INTRO: From autumn 2005, Tsukuba Science City will be directly linked to Akihabara station in the centre of Tokyo by a 58 km radial line now being built through the Greater Tokyo area. The opportunity to get even closer to ’pure’ regenerative braking is being seized, reserving friction brakes for parking and emergencies
BYLINE: Prof Satoru Sone
Dept of Electrical EngineeringKogakuin University
JAPAN HAS at least twice as many EMU cars in service as any other country by a wide margin - around 50000 ranging from the Shinkansen high speed fleet to metros and trams. There are even plans to use them for express container trains, and a prototype is already on test (RG 5.03 p265). Although friction brakes actuated by compressed air survive as the basic form of brake, more than 90% of the kinetic energy of electric trains in Japan is dissipated by electric braking, using the traction motors to generate power.
Dynamic (rheostatic) braking became popular from the early 1950s, and field chopper control born in the late 1960s made it possible to use regenerative braking for stopping at stations. Up to that time, regenerative braking had been used exclusively for retarding trains descending steep gradients. The biggest breakthrough came in 1984 when 4500V gate turn-off thyristors became available. From that year onwards, almost all EMUs manufactured in Japan were fitted with AC motors and regenerative braking.
The ’pure’ electric brake
In the early days, electric brakes were unable to stop a train because the retarding torque of the motor falls to zero as the speed drops away. This required the substitution of friction braking at low speed.
But the characteristics of the two brakes are quite different. Thanks to power electronics, an electric brake can precisely control the braking force. Friction brakes cannot, because the coefficient of friction varies by speed and surface condition. The response time of an electric brake is fast, but pneumatic brakes react slowly. Thus fading out electric braking and simultaneously blending in the air brake is not easy to synchronise, leading to longitudinal surges that affect riding comfort and sometimes result in wheel flats.
It was in 1997 that I put forward a proposal to eliminate friction brakes from normal service braking by generating braking force electrically right down to a complete standstill. Most of the braking would be regenerative, but as this faded the traction motors would be powered in reverse, all as a single smooth operation. This was basically not a new idea because the technique was already being used to stop lifts smoothly and accurately.
In a strict sense, ’pure electric braking’ means total regeneration. However, in normal operation a relatively high braking force is often applied at high speed. This is achieved electrically in the dynamic braking mode, thanks to the over-voltage capability of traction motors when operating as generators feeding resistor banks independent of the catenary voltage. Indeed, the highest traction motor voltage at braking is typically 2·5 times that of powering.
This is very difficult to do economically when regenerating. In practice, the catenary voltage is typically only 15% higher in regenerative mode than it is in powering mode. This means that regenerative brake force at high speed would be less than half that of dynamic braking. Using the word ’pure’ in a broader sense, we tend to call any system that is effective down to standstill a pure electric brake, and these are rapidly gaining popularity.
In 1997, Shin-Keisei Electric Railway Co began to develop this kind of pure electric brake with Mitsubishi Electric Corp and the University of Tokyo, where I was working at that time. Tests were very successful, and pure electric brakes have been used in revenue service since 1998.
However, there remains a technical difficulty with this broader form of ’pure’ electric brake. Despite the fact that electric braking to a complete stop is widely used for lifts, a problem emerged in the rail application. Direction and speed measurements at very low speeds are difficult, because of the inferior rotary encoder used for trains in Japan. There are only 60 pulses per revolution compared to several hundred for more general use.
It was found in due course that extrapolation of train speed from the 60 pulse detector, or even speed sensorless control using a reference model monitoring the input voltage and current of the traction motor could solve this problem. Today, all traction manufacturers in Japan are ready to supply a pure electric brake.
To avoid damaging on-board equipment by over-voltage, regenerated power has to be limited or ’squeezed’ to prevent the catenary voltage rising too high. In the early years of regenerative trains, squeezing characteristics were over-protective, as shown in Fig 1a. Up to line voltage V1 the regenerated line current remains constant at Imax, but above this limit the current is ramped down to zero at V2 and reversed to prevent the line voltage exceeding Vmax.
The higher both V1 and V2 are, the more power can be regenerated within the limits of what on-board equipment can withstand. If all trains on the same line are similarly controlled by power electronics, the characteristics of V1 = V2 = Vmax as shown in Fig 1b can be used because ’powering off’ and ’regenerating off’ is slow enough and simultaneous. Thus regeneration can continue at Imax right up to the fringe of Vmax.
If a rheostatically controlled train stops powering, typically in two steps, the voltage trajectory of a regenerating train goes much higher than Vmax as shown in Fig 1c. Similarly, the power balance between a combination of a long train of light powering and a short train of strong regeneration changes by powering off. The jerk rate (rate of change of deceleration) can be outside ride comfort limits and/or the voltage locus considerably exceeds Vmax.
One effective countermeasure against an insufficiently receptive line is blending rheostatic and regenerative braking. The simplest way to achieve this is to add a chopper-controlled rheostat in parallel with the filter capacitor in a regenerative brake circuit, so that this rheostat absorbs power in excess of receptivity. A rather complicated circuit with sophisticated control can thus enlarge braking power beyond that of pure regenerative braking.
Although it has not been tried in Japan, some attempts have been made to use flywheel energy storage either on the train or at the substations, for example on the London Underground (RG 4.01 p241).
Power supply changes
Fig 2a shows the typical characteristics of a DC substation. As the power drawn by trains varies substantially and rapidly, as much as three times the rated current (usually for 1h) can be drawn for short periods. Nominal voltage regulation of typically 6% to 8% means that the voltage will drop by 20% at the substation during these peak current conditions. The no-load voltage is therefore set high - a typical value is 1620V for a nominal 1500V system.
On the other hand, electrical equipment on trains can generally withstand 1800V. This means that Vmax in Fig 1a should be set at 1800V, but it is very difficult to transmit heavy regenerative current to a remote powering train under this voltage limit.
Using phase-controlled thyristors to regulate the output voltage at the substation is one way of improving this situation. One example with the characteristics shown in Fig 2b proved effective, but mixed usage with adjacent non-controlled substations reduces effectiveness and brings difficulties in balancing the load between substations.
To prevent regenerative current produced by trains from being squeezed because the voltage is too high it is neccessary to keep the circuit resistance between substations and trains as small as possible. However, the installation of return conductors parallel with the running rail is not practical because of track circuit issues, and total resistance per km can hardly be smaller than 30m