Evaluating Singapore Electricity Market Risks: In Depth Analysis Using Data
Disclaimer: All information / data presented here are publicity available data and of my opinion. This does not represent the comment of any companies I have worked for, or with.
In Singapore, we benefited from relatively stable wholesale electricity prices, with minor peak and off-peak price differences. This is largely attributed to the highly efficient power transmission grid and customer focused market policies. It is useful to understand the electricity market in greater detail by analyzing the existing trends. To start off, I will show certain data points that demonstrates the market realities.
Historical Unified Singapore Electricity Price (USEP) / Volume — Source: Energy Market Company
USEP stands for Unified Singapore Electricity Prices, which is the half-hourly electricity settlement price by the market participants. Examples of market participants are electricity retailers and electricity generators. The following graph shows the USEP data from 23 April 2017 to yesterday (5 April 2018), with a total of 16618 datasets for analysis. The following graph shows the USEP prices across the whole year.
The following graph shows the USEP Prices separated by each month. The vertical axis shows the half-hourly USEP prices while the horizontal axis shows the half-hourly price. Naturally, February has a shorter x-axis as compared to January as February has lesser data points as compared to January.
Wholesale Energy Consumption — Source: Energy Market Company
The following graph shows the data for energy consumed in Singapore. We can deduce that the energy consumption follows a predictable pattern on weekly basis.
Historical Oil Prices — Source: Quandl
Historical oil is also used as a reference of energy prices.
Singapore Power Electricity Rate — Source: Singapore Power
Traditionally, the regulated tariff is known to mimic the oil trends and USEP prices, covering all underlying volatility risks.
Myth 1: The wholesale electricity prices are subjected to sudden spikes and the volatility is significant.
First, let’s establish the facts that occasional “spikes” do occur on wholesale electricity due to various reasons — it could be power plant generator failure, demand side instability, transmission grid instability, etc. These are unpredictable random events that we will classify as rare events. These rare events disrupt the supply-demand equilibrium of electricity market in varying magnitudes. Depending on the severity of the event, the impact on USEP prices could be significant.
Based on the historical USEP price, we calculate the basic statistical data for ease of comparison. Here we evaluate 3 separate cases:
Base case: Rare events are included as the basis of assumption. This demonstrates the simplest form of analysis on the data sets. No data sets has been excluded from analysis.
Adjusted case: the mean energy cost is calculated in consideration of the energy (kWh) consumed. This equation averages the energy cost with respect to daily consumption. This shows the true average USEP cost of a consumer / generator. The following formula is used to calculate the mean:
Please note that the adjusted case will vary according to individual usage pattern. For example, a datacenter, a manufacturing hub, and a commercial office will have different usage patterns that lead to a different adjusted USEP. You will need the actual profile of energy consumption to calculate the adjusted mean.
Eliminated Case: This model is used to calculate business as usual scenario assuming no rare events. This model improves adjusted case by statistically removing data points from random events for consideration. The random event is assumed as μ+2d (mean and two times standard deviation). The corresponding USEP demonstrates the amount that the market participant would be paying should there be no rare events during the event of dataset. The corresponding formula is below.
The following table summarizes the findings.
We derive several findings from these datasets. By looking into the data, we can be confident that the impact of rare events constitutes a certain percentage of risks. For a consumer assuming constant load throughout the day (i.e. no peak and off-peak power consumption), the expected average cost of electricity is would be as Base Case. Additional considerations should be factored in for customer with peak-load pattern (such as office hour or manufacturing facilities).
Conclusion: Considering the average mean and the eliminated mean, the impact of sudden fluctuation represented 3.137% additional risk into the USEP price. The long-term impact of market volatility represented $ 2.26 / MWh into energy cost.
Myth 2: The price of USEP follows a normal distribution curve
Statistically speaking, one could assume that a large enough data points could form a normal distribution curve. To understand this problem better, the data points are binned into 100 dollars brackets to evaluate the distribution model to be used.
Plotting the data points according to the bin ranges shows the following exponential and normal distributions.
Depending on how the data is being segmented, the representing statistical model can be different. We see that the probability of rare events follows exponential distribution while the probability of price distribution follows normal distribution.
It seems like plenty of surge in pricing are due to the rare events take place in Feb 2018 and March 2018. The data also shows that the mean of electricity is S$ 80–85 / MWh.
What we do not show is the moving distributions. As the graph shows, the USEP price can increase gradually over the months and hence affect the statistical data. This can be further studied.
Conclusion: USEP does follow a normal distribution for prices below S$ 100 / MWh
Myth 3: There is a noticeable growth in electricity consumption by observing the monthly data.
The daily total electricity generation over the historical trend is demonstrated below. In the following scenario, only monthly data is compared from May 2017.
There has been insignificant growth when comparing historical month-on-month data. According to EMA, there has been a graduate single digit growth when comparing historical year-on-year data. This myth remains unconfirmed until additional data can be provided. Note that as of writing, only data points up till 5 April have been collected.
Conclusion: Month-on-month data is insufficient to evaluate electricity growth.
Myth 4: Rising USEP Pricing is due to Rising HSFO
The current market perception is that the oil prices has been volatile in the past few months and is currently trending on the rise. Typically USEP should follow the oil prices as this constitutes the cost of electricity. We look at the monthly average USEP price and visualize the current trends. This has factored in the rare events. The following chart shows the monthly average USEP price.
The daily moving average shows a gradual upward trend since August 2017, and a drop since March 2018. This is also largely contributed by the recent HSFO price volatility. To provide a better intuition, the historical HSFO and USEP data are compared in the same chart. The data set was extracted from Quandl [SGX Platts Singapore Fuel Oil 180cst Index Futures, Continuous Contract #1 (1MF1) (Front Month)]
For those uninitiated, HSFO (180ct) is the oil price that is used by EMC according to this link. According to Singapore Energy Statistic, 95.2% of the energy in Singapore is generated by cost. It is assumed that oil prices and gas prices are perfectly correlated.
This demonstrates a that USEP loosely follows HSFO prices. There is a graduate and noticeable upward trend on USEP and HSFO at different period. By comparison, HSFO increased from 300 to 370 while USEP increased from 70 to 90. The intuition shows that there is a correlation between these two values, but the instantaneous correlation is not of statistical significant.
One other argument is that the oil prices might not impact the electricity prices immediately due to the commercial terms and agreement between gencos and suppliers. The structure of the oil prices could only impact the electricity price at a fixed lag period. From the chart above, USEP seems to lag HSFO by a quarter in moving trend. Further analysis could be evaluated to determine the magnitude of correlation between HSFO and USEP and the time lag for USEP to be impacted by HSFO.
Myth 5: There is a correlation between USEP price and demand
The is a perception that the USEP is directly correlated with demand and supply. By demand we mean the spot demand, and by supply we mean the electricitys selling price. The higher the demand, the higher the electricity price. We first remove the rare events (sudden price spikes), using data with USEP < SGD 100 MWh. The resulting scatter plot of USEP vs demand is shown below.
We exclude the consideration of rare events it upsets the supply-demand equilibrium drastically, which resulted in direct correlation.
Medium does not allow for high resolution picture update, here is the link to the high res image of this chart.
We attempt to fit a linear model to show the correlation between X(USEP price) and Y (demand). For a perfectly correlated relationship, at y= MX + C, the value is 1. For a perfected independent relationship, the R2 value is 0. Using linear model to represent the data, the corresponding R value is 0.0407. This shows there is not a strong statistical correlation between half-hourly demand and corresponding USEP prices.
Interestingly, this implies that USEP price is not necessarily positively correlated with demand. I will write more about the analysis of this data next time. This requires a more in-depth analysis to unravel the correlations. One quick intuition to look at is the impact of month-on-month USEP distribution and the need to further dissect the data for analysis.
This analysis could be proven useful for businesses to understand how they can generate business model that would be beneficial to the utilities or customers.
These are some initial analysis that I have made using publicity available data. I will expand in detail on the USEP data in the next post. If you have any comment / questions, you can reach out to me at hanwen.low@gmail.com. If you want to be notified of the next update, you may drop me an email as well.
Claps once if you like it, more if you love it.
