This paper introduces a semi-nonparametric (SNP) methodology for the modeling of skewness, leptokurtosis, and high-order moments in electricity markets. We highlight the importance of accurately measuring these features in many contexts (e.g. design of power plants with different technologies, fuel prices, and energy demand) and the need for using flexible non-normal densities. We show that the SNP approach describes the uncertainty in an electricity market, reducing the limitations that normality and parametric density functions impose. Our application covers a wide variety of Colombian electricity variables, including spot price, national energy demand, the climate index ONI, and the series of hydrologic inflows for different rivers. For such variables, we find that the SNP outperforms the normal distribution in terms of accuracy measures based on maximum likelihood estimation. As a result, our methodology has direct applications for risk analysis and portfolio choice related to electricity markets and for implementing policies on electricity markets that improve efficiency and sustainability.