----------------------------------------------------------------------------------------------------
Step |   Time |      Value |         D |         P |         Q |         d |         p |         q |
16 | 00m01s | 4282.29923 |    0.0000 |    2.0000 |    2.0000 |    1.0000 |    4.0000 |    0.0000 |
17 | 00m01s | 4241.97743 |    0.0000 |    0.0000 |    2.0000 |    1.0000 |    4.0000 |    4.0000 |
18 | 00m01s | 4267.45176 |    0.0000 |    0.0000 |    2.0000 |    1.0000 |    0.0000 |    4.0000
...
22 | 00m02s | 3720.84820 |    1.0000 |    0.0000 |    2.0000 |    1.0000 |    2.0000 |    4.0000 |


## Renewable Energy is Growing

### Traditional forms of energy are being replaced by renewable sources.

A hierarchical time series approach was taken to understand trends by various generation sources.

#### Generation from IPP

When stratifying the utility from the IPPs, we see that less energy generated by the utilities is being delivered to the system. These trends reflect anticipated shifts as the utility companies expand their portfolio to include renewable energies by independent companies such as grid-scale solar farms.

#### Generation from hydro/wind

In addition to renewable energy being provided by IPPs, utilities are also diversifying their own energy portfolios by replacing steam and diesel with energy from biodiesel, hydro and wind. However, it seems that IPPs tend to have a much larger share of the portfolio.

## Conclusion

The analysis sought to explain policy impacts on energy generation. The results suggest that energy dependence from utilities are on the decline. In addition, the findings show that Hawaii’s energy portfolio is increasingly comprised of renewable energy. When it comes to making changes in the energy sector, policy definitely plays an influential role.