What is chilled water load forecasting and what is its basic idea?

1, redundant energy consumption: the energy consumption difference between the non-predictive optimization process and the ideal optimization process (the optimization time step is infinitely reduced).

2. Scheme objective: to reduce redundant energy consumption as much as possible through predictive control without shortening the time step.

3. Implementation steps:

Firstly, the cooling load forecasting model, the supervised optimal control model and the air conditioning system model are established.

At the beginning of a time step, execute: 1) predict the cooling load of the next time step through the cooling load prediction model (it is actually necessary to predict the outdoor wet bulb temperature, and to simplify the problem, it is assumed that the outdoor wet bulb temperature remains unchanged at this time step); 2) Transfer the prediction results of cooling load and wet bulb temperature to the optimization algorithm to search for the optimal control strategy (the optimization goal of the optimization algorithm is to minimize the sum of current power and power at the end of time step, and the system power is calculated by the air conditioning system model); 3) Implement control strategy.

Repeat step 2.

4. Model algorithm:

Air conditioning system model: a "grey box" model consisting of cooling tower (ANN model), cooling water pump (mathematical formula) and water chiller (ANN model). Training ANN model based on field test data and data generated by TRNSYS simulation; The building model is not established, and the side characteristics of the building are included in the load forecasting model and the cold machine model;

Dynamic cooling load prediction: According to historical load, building information and environmental weather, predict the cooling load in the next 65,438+0 hours. Unbiased random walk, ANN and integration model are tried (two sub-models are established according to historical cooling load and system state respectively and then integrated), among which the integration model has the best effect, R^2=0.9605 (the data comes from a real office building in Hong Kong, which seems to take only four working days, and only the sum of cooling load is used.

Optimal control: Genetic algorithm is adopted, and the variables of optimal control are chilled water supply temperature, cooling water supply temperature and cooling water flow (excluding chilled water flow).

5. Case study

MATLAB is used to replace the actual air conditioning system to analyze and verify the scheme.

When the cooling load rises rapidly (8: 00 am ~ 9: 00 am), the energy-saving effect of predictive control is obvious;

Cooling water flow is more sensitive to air conditioning load. Compared with non-predictive control, the cooling water flow rate given by predictive control changes obviously, and the chilled water supply temperature and cooling water supply temperature are almost equal.

After calculation, prediction and optimization, redundant energy consumption can be reduced by more than 80%. However, in this case, the cumulative redundant energy consumption for four days is only 14 kWh, which is a bit outrageous. The author explains that the possible reason is that the cooling water flow can only be adjusted between 95.5%-65,438+000% due to the system parameters of the cold station, and the cooling water flow is the main influencing factor of the system energy consumption, so the redundant energy consumption is very limited. If the adjustment range of cooling water flow is limited,

think