Distributed objects on the distribution side of the network are currently massively assessed so that the tasks carried by the grid distribution and consumption terminal present the characteristics of multiple time scales and large differences in demand，which leads to the complex equilibrium problem of random fluctuation in the computing resource demand.Traditional terminals are limited by the fixed application scenarios and relatively certain resource allocation，and the terminals can only adapt passively through the “quantity for quality ” method，which is unable to solve the persistent contradiction of imbalance between the supply and demand of computing resources in terminals from the root.Therefore，a method of demand forecasting and dynamic equilibrium of computing resources for grid distribution and consumption tasks based on state iteration is proposed.Firstly，the task computing resource demand model is established based on the analysis of scenario attributes and characteristics of grid distribution and consumption tasks.Secondly，the short-term effectiveness forecasting is predicted by the traditional Markov model.Then，the first-order difference equations of the state are used to train the data and track the state fluctuation.The historical state and the forecasted state are used for state iteration to avoid the convergence of long-term forecasting.Finally，a dynamic equilibrium model is established according to the time-scale features of cyclical and non-cyclical tasks.The optimal configuration of the imbalance in computing resource requirements is achieved by shifting staggered peaks and adjusting differentiation.The results have shown that the improved Markov model based on first-order difference and state iteration has the short-term accuracy of the traditional model and the long-term traceability of data fluctuations.The dynamic equilibrium model can effectively reduce the imbalance of resource demand for computing resources and show good ability to cope with resource imbalance deviation.