LCOE – levelized cost of electricity – is an unstandardized model used by governments, companies, consultants and others to make an economic assessment of the cost of generating electricity from a specific source.  LCOE models are often filled with assumptions (what people believe) instead of field data and even when field data is used, the forecasts generated are based on assumptions (beliefs) that typically skew towards the amorphous phrase: grid parity.

The LCOE calculation typically includes a variety of inputs, all of which are susceptible to the modelers experience and or bias for the future direction of the energy technology being modeled.  Typical inputs include: the cost of the installation (components, labor), financing costs, capacity factor, cost of O&M, system production, and so on.  The strongest models use data from field experience (an actual system).  LCOE models based on systems that have been in the field for a number of years are the most robust, however, when the model is used to forecast the future, the bias of the modeler can and sometimes does insinuate itself into the results.

The major flaws of LCOE models are:

  1. The use of assumptions instead of data
  2. Using data that supports the modeler’s goal (that is, assuming too low costs for hardware or O&M) and too high production values
  3. Adjusting model inputs to make a particular point
  4. The assumption of that cost and price are synonymous

Concerning number 1, assumptions about the cost of components or financing (instead of data), using estimates for O&M that are lower than the actual cost of O&M, forecasting down from an assumption about cost that is already too low, assuming system production that is too high and assuming that the comparative energy technology will continue to increase in cost can lead to misleading results.

Concerning number 2, it is not good practice to select the lowest available value and then forecast down from that point.

At this point, there are as many proprietary LCOE models as there are developers, manufacturers and consultants.  The word proprietary does not necessarily confer excellence on the model in question; all it does is state ownership of IP.

When filled with hard data from systems operating in the field LCOE models can be good tools to observe trends overtime. The older an LCOE model gets (provided it is consistently updated with real data from an installation) the better it is as tool for learning. Overtime such a model can become a superior forecasting tool as long as the modeler remains focused on developing an unbiased tool.

As a forecasting tool LCOE models are highly vulnerable to bias. Types of bias include the belief that the cost or price of one energy source will continue decreasing and perhaps accelerate, while the cost or price of the competing energy source will continue increasing.  The behavior of prices for any good or service (including electricity) is variable, that is, prices do not typically increase or decrease in a straight line. LCOE models can be vulnerable to the belief of the modeler that prices for conventional energy will continue increasing while prices for solar generated electricity will continue decreasing.

Assumptions are a form of bias. For example, the assumption that operations and maintenance costs (O&M), which are currently undervalued, will continue decreasing from a point that is already too low, could insinuate bias into an LCOE model, leading to misleading results.

Mistakes in assumptions made around inputs such as the price of modules and other components institute bias.  Assuming that replacement parts will be inexpensive in the future adds bias.  Incorrect assumptions about the running life of a system or its production (output) add bias.  Developing blanket generalizations based on closely held beliefs adds bias.

Unfortunately, LCOE models are highly vulnerable to manipulation to prove whatever point is the goal of the manipulator.  As a sales tool, these models can (and often are) adjusted to make an impression.  As a forecasting tool, these models can (and often are) adjusted to make a point.  Once bias and assumptions in place of data make their way into any model robustness and usefulness suffer and learning is lost. This is truly unfortunate because as deployment of solar continues, the industry needs all the learning it can get. The upfront cost of solar and the time to recover the investment should not be the point. Unlike conventional energy, once the hardware is installed the fuel for a solar installation is free and the system itself typically requires minimal maintenance depending on where the system is installed, size of the system, etc.  Use of solar has a significant role to play in the fight to save our climate and it is far less expensive to install solar now than to fix (after the fact) the damage done to our environment due to climate change.