APPARATUS FOR DETERMINING REACTANT PURITY This invention relates to an apparatus configured to determine the purity of a reactant. In particular, it relates to an apparatus for determining the purity of a fuel, such as hydrogen, and/or an oxidant, such as air. It also relates to a filling station or stationary power device including the apparatus and a method of determining reactant purity. The use of hydrogen as a fuel for the generation of electrical power in fuel cells is becoming of increasing importance. Purity of the hydrogen supply is important for optimal electrical power generation and for maintaining fuel cells using that hydrogen in optimal condition. Currently, hydrogen used in fuel cell systems is often synthesized through the steam reforming of natural methane gas. Even where best quality practices are used, a number of contaminants may be present in the hydrogen fuel which are harmful to fuel cell operation. Although the harm is usually reversible, in the worst cases a high degree of contamination may be present including some compounds which may cause irreversible harm to the fuel cell. According to a first aspect of the invention, we provide an apparatus configured to determine reactant purity comprising; a first fuel cell configured to generate electrical current from the electrochemical reaction between two reactants, having a first reactant inlet configured to receive a test reactant comprising one of the two reactants from a first reactant source; a second fuel cell configured to generate electrical current from the electrochemical reaction between the two reactants, having a second reactant inlet configured to receive the test reactant from a second reactant source; a controller configured to apply an electrical load to each fuel cell and determine an electrical output difference, ODt, between an electrical output of the first fuel cell and an electrical output of the second fuel cell, and determine a difference between a predicted output difference and the determined electrical output difference, ODt, the predicted output difference determined based on a historical output difference and a historical rate of change in said output difference determined at an earlier time, said controller configured to provide a purity output indicative of the test reactant purity at least based on the difference between the predicted and determined output difference. This is advantageous as it has been found that determining a prediction of what the output difference is going to be at a later time based on historical measurements and then, at the later time, determining the output difference and making a comparison provides an effective way of determining or monitoring reactant purity. The historical values may comprise a previously determined difference or rate of change or a historical average of the difference or rate of change. The controller may be configured to determine at least three indicators at time t using two of the same indicators previously determined at an earlier time t-1, the indicators comprising a Deltat indicator representative of the difference between a predicted output difference and the determined output difference, a smoothed level indicator SLt, and a rate of change indicator ROCt, wherein; Deltat = ODt - (SL + At x ROC ) SLt = (SLt-i + At x ROCM ) + x Deltat ROCt = ROCM + 2 x Deltat and At comprises the time difference between time t and t-1, and C and a2 comprise predetermined values, wherein the purity output is determined using said indicators. These indicators have been found to be effective at identifying deterioration of reactant quality over time, while being computationally efficient. The purity output may be determined using the electrical output difference, ODt. The first reactant source may provide a reference reactant of the test reactant, the reference reactant having a known purity and the second reactant source provides a fuel of an unknown purity. The first reactant source may comprise a purification device, the purification device configured to purify part of the test reactant supplied from the second reactant source. Thus, the apparatus is configured to determine the electrical output of two fuel cells that are substantially identical other than the fuel that is supplied to them. It may be assumed that the reference reactant is pure while the test reactant has an unknown purity which the present apparatus may determine relative to the reference reactant. Optionally the test reactant comprises a fuel. The fuel may comprise hydrogen. Optionally the test reactant comprises an oxidant, such as atmospheric air. The reactant may be for supply to a fuel cell power source, such as a fuel cell powered vehicle or a stationary power device. The output difference may be determined based on an average of a plurality of sampled electrical output values from the first fuel cell and the second fuel cell. Thus, a plurality of output values may be averaged and the difference determined or a plurality of output differences determined and then averaged. The average may comprise a modal, mean or median average, measure of central tendency or any other average. The controller may be configured to determine if the rate of change indicator exceeds a predetermined threshold range and, if so, provide a warning of changing reactant purity. The controller may be configured to determine if the smoothed level indicator exceeds a predetermined threshold range and, if so, provide an indication that the reactant purity is unacceptable. The controller may be configured to determine if the Delta indicator exceeds a predetermined threshold range and, if so, provide an indication that the reactant purity is unacceptable. Thus, the controller may only raise an alarm or warning if a predetermined threshold is exceeded. Alternatively, it may provide a plurality of warnings based on a plurality of thresholds. The apparatus may include a third fuel cell configured to generate electrical current from the electrochemical reaction between the two reactants, wherein the test reactant comprises a first test reactant and the other of the two reactants comprises a second test reactant; the first fuel cell configured to receive the first test reactant from the first reactant source and the second test reactant from a fourth reactant source; the second fuel cell configured to receive the first test reactant from the second reactant source and the second test reactant from the fourth reactant source; the third fuel cell configured to receive the first test reactant from the first reactant source and the second test reactant from a third reactant source, the controller configured to determine an electrical output difference, ODt, between an electrical output of the first and second fuel cell; first and third fuel cell; and second and third fuel cell, the controller configured to give an indication of the first test reactant purity and the second test reactant purity at least based on a difference between a predicted output difference and the determined output difference, ODt, the predicted output difference determined based on a historical output difference and a historical rate of change in said output difference for each of the output differences. The first test reactant may comprise a fuel and the second test reactant comprises air; the first reactant source comprising a pure fuel source; the second reactant source comprising a fuel source of unknown purity; the third reactant source comprising a pure air source or an air source of unknown purity and the fourth reactant source comprising the other of the pure air source and air source of unknown purity. The controller may be configured to output an indication of the performance difference between the third fuel cell and the second fuel cell. Thus, the output difference between second and third fuel cells may provide an indication of fuel cell health. The first fuel cell may comprise a plurality of series-connected fuel cells in a stack and/or in which the second fuel cell comprises a plurality of series-connected fuel cells in a stack. According to a second aspect of the invention we provide a method for determining an indication of reactant purity comprising; measuring an electrical output of a first fuel cell having a load applied thereto and configured to generate electrical current from the electrochemical reaction between two reactants, one of the two reactants comprising a test reactant supplied from a first reactant source to the first fuel cell; measuring an electrical output of a second fuel cell having a load applied thereto and configured to generate electrical current from the electrochemical reaction between the two reactants, the test reactant supplied to the second fuel cell supplied from a second reactant source; determining an electrical output difference, ODt, between an electrical output of the first fuel cell and an electrical output of the second fuel cell, providing an indication of the test reactant purity at least based on a difference between a predicted output difference and the determined output difference, the predicted output difference determined based on a historical output difference and a historical rate of change in said output difference. The step of providing an indication may comprise; determining at least three indicators at time t using two of the same indicators previously determined at time t-1, the indicators comprising a Delta indicator representative of the difference between a predicted output difference and the determined output difference, a smoothed level indicator SU, and a rate of change indicator ROCt, wherein; Deltat = ODt - (SLM + At X ROCt- ) SU = (SLt-i + At x ROCn) + x Deltat ROCt = ROCt-i + a2 x Deltat and At comprises the time difference between time t and t-1, and c