The invention discloses a short-term load coordination and forecasting method based on branch-circuit measurement data. According to a power network structure, load nodes are subjected to regional division; each load node predicts power of upper and lower layer voltage class equipment respectively; the prediction result of the upper layer and lower layer in each load node is coordinated; the prediction results of the load nodes are coordinated; whether whole-network prediction and coordination are completed or not is judged; and the whole-network load prediction result obtained after coordination is output. The method is based on circuit measurement data and uses a vector machine support method for prediction, firstly, the prediction results of load on the upper and lower layers in the nodes are coordinated, secondly, the electricity transmission branch prediction result is coordinated through interaction of key information between the nodes, the prediction accuracy is improved, meanwhile the coordination of the prediction results of the node load on the upper and lower layers can be achieved, and technical support can be provided for the operation and scheduling of an electric power system.