Under the guidance of Prof. Dr. Sanjay Kumar Singh , IIT(BHU) CSE Dept Under the guidance of PhD Scholar Ananth Padmanathan , IIT(BHU)
- Implemented Support Vector Machine Model to predict the output current for an Electric car, in the dataset containing input currents and phase value of current .
- Using the percentage loss of current , predicted the battery life.
- Achieved 97% accuracy. Exposure : Python ,XgBoost Regressor , Machine Learning , Support Vector Machine , Linear Regression . Link to the code : https://github.com/neshvig10/AI_Project
Link to the dataset used : https://www.kaggle.com/datasets/hankelea/system-identification-of-an-electric-motor
Fields in the dataset :
- id_k : d-current at time k in A (Ampere)
- iq_k : q-current at time k in A (Ampere)
- epsilon_k : rotor angle at time k in rad (Radian)
- n_k : elementary vector (inverter switching command) applied between time k and time (k+1)
- n_1k : elementary vector (inverter switching command) applied between time (k-1) and time k
- id_k1 : d-current at time (k+1) in A (Ampere)
- iq_k1 : q-current at time (k+1) in A (Ampere)
The dataset used here is derived from PADERBORN University , Germany and it has around 10 million rows of data
Contributors :
- S Vigneshwaran (21075073) - myself
- Saikat Mondal (21075075)
The drivetrain of an electric vehicle consists of a battery, an inverter, an electric motor, and a controller. In our experiment, we always keep our battery ideal throughout the experiment. The inverter is a power electronic gadget with switchable semiconductors that converts the electric energy provided by the battery from a two-phase DC voltage to a three-phase AC voltage with varying amplitude and frequency and the generated three currents are transformed to d-q currents using Clarke and Park transformation. This is required to operate the motor at different rotational speeds with a certain torque generated. Since we are considering the battery to be ideal, we are mainly focusing on the interaction between the inverter, motor, and controller.
Figure shows the basic setup of an electric drivetrain of an electric vehicle. Consider, the voltage of the battery be UDC and the inverter consists of three switches (sa, sb, sc) each having a possibility of (+1, -1). Thus, with these 2 different possibilities and 3 switches eight different combinations of switching states are possible generating different voltages at terminals a, b, c as shown in table-1. The instance of all the possibilities of the switches is known as elementary vectors (vn) where n is the index of the vector.
Because of the structure and setup of the inverter the v1 and v8 produce the same voltage so we consider both of them as a single vector. The motor is generally operated in the dq coordinate system. The dq coordinate system is commonly used in motor control because it simplifies the mathematical modeling and control of the motor. In this coordinate system, the stator currents and voltages are represented in two orthogonal axes: the d-axis and the q-axis.
Support Vector Machine (SVM) regression is a popular machine learning algorithm used for regression tasks. In contrast to other regression algorithms, SVM regression tries to find the best line or hyperplane that can accurately predict the target variable.
In SVM regression, the algorithm tries to find a hyperplane that has the maximum margin from the training data. The margin is the distance between the hyperplane and the closest data points from both sides. The goal is to minimize the errors made by the hyperplane while maximizing the margin.
To implement SVM regression, the training data is first transformed into a higher dimensional feature space, where the data is more separable. This transformation is done using a kernel function, which takes the input variables and maps them to a higher dimensional space. The most common kernel functions used in SVM regression are the linear, polynomial, and radial basis function (RBF) kernels.
Once the data is transformed, the algorithm finds the hyperplane that has the maximum margin and separates the data into two regions: positive and negative. The predicted value for a new data point is then based on its position relative to the hyperplane.
SVM regression has several advantages, including its ability to handle high-dimensional data and its ability to work well with non-linear relationships between the input variables and the target variable. However, it can be sensitive to the choice of kernel function and regularization parameters, which may require some tuning to achieve optimal performance.
