The invention discloses a forecast and monitoring method for peak-hour power usage. The method comprises the steps that classification for power usage data is conducted according to preset temperature threshold, and the classification of the power usage data in a period of time pending to be forecasted is determined; collecting the power usage data of specified classification of all the electric equipment pending to be tested is conducted in a period of time pending to be forecasted; conditional random field model is established, and the collected power usage data which is taken as a training sample is used for estimating the conditional probability distribution function of the power usage; anticipated power usage of the next forecast period of time and whether the high-demand in electricity occurs are forecasted according to the the conditional probability distribution function of the power usage and current power usage; when the peak-hour power usage is forecasted, alarm amount of information is sent to users. Accordingly, good monitoring of the power usage of the power grid can be conducted, so that the network load can be released through regulating the power of appliance in time before the peak-hour power usage occurs.