login name
ITA english

Report RSE 16001480

AMaCha (stochastic Analysis with Markov Chains) 2.0 – Descrizione funzionale


This document describes in detail the tool AMaCha - stochastic Analysis with Markov Chains – version2.0. It collects and amalgamates all the material produced in the years together with the latest changes,in order to provide a comprehensive, complete and final description. The document is divided into fourparts: the theoretical support, the description of the code, a proposal of comparison between results andinput, possible future developments.

In actual power grid, national, European and non-European, there are an increasing number ofgenerators that produce energy from renewable not programmable sources, whose production entered inthe network is not decidable a priori. This, together with the energy policies of the various countries,involves a series of issues to cope with, for example the need to ensure the load cover without knowinghow much energy will be produced by the FER, the reserve management, the expansion of the powernet. To be able to describe more realistically the electricity network and the electricity market, within thesimulation tool is necessary to insert the random features of generation from renewable sources.

AMaCha has exactly this goal: after analysing the historical series of power production from wind orphotovoltaic sources, it captures the statistical parameters and assigns them to the sequence of valuesthat generates in output; these are sets of 8760 hourly values that likely can be considered series ofpower production from wind or solar sources.

The algorithm is divided into two parts.

The first part, AMaCha1, adjusts the input data to remove the missing values; then it deseasonalizesthem with two different methods depending on the source of the data: in the wind case it uses a versionof the algorithm Seasonal Trend decomposition with Loess (STL) to extract the annual seasonality, inthe solar case instead it uses an algorithm only inspired to STL, and it extracts both annual and dailyseasonality; if the user wants and just in case of wind, AMaCha1 identifies the spatial correlation too bythe algorithm of Principal Component Analysis (PCA). The residue is used to build the parameters of adiscrete Markov Process.

The second part, AMaCha2, receives in input the parameters of the Markov model and the informationabout the seasonality and the cross-correlation, and it generates some random series that are comparablewith those provided in input.

Related Links