Philip Taylor explains how distributed intelligent load controllers can ensure efficient and stable operation of small stand-alone hydro projects
Renewable energy systems, such as hydro powered networks, experience real time variations in energy input and load. Control methods are required to maintain stability and achieve maximum penetration. Existing load controllers use a threshold system based on the generator droop characteristic to make switching decisions about the load. This method has the disadvantage that load controllers require configuration for each system involving a difficult process of finding suitable sizes of loads, and setting the thresholds at the optimum points on the droop characteristic. Also, certain loads at the low end of the threshold range experience a much better service than others.
Phase angle switching systems are also used. The main problem in this case is that the voltage waveform is severely distorted by the switching actions. This is a matter of concern as EMC compliance is unlikely. These load control systems are usually centralised controllers and therefore a lot of power is often wasted in dumping.
To overcome these problems a new load control system was developed — a distributed intelligent load controller which uses a low cost micro-controller based device. These connect individual single-phase loads to the supply. If one load controller fails or is disconnected, the rest of the system can continue to function adequately. The load control system is highly decentralised, which improves control accuracy and allows for very small changes in load.
Software embedded in the micro-controllers measures the frequency and voltage, and this information is used to connect or disconnect the loads. Systems with capacitor excited induction generators will rely on voltage measure-ments only. The distributed nature of the control means that real consumer loads can be manipulated resulting in little or no energy being wasted due to dumping. The new load control system uses zero voltage switching, causing negligible distortion to the voltage waveform. The controllers share a single target frequency and therefore do not require a complex set up procedure.
An artificial intelligence technique known as fuzzy logic was used to develop this intelligent controller which can provide real time advice regarding load switching. Fuzzy logic is an extension of classical set theory, which uses sets with blurred boundaries such as ‘warm’ or ‘cold’ rather than the rigid ‘yes’ or ‘no’ decisions traditionally made by computers. Fuzzy logic is particularly appropriate for systems where the sources of information are interpreted qualitatively, inexactly or uncertainly. The problem in question exhibits each of these features. Interpretation of the power system status at any instant is qualitative, all measurement algorithms are inexact and future generation and load is uncertain.
The load controllers are designed to work together to achieve the following control objectives:
•Stabilise frequency and voltage.
•Maximise penetration from renewable energy sources.
•Reduce short term peak loads.
•Reduce wasteful energy dumping.
•Avoid blackouts or brown outs.
•Share energy equitably between loads/consumers.
Load control software
Measurement algorithms are embedded in the micro-controller and provide information regarding frequency, voltage and rate of change of frequency and voltage. The frequency measurement algorithm uses a level crossing technique rather than zero crossing to calculate the frequency. This method was used to avoid spikes and distortion on the voltage waveform affecting the accuracy of the measurement.
The measurement algorithms provide the inputs to the fuzzy control system. The fuzzy control system then processes these inputs and uses its rule base and various fuzzy operators to recommend whether to switch the load on, switch off or take no action.
The application is a multiple input single output (MISO) system and uses state evaluation control rules. The process state is evaluated and a fuzzy control action is computed at time t as a function of the inputs and the control rules.
The rule base was designed using knowledge of power system dynamics. A trade off between system complexity and a simple low memory solution was required so that the final controller software could be stored in the limited program memory of the micro-controller.
The fuzzy controller produces a crisp value that advises on the switching action to be taken. The fuzzy controller also indicates the degree of truth in the control action advised. There is no overall supervisory controller for the load controllers and they do not communicate with each other. Therefore, in order to ensure that the load controllers operate individually and that the energy is shared equitably among the loads, code was developed to provide a random element to the load controllers. The code also ensures that the overall load connected is proportional to the degree of truth indicated by the fuzzy controller.
The optimum design
Extensive experimentation was carried out, using computer models and also hydro and wind powered systems, in an attempt to find the optimum controller design. It became evident that there was no single best design. Therefore for best performance the controller was programmed to adapt in real time. The controller receives feedback on its current performance, permitting on-line modification of the controller design in the pursuit of optimum performance.
Basic alterations can be made to the controller causing a bias towards particular desired behaviour. For example, if the load controllers were not connecting sufficient load fast enough, the frequency would begin to rise beyond acceptable limits. The controllers would learn from this and adapt so that they would be more likely to add load earlier given the same circumstances again.
If the performance of the load controllers is satisfactory no learning is undertaken and the controller design remains constant. The algorithms with the learning capability outperform those with a static controller design.
Load control hardware
The hardware design is deliberately simple with few components to reduce cost. UK company Econnect uses load control products suitable for 230V or 110V, 50Hz or 60Hz and with different mounting and packaging options.
The load controller uses a low cost single chip PIC micro-controller. These controllers are RISC architecture ICs which are capable of faster operation than the traditional von Neumann architecture and also promote simple yet efficient code.
Initially the load controllers used an electromagnetic relay to carry out the load switching. The latest generation of load controllers uses a TRIAC to enable faster switching rates. Zero crossing optical drivers are used to control the TRIACs to ensure that switching actions are taken at voltage zero crossings. This slightly compromises the speed of response of the control system but minimises the distortion of the voltage waveform, and is essential as EMC regulations become increasingly stringent. It can be seen (bottom right), that the distortion of the waveform is negligible in comparison to that caused by the phase angle load controller (centre right). The bottom figure was taken from a micro-hydro scheme in Polmood Scotland whilst under fuzzy load control. The figure in the centre is from the same system under centralised phase angle control.
The operating experience of the Polmood micro hydro power system in the UK can demonstrate the advantages of fuzzy logic load controllers. This system (rated at 18kW, single phase, 230V) has an existing centralised phase angle load controller which heavily distorts the voltage waveform. The fuzzy load controllers were connected to the micro hydro site whose system generates up to 14kW. Fifteen 1kW heating loads were connected to the fuzzy load controllers and switched to maintain a stable frequency. It can be seen (top figure) that the frequency control was within +0.2Hz. This is well within the limits for an autonomous system. The load controller was controlling the system at 52Hz initially, after that the system was under the sole influence of the fuzzy load controllers.