Weak current College】 【VAV system of intelligent control.

First, an overview

In 1984, the United States Connecticut (Connecticut), Hart Ford City (Hargford), will conduct an old financial Tower, known as the "city of palaces" (CityPalaceBuilding), which is recognized as the world's first intelligent building. Domestic in the early ' 90s built some intelligent buildings, such as Beijing's China World Trade Center and Beijing-Guangzhou Center, Shanghai Jin Jiang Tower Hotel, Shanghai municipal building in the market. Intelligent building is a building, it creates an environment of building occupants of maximum work efficiency, while at the same time with a minimum of one time consumption guarantee effective resource management. Intelligent building intelligence, so their workplace environment design principle is "people-oriented", all around for the user to create a comfortable environment, improve user productivity. Among them, the HVAC industry played an important role.

Intelligent building, room number, large load changes within the zone; at the same time, since air conditioning energy consumption of the whole building, generally 40% energy saving issues have become increasingly prominent. Air conditioning in various ways, VAV system has its own advantages:

1. as a result of the air conditioning system for most of the time running under partial load, the air volume of energy consumption reduction brought fans and terminal equipment for reducing the energy consumption of the heater again;

2. to achieve the flexible control of the local region;

3. use of system diversity makes the initial cost of the central system;

4. Similarly, because of the diversity of available system, future expansion costs reduction;

5. the system is balanced (Self-balancing), etc. Therefore, foreign intelligence building air-conditioning systems the use of VAV system, or with CAV air-conditioning system, the combination of air conditioning systems FCU.

Although the VAV system has these advantages, but it is the most complex of the control. Currently, VAV system control basically consist of multiple loop of PID control. Changes in the system model parameter is not large, PID control to good effect. However, VAV system is an interference with large, highly non-linear, non-deterministic system, this is due to:

1. weather and air conditioning area personnel activities vary greatly, the formation of the system is taking a significant interference;

2. air conditioning process is highly non-linear; the operation of the actuator is linear;

3. the coupling between the control circuit is completely decoupled, is not possible;

4. over time, the device will replace the ageing and the resulting changes in the system parameters;

5. in many systems, the system of the digital model is difficult to establish.

Therefore, the effect of PID control is very bad. In the field of HVAC control, many new control methods, such as MacArthur and Grald method using adaptive control, Dexter and self-regulating Haves use predictive controller. However, modern control theory and the theory of large systems used in VAV system due to their analysis, synthesis and design are strict and accurate digital model, so the same encounter such a problem. And intelligent control theory for charged object and its environment and tasks of the uncertainty raised, VAV system control areas should have broad prospects. At present, intelligent control theory there are three directions, i.e., neural network, fuzzy control and expert system. Their main VAV system for diagnosing exceptions, forecast energy consumption. For VAV system control, artificial neural network control and fuzzy control of study has been started, and expert system because of its knowledge base, design is difficult, it is difficult to apply, the following, for fuzzy control, neural network control of VAV system for little explored.

Second, fuzzy control

Fuzzy control is based on the rules of intelligent control to mathematics as a basis. The basic structure of the system as shown in Figure 1. Controller of four basic components, fuzz interfaces, knowledge base, decision-making logic unit, go to fuzzy interface.

In the past few years, some to the HVAC system by fuzzy logic control, S.Huang and r.m.nelson PFC (PID and fuzzy control) introduced into the HVAC control and second-order for unit transfer function for the simulation. The two authors and describes a kind of fuzzy control rules of adjustment methods, applied to the control of HVAC systems, used to control a heat exchanger of pneumatic valve will return air temperature as input, the experimental results show that the control programmes is much better than the PID control. These control thought also can be used for the control of VAV system. RobertN.Lea and fuzzy controller EdgarDohmann application compressor and blower, input is set point temperature, relative humidity and to be adjusted in six regions. Soetal. launched a control based on fuzzy logic controller, four apartment state of a parameter (for wind pressure on the throttle, the indoor temperature, indoor humidity and air temperature) by adjusting the five Executive command (for wind fan speeds, air supply ventilation, cooling water flow rate terms, heater power and humidifier for temperature ratio) to control.

Fuzzy controller for control of the VAV system of parameters as input, the output is the VAV systems of command execution. A simple VAV system FLCD fig.3 as shown in Figure 2 State needs according to the different control scheme to select.

General of fuzzy controller, there are many shortcomings, such as system increases performance, overshoot bigTo regulate the time, or even production line oscillation, a antijamming ability, steady-state error, have these defects of the main reasons is the general structure of fuzzy controller for too simple, but also in the design process has many subjective factors, but once the fuzzy rules determine doesn't change, and we hope that the fuzzy controller to dynamically adjust itself, with the ability to learn, to achieve the intended control quality, fuzzy control, an important research direction is adaptive fuzzy control, there are two major, model reference Adaptive fuzzy control (Figure 3a) and adaptive fuzzy control (Figure 3b). Fuzzy controller adjustable parts: control rules, membership function and normalization factor.

Various intelligent control of all available between knot complement each other, such as fuzzy neural network control, is the use of neural network learning ability, and the combination of expert system, or with the classic control methods, modern control method of combination, these are the control of current topics in the field of research, the need to control who studied and applied as soon as possible to the VAV system.

Third, the neural network control

Neural networks in HVAC control areas of research and applications more. Curtissetal. first apply hot water coil of predictive control of hot water valve, the work for future research provides a good foundation. It is based on valve position, cooling load, air temperature, air velocity, water temperature, hot water flow, as well as their historical data as input to predict the cooling load. The author has introduced a neural network on HVC central unit AHU for energy management, completed on the optimization of a local ring run control, the neural network control of same as VAV system control provides a number of ideas.

Neural network control usually take identification system, identification is kin neural network training process on neuronal connections between the value of the right to modify, and then can be used for superior control. Because of the neural network training requires a significant amount of time, real-time identification very difficult to achieve, so generally the online identification, real-time control. A VAV systems to the central unit of VAV control system with simple block diagram as shown in Figure 5:

ANN identification is composed by three layers of neurons in the input layer, hidden and output layer, training samples the input should be a system in the moment of action t-1 (then heating clockwise power output, humidifier output for wind fan speed, cooling water valve control for air throttle angle, etc.) and characterization system characteristics of State parameter (indoor temperature, indoor humidity, air supply pressure, the throttle for air temperature, wind velocity, wind velocity, wind moisture content, etc.), the output should be a system of State parameter t at all times. As a result of VAV system inertia is large compared with ANN training speed, the unit features change very slowly, so no real control problems that once concluded, the identification process began. Controller tuning into the moment t perform commands in the next moment arrives a hope of control. Control of aim with minimal delay and power to obtain the desired room temperature and humidity, so that the system of static and dynamic characteristics to meet your performance requirements. Therefore, the controller of the objective function should include two parts, one is the set point (such as the indoor temperature, indoor humidity, pressure for fans, or after the PMV indicators, etc.), part of the system of total energy consumption (such as the power consumption of fan, heater power, etc.), specific parameters to select depending on the design of the air-conditioning system and control programmes to determine, and can adjust the scaling factor between them to focus on any one.

At present, the neural network training is mainly made of BP, BP has two distinct disadvantages: convergence is slow, easily into local Optima. And simulated annealing algorithm, the simplex method and the genetic algorithm (GA) to solve these problems, in particular genetic algorithm in recent years of increasing attention. But GA can only find in a short period of time to close the global optimum solution of near-optimal solutions, because the process of seeking GA is random, with a certain degree of probability of blindness and even reached the vicinity of the optimal solution is also likely to turn a blind eye "," "walk on by" BP and genetic algorithm can overcome these problems.

IV. conclusions

Abroad in intelligent buildings air conditioning system most used VAV system, this article only on fuzzy and neural networks in VAV system of application for some of the VAV system is usually carried out energy-saving air conditioning process multiple load partition, and then use multiple loop of PID control. The VAV system is an interference with large, highly non-linear, non-deterministic system, and each single loop coupling between strongly, the use of PID in static and dynamic characteristics often cannot satisfy the performance requirements. Domestic and foreign scholars trying to use smart control, such as fuzzy logic, neural networks and a variety of hybrid control methods in place of the PID control, but are in the application of the single circuit, single-loop optimal performance on reach, in global terms not reach optimal performance, while at the same time as the circuit of the regulation have offset the effect of temperature and humidity, resulting in increased energy consumption. Some scholars use some optimization methods for global optimization control or on the large system globally coordinated control theory, to determine each single loop of a given value, but they are also based on the system model comparison to determine, on the basis of the system difficult to modeling, use these control methods are very difficult, but not reach the optimal global performance. Use intelligent control method in the global control of the system, eliminating the need for system modeling, resolves the previous control loop due to coupling many control performance problems.

The development of the theory of intelligent control is the control of VAV system provides more advanced methods, but also researchers presented great challenges. Need more control over the personnel into the area of VAV control, and air conditioning professionals communicate together, design a better VAV control system.

>