Itaú BBA - Energy rationing risks are likely in 2015 if rain repeats recent pattern

Macro Vision

< Back

Energy rationing risks are likely in 2015 if rain repeats recent pattern

November 25, 2014

In 2014, risks increased further with low rainfall in January and February.

• The low levels of reservoirs once again bring rationing risks, now for 2015. To evaluate these risks, we estimate reservoirs levels on April 30, 2015 (end of the rainy season) based on the  rainfall observed in the past 14 years. 

• If rainfall over the next six months repeats the pattern seen in the last three years, the reservoirs will reach the end of the rainy season at even lower levels than in April 2014, which will likely lead to some form of rationing.

 • If rainfall is stronger, in line with the amount seen between 2000 and 2011, the reservoirs will reach the end of April still below the seasonal average for the period, which will likely require above-average use of thermal plants, but not rationing. 

Low reservoirs (and higher risks) since the end of 2012

The perception of risks for the electricity supply began in 2012, after a sharp fall in reservoirs, to 30.5% of their capacity[1] by the end of December of that year. Rainfall above the historical pattern between March and July 2013 and increased use of thermal plants eased the situation, leading reservoirs to 42.7% capacity by the end of 2013.

In 2014, risks increased further with low rainfall in January and February. Despite the strong use of thermal plants throughout the year and rainfall at 99% of the historical average between the second half of February and the end of September, the reservoirs remained at levels that suggest a risk of rationing. The dry weather in October (the beginning of the rainy season) aggravated the problem.

Low rainfall (94% of the historical average between November 2012 and October 2014, according to our weighting) does not in itself explain the long period of low reservoirs. The situation is also caused by:

• Delay in construction of new power plants and transmission lines;

• High consumption growth, despite the recent slowdown of GDP;

• Evidence that the official models overestimate the power-generation capacity for a given intensity of rainfall. This discrepancy may be associated with difficulties in transmission lines and diversion of water from rivers and reservoirs for industrial and agricultural use, among other reasons.

Weak generation at hydroelectric plants requires greater use of thermal plants, which increases electricity-distribution costs and free-market energy prices. The result is an overall price increase, which translates into a negative supply shock to economic activity.

Framework for the assessment of rationing risk in 2015

To assess rationing risks in 2015, we used an approach based on:

• Aggregation of data from the National Integrated System;

• Electrical energy consumption as a function of temperature, spot market prices, economic activity and working days;

• Potential hydroelectric generation (Affluent Natural Energy – ANE) based on calculations of rainfall in the country;

• High usage of thermal plants, for as long as reservoirs remain at seasonally low levels.

The Annex contains details on the i) limitations/possibilities of aggregation, ii) modeling of electrical energy consumption and iii) relationship between our weighted calculations of rainfall and ANE.

Among the model’s exogenous variables, rainfall is the main source of variation for the level of the reservoirs. With other variables kept constant, we estimated scenarios of reservoir levels based on different accumulated rainfall levels during the rainy season (November to April of the following year).

Finally, the decision involving possible rationing would occur at the end of the rainy season (end of April), based on the aggregate level of the reservoirs. Two levels serve here as reference: reservoir levels on April 30, 2014 (42.6%) and the Risk-Aversion Curve[2] of the National Electric System Operator (48.4%). Rounding up the results, we assess that an aggregate level below 50% on April 30, 2015 leads to the continuity of a risk situation, while a level below 40% would possibly lead to some sort of energy-consumption-reduction program.

Results: There are rationing risks if rainfall repeats the pattern of last three years

Given this framework, we forecasted the aggregate level of reservoirs on April 30, 2015 based on rainfall trajectories equal to those observed in each of the last 14 years. Simply put, we replaced rainfall forecasts for the period between now and April 2015 with the rainfall observed in each rainy season (from November of a given year to April of the following year), starting in 2000.

The results are shown in the table below. If rainfall over the coming months repeats the pattern of any of the rainy periods between 2000/01 and 2010/11, we would reach the end of April in a relatively comfortable position, on average 7 pp above the risk-aversion curve. If rainfall, on the other hand, repeats the patterns observed in 2011/12, 2012/13 and 2013/14, the resulting levels imply the need for some kind of action to reduce consumption.

Source: ONS, Itaú

It is important to note that, of all the rainy periods since 2000, only the last three would lead to levels near or below those observed on April 30, 2014 (the minimum historical value for the month, below the risk-aversion curve). This result may suggest changes in the weather pattern, with less rainfall in the SE/CW area. Although we have not tested this hypothesis, this is a possibility that deserves to be closely watched over the coming months.

Over the short term, regardless of the normalization of rainfall ahead, the current situation poses several operational challenges for the National Electric System Operator (ONS). In order for the system to meet the increased consumption of energy in the summer, it is essential that rainfall follows the seasonal pattern during the remainder of the rainy season.

Artur Manoel Passos and Pedro Renault Coelho


1) Aggregation of river basins: does the aggregated analysis make sense in the current scenario?

In our simulations we use the aggregate level of the system’s reservoirs, without looking at individual levels in each region. An important condition for the use of this approach without prejudice to the analysis is the assumption that it is possible to transfer sufficient energy from surplus to deficit regions.

In order to verify the plausibility of this hypothesis, we analyzed the historical series (since 2000) of generation, consumption and energy transfer between regions and found that, in all cases, the power-transmission capacity between a region and the rest of the system is greater than or equal to the largest energy deficit historically observed in that region. More importantly for the current context, the SE/CW region has capacity to receive energy transfers that is about 60% higher than its largest historical deficit (about 4,000 average MW in August 2009). Thus, we conclude that the aggregation of reservoir levels does not adversely affect our analysis.

The map below shows the power-transfer capacity between the system’s aggregate regions. To estimate the maximum volume that may be transferred between two regions, we analyzed the historical series of transfers and assumed that the highest value observed is the maximum transfer capacity[3], in either direction[4]. Following the approach used by the ONS, we aggregated the transfers made in the triple border between the N, SE/CW and NE regions at a single point, which distributes power between the three regions.

Source: ONS, Itaú

It is worth mentioning that the low aggregate level of reservoirs hides a situation of greater concern in the Southeast/Center-West and Northeast river basins, presently at levels below the weighted average (21%), at 15.5% and 13.1%, respectively.

2) Modeling the electricity consumption

As much as the amount of water entering the system is important to forecast the final level of reservoirs in April, the amount of power generated is also relevant to the results.

In order to estimate the demand for energy ahead, we built a model based on energy prices in the free market, industrial production, proportion of working days in the month and temperature. The intuition is simple: more-intense economic activity, a higher number of working days and higher temperatures lead to higher energy consumption, while higher prices lead to reduced demand.

With most of the data for 2014 already available, we estimate consumption growth of 1.9% compared with last year’s average, which is well below the average annual growth of 3.7% in the last ten years. For 2015, the model forecasts a reduction of 0.8% in energy demand (due to still relatively weak activity, high prices and a higher number of holidays on weekdays).

3) Modeling of affluent natural energy

Our modeling of Affluent Natural Energy (ANE) starts from a weighting of rainfall in several river basins, based on the significance to the generation of run-of-river hydroelectric power plants or the accumulation of reservoirs. Gross ANE as a percentage of the Long-Term Average (LTA, energy generated on historical hydrological conditions) is defined for each month in terms of weighted rainfall of that month and the previous two months (also as a percentage of the historical average).

In addition to rain, we incorporated evidence in our models that the official models overestimate the power-generation capacity for a given intensity of rainfall. That is, rainfall levels in line with the historical average have been generating gross ANEs below the historical average. This discrepancy may be associated with difficulties in transmission lines and diversion of water from rivers and reservoirs for industrial and agricultural use, among other reasons.

Our regressions suggest two patterns for these inefficiencies: first, inefficiencies appear to have increased since April 2012. Second, inefficiencies are higher in the period characterized by increased rainfall and greater potential generation.

The chart above shows the actual gross ANE and the gross ANE forecasted by modeling (6-month moving average, to help visualization).

Given the gross ANE, the storable ANE is estimated by a model that considers monthly energy losses in line with the historical pattern.


[1] Aggregate level of the four basins of the National Integrated System.

[2] We aggregate the Risk-Aversion Curve (RAC) of the Southeast/Center-West, South and Northeast basins. The basin of the North region does not have an approved RAC, thus we use the lowest historical level since 2000.

[3] This is a relatively conservative assumption: the lines may never have been used to their maximum capacity.

[4] Except in case of the connection between the S and SE/CW, where loads transferred to the South have already reached 5,880, but the loads transferred to the SE/CW never exceeded 3,304 average MW (we use the latter value as a limit to estimate the transfer capacity to the SE/CW).


< Back