Building Thermal Load Prediction And Control Based On Artificial Neural Network

At present, the most central heating sent hot water from heating centers or exchange stations to the clients directly in many northern cities, then the heating mode is easy to cause uneven heating and energy waste. Although parts of the state began to try the system of thermal measurement and controlled management based on the national related suggestions and requests, it’s slow to take off since it requires large disposable investment and strong heating energy saving consciousness. Therefore, instead of it, building thermal measurement and controlled management device has moderate investment, higher automation and industry reliability. It is an effective mean to achieve the goal of terminal heating energy saving.Since the heating system is a complex dynamic system with strong time lag, large inertia as well as some nonlinear factors including outdoor environment and building maintenance, it is difficult to model the heating system based on mathematical mechanism because of many parameters and big error of the forecast result. By means of the nonlinear approximation capability of BP (Back Propagation) and RBF (Radial Basis Function) artificial neural networks, this thesis model and forecast the heat load according to outdoor dry bulb temperature, illumination, wind speed, time and indoor temperature. Comparison studies between BP and RBF artificial neural networks reveal that the RBF neural network has higher prediction accuracy with 5.3% than BP artificial neural networks. In a word, the RBF artificial neural network is more suitable for forecasting the heat load.Furthermore, it’s required to regulate effectively the thermal heating load based on reasonable forecast.Usually,the valve opening adjustment of heating medium in building heating pipe network perhaps cause uneven heating of the interior of buildings because of increasing of negative loops,even the accident of freeze plug.Aiming at the problem of the uneven heating with continuous adjustment of flow this thesis designed a new intelligent Bang-Bang control. In other words, valves opening need to achieve maximum and certain time when valves open, or minimum when valves close, to ensure the hot water cycles in whole buildings with certain pressure and flow rate when valves open and cut off the heating when valves close. But notably, The heat dissipation process in buildings is still continuous. Valve closing time which should meet the process demands is calculated relatively accurately by comparing the heating in a period of valve fixed open time and forecast heating demand, thus a something new control mode of variable cycle which is fixed valve opening time and variable valve close time is formed. Combining the mode with indoor comfort, the pipe antifreeze, equipment failure prevention and so on, a new controller which has high degree of automation and industry grade reliability is bring out. By the experimental determination, It realized the 10% of average rate of energy saving in the case of stable weather and adequate heat from heat source or heat supply network, having great value of social and economic.