fitting, good self-learning and fault tolerance. However, the
traditional BP network requires a lot of training samples. Its
convergence time is long, and the initial weight or the threshold
value is difficult to be determined. Beyond that, it is easy to fall
into local minimum. These defects have affected the prediction
accuracy of the model. Thus, the researchers use different
intelligent algorithms to improve the BP neural network, but
this kind of research is relatively small. Researcher established
the energy consumption of the RBF prediction algorithm [3],
combined the grey system requires less training data and neural
network has the advantages of self-learning and self-organizing.
It reduces the relative error of the experiment, but the precision
and stability of the model is not ideal. Researcher used genetic
algorithm to optimize the connection weights of neural network
[4], so the prediction model of building energy consumption
and indoor thermal comfort of GA-BP network is established.
The results show that the model has higher prediction accuracy,
but genetic algorithm has some defects such as complex
encoding and variation,and strict requirements for initial
parameters. There are deficiencies in the above literature. The
combination of BP neural network and other optimization
algorithm for the daily energy consumption prediction research
is less.