Switching from fossil-fuel-powered vehicles to electric vehicles has become an international focus in the pursuit of combatting climate change. Regardless, the adoption of electric vehicles has been slow, in part, due to range anxiety. One solution to mitigating range anxiety is to provide a more accurate state of charge (SOC) and range estimation. SOC estimation of lithium-ion batteries for electric vehicle application is a well-researched topic, yet minimal tools and code exist online for researchers and students alike. To that end, a publicly available Kalman filter-based SOC estimation function is presented. The MATLAB function utilizes a second-order resistor-capacitor equivalent circuit model. It requires the SOC-OCV (open circuit voltage) curve, internal resistance, and equivalent circuit model battery parameters. Users can use an extended Kalman filter (EKF) or adaptive extended Kalman filter (AEKF) algorithm and temperature-dependent battery data. A practical example is illustrated using the LA92 driving cycle of a Turnigy battery at multiple temperatures ranging from -10C to 40C. Current range estimation methods suffer from inaccuracy as factors including temperature, wind, driver behaviour, battery voltage, current, SOC, route/terrain, and much more make it difficult to model accurately. One of the most critical factors in range estimation is the battery. However, most models thus far are represented using equivalent circuit models as they are more widely researched. Another limitation is that any machine learning-based range estimation is typically based on historical driving data that require odometer readings for training. A range estimation algorithm using a machine learning-based voltage estimation model is presented. Specifically, the long short-term memory cell in a recurrent neural network is used for the battery model. The model is trained with two datasets, classic and whole, from the experimental data of four Tesla/Panasonic 2170 battery cells. All network training is completed on SHARCNET, a resource provided by Canada Compute to researchers. The classically trained network achieved an average root mean squared error (RMSE) of 44 mV compared to 34 mV achieved by the network trained on the whole dataset. Based on the whole dataset, all test cases achieve an end range estimation of less than 5 km with an average of 0.29 km. / Thesis / Master of Applied Science (MASc)

This study investigated the implementation of a remaining range estimation algorithm for electrical vehicles, an essential feature to define a vehicle's reliability on the road. The implementation was made on an electrical motorcycle, comparing three models: a dynamic force based model, a power based model and a mass estimation model. The mass model estimated the mass with the help of a RLS algorithm and is a combination of the force based model and the power model. It investigates the possibility to further increase the accuracy of a range estimation algorithm by estimating the total mass of the vehicle over a driving session. On top of these models, two kinds of prediction methods for future consumption were evaluated: the average-past prediction and the home-intention prediction. Both models uses past data to predict the future, but the home-intention prediction is a suggested method to further improve the classic average-past method, where the beginning and end of the vehicle's driving sessions is assumed to be the same location. Tests were executed for the models on an electrical motorcycle provided by the company CAKE. A test equipment were put on the motorcycle, consisting of microprocessors and sensors, used for computation and collection of data. With this equipment, experiments were performed on three test routes with different conditions, comparing the models’ accuracies. The results showed that the Power Model, even with its lower complexity performed best overall, while the Force Model showed mixed results. Depending on the prediction method the Force Model performed either at the top or at the bottom. When the results were analyzed, this behavior seem to be the result of insufficient/faulty hardware which were essential for the average-past prediction to achieve proper results. The Force Model using home-intention prediction consistently performed better, as long as its prediction was correct. The Mass Model was executed offline and were then used to simulate the effect it could have had online. This showed promising result, suggesting improved accuracy if implemented online, but which in this thesis is left as a suggestion of improvement for future work. / Den här studien utforskade implementationen av en algoritm för att estimatera kvarstående räckvidd för ett elektriskt fordon, som är en viktigt funktionalitet för att utvärdera ett fordons pålitlighet på vägen. Implementeringen gjordes på en elektrisk motorcykel på tre modeller: en kraftbaserad-, en effektbaserad- och en massestimeringsmodell. Massestimeringsmodellen estimerar fordonets massa med hjälp av en RLS algoritm och är en kombination av kraft- och effektmodellen. Den utforskar möjligheten att förbättra räckviddsestimeringen ytterligare genom att kunna estimera den totala vikten av fordonet under körningen. På dessa modeller så utvärderades två typer av prediktionsmetoder för att förutspå framtida energiförbrukning: genomsnittliga-datametoden, en metod som använder genomsnittlig data i dåtid, och hem-avsiktsmetoden, en metod som förutspår förarens avsikt att åka hem. Båda modellerna använder gammal data för att förutspå framtiden, men hem-avsiktsmetoden är en föreslagen metod för att ytterligare förbättra den klassiska genomsnittliga-passerade metoden, där början och slutet av körningen antas vara samma position. Test utfördes för modellerna på en elektrisk motorcykel från företaget CAKE. En testutrustning monterades på motorcykeln som består av mikroprocessorer och sensorer och användes för samla och bearbeta data. Med denna utrustning genomfördes experiment på tre olika rutter som hade olika förutsättningar där modellerna träffsäkerhet sedan jämfördes. Resultatet visade på att Effektmodellen, även då den har en lägre nivå av komplexitet, faktiskt presterade generellt sätt bäst, medans Kraftmodellen visade på blandat resultat. Beroende på prediktionsmetod som användes så presterade Kraftmoddel antingen i toppen eller botten. När resultatet analyserades så verkar detta beteende bero på otillräcklig/problematisk hårdvara som var avgörande för den genomsnittliga-passerade metoden. Kraftmodellen tillsammans med hem-avsiktsmetoden ökade prestandan konsekvent så länge som förutsägelsen var korrekt. Massmodellen utfördes offline och detta resultat användes sedan för att simulera massmodellens påverkan på estimering online. Detta visade på lovande resultat och visar på att ifall den metod kördes online så skulle pricksäkerheten kunna ökas, men är något som inte utförs i detta arbete utan lämnas som ett förslag på förbättring för framtida studier.