Abstract: The coding and decoding of neural information is the core research content in neuroscience, and it is also very challenging. The traditional neural coding theories have their own limitations, and they are difficult to provide effective theory from the global operation mode of the brain. Since energy is a scalar and has superposition, the theory of energy coding can study the global neural coding problem of the brain function from the prospective of energy characteristics of neuron activities, and has achieved a series of research results. Based on the Wang-Zhang neuron energy calculation model, this paper constructed a multi-level neural network, and we obtained the changes of the energy consumption of the neural network and energy supply of glucose in the blood by numerical simulation. The calculation results showed that the time of peak supply of glucose in the blood is delayed about 5.6 seconds compared to the time when the neural activity of the network reaches its peak, which reproduced hemodynamic phenomena in functional nuclear magnetic resonance (fMRI) from a quantitative perspective: after a five to seven seconds delay in the activation of a brain region, the change in cerebral blood flow increases dramatically. The simulation results showed that negative energy mechanism, which was previously reported by our group using Wang-Zhang neuronal model, played a central role in controlling the hemodynamics of the brain. Also, it predicted the neural coupling mechanism between the energy metabolism and blood flow changes in the brain under the condition of stimulation, which was determined by imbalance and mismatch between the positive and negative energy during the spike of neuronal action potentials. The research results in this paper provided a new research direction for further exploring the physiological mechanism of hemodynamic phenomena in the future, and gave a new perspective and research method in the modeling and calculation of neural networks.