The invention relates to an electricity market early-warning technology, in particular to an establishment method for an electricity market supply-demand early-warning model based on dynamic clustering. The establishment method comprises the following steps: 1.1 selecting electricity market pre-trade early-warning indexes which comprise the following indexes: (1) a declared space share; (2) a price rise rate; (3) a declared high price rate; (4) a declared high price success rate; 1.2 according to the ability indexes comprising the declared space share, the price rise rate, the declared high price rate and the declared high price success rate which are determined in step 1.1 and capable of reflecting a market bidding status and bidding strength and of controlling the market price, adopting k-means clustering, wherein an initial center is automatically determined by SPSS (Statistical Product and Service Solutions); 1.3 obtaining final class-center difference and class-center excursion degree between classes by utilizing the k-means algorithm so as to establish the electricity market supply-demand early-warning model.