An electric bus energy consumption prediction method combined with markov working condition prediction

Abstract

The application provides a kind of electric bus energy consumption prediction method combined with Markov working condition prediction, in which, in the establishment of electric bus vehicle physical model and energy consumption prediction model, various factors influencing vehicle energy consumption and speed distribution law of vehicle at station and different road sections are fully considered, so that the vehicle speed curve generated based on Markov Monte Carlo method is as close to the real working condition as possible, thereby significantly improving the accuracy of the prediction result of bus energy consumption.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of new energy vehicle energy consumption measurement, and particularly relates to a method for predicting energy consumption of an electric bus in combination with Markov working condition prediction. BACKGROUND

[0002] Due to the complex and changeable working conditions faced by electric buses in daily operation, it is still difficult to measure the driving energy consumption of the electric buses. In some existing technologies, historical sample data is used to predict the energy consumption of electric buses, but the future working conditions are not considered, which makes this method not practical. Some existing technologies, such as Chinese patent application CN108806021A, establish a speed-acceleration state transition matrix for different types of road segments, and then use Markov prediction to obtain the acceleration value of the target road segment, and further obtain the time-speed curve of each target road segment, which can improve the accuracy of electric vehicle energy consumption prediction to a certain extent. However, the generation of this speed curve is only based on the Markov Monte Carlo method, without considering that the parking speed is reduced to 0 at a specific location (such as a bus station), and the speed difference of the bus at different time periods and road segments is also rarely considered, resulting in a relatively large gap between the generated speed distribution and the actual speed distribution. SUMMARY

[0003] Therefore, in view of the technical problems in the field, the application provides a method for predicting energy consumption of an electric bus in combination with Markov working condition prediction, which specifically includes the following steps:

[0004] Step 1: Collect original vehicle driving data including running time, speed, temperature inside and outside the vehicle, air conditioning state, battery SOC, current and voltage of the battery, etc. during the historical operation of the electric bus, and perform data cleaning processing; divide the original vehicle driving data after data cleaning into multiple trip segments according to different trips of vehicle operation, and divide the inter-station segments according to the positions of each station of the bus; extract the feature parameters related to the energy consumption of the electric bus from each segment, and calculate the energy consumption of each trip and the motor energy consumption of the electric bus;

[0005] Step 2: Construct a physical model of the electric bus based on vehicle longitudinal dynamics, which is used to reflect the relationship between motor energy consumption and vehicle physical model parameters, vehicle speed and acceleration; use the speed, acceleration and motor energy consumption data corresponding to the bus trip segment to perform least squares identification to obtain the vehicle physical model parameters;

[0006] Step three, statistics of the speed distribution in each segment and calculation of the state transition matrix of the electric bus speed at different times and road segments, generation of a first half of a driving segment by Markov Monte Carlo method, selection of a suitable segment from the corresponding deceleration segment between stations, and splicing of the first half of the driving segment to form a complete driving segment and obtain the speed curve corresponding to the complete driving segment;

[0007] Step four, using the physical model identified in step two to calculate the motor energy consumption characteristics of different trips based on the speed curve obtained in step three, and inputting the characteristic parameters related to the energy consumption of the electric bus obtained in step one and the trip energy consumption into the LightGBM model for training; after training, based on the real-time vehicle driving data and the corresponding speed curve, the LightGBM model is used to predict the energy consumption of the electric bus in the corresponding trip.

[0008] Further, the data cleaning process in step one includes filling the missing values in the original vehicle driving data by linear interpolation; when extracting the characteristic parameters from the divided segments, the 3σ rule is used to remove outliers; the energy consumption of each trip of the electric bus and the motor energy consumption are calculated by the following formulas respectively:

[0009] EC trip =∑U b (t)×I b (t)×Δt

[0010] EC motor =∑U m (t)×I m (t)×Δt

[0011] Wherein, EC trip represents the energy consumption of the bus in a trip, U b (t), I b (t) and U m (t), I m (t) represent the voltage and current of the battery and motor at time t, and Δt represents the time interval of data acquisition.

[0012] Further, in step two, the influence of air resistance, rolling resistance, slope resistance and inertial force on the vehicle is considered when building the physical model of the electric bus, and the following relationship between motor energy consumption and vehicle physical model parameters, vehicle speed and acceleration is established:

[0013]

[0014] In the formula, ρ represents the air density, C dis the air resistance coefficient, A is the vehicle forward area, M is the vehicle weight, g is the gravity acceleration, f is the rolling resistance coefficient, θ is the road slope, δ is the rotational inertia coefficient, v(t) and a(t) are the speed and acceleration at time t, η b is the battery discharge efficiency, η r is the energy recovery rate of the electric bus, k1, k2, k3,..., k6 are physical model parameters to be fitted.

[0015] Further, in step three, when calculating the state transition matrix, the speed is first discretized into different speed intervals and a plurality of state values are assigned to each interval; the process of generating the speed curve includes:

[0016] The target driving distance L, the state transition matrix TPM and the deceleration segment set Decs are input into the Markov Monte Carlo algorithm, and the following steps are sequentially executed:

[0017] ① Initialize the algorithm, set the cumulative driving distance = 0, the current speed = 0, the current state = 0, and the speed list = [0];

[0018] ② Randomly generate a number s in the range of (0, 1];

[0019] ③ Obtain the next state i according to the state transition matrix and s;

[0020] ④ Randomly generate a speed v in the corresponding speed range according to the state i;

[0021] ⑤ Add the speed v to the end of the speed list;

[0022] ⑥ Update the cumulative driving distance = cumulative driving distance + v x Δt;

[0023] ⑦ Judge whether the following conditions are met: |initial speed of deceleration segment - v| < 5 km / h, and |driving distance of deceleration segment - remaining driving distance| < 5 m; if yes, place the deceleration segment at the end of the generated segment and exit the loop; if not, return to step ③.

[0024] The electric bus energy consumption prediction method provided by the present application in combination with Markov working condition prediction, in establishing the physical model and energy consumption prediction model of the electric bus, fully considers various factors affecting the energy consumption of the vehicle and the speed distribution law of the vehicle at the station and on different road sections, so that the vehicle speed curve generated based on the Markov Monte Carlo method is as close to the real working condition as possible, thereby significantly improving the accuracy of the prediction result of the energy consumption of the bus. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The route framework diagram of the method provided by the present application. DETAILED DESCRIPTION

[0026] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] The energy consumption prediction method for electric buses that combines Markov operating condition prediction provided by this invention, such as... Figure 1 As shown, the specific steps include:

[0028] Step 1: Collect raw vehicle driving data from the historical operation of electric buses, including running time, speed, interior and exterior temperatures, air conditioning status, battery SOC, battery current and voltage, etc., and perform data cleaning processing. Divide the raw vehicle driving data after data cleaning into multiple trip segments according to different trips of the vehicle, and divide them into inter-station segments according to the location of each bus stop. Extract characteristic parameters related to the energy consumption of electric buses from each segment. In addition to the parameters in the raw vehicle driving data, parameters such as vehicle mileage, departure time in a day, departure time in a week, and air conditioning on or off status can also be selected. Calculate the energy consumption of electric buses for each trip and the energy consumption of the motor.

[0029] Step 2: Construct a physical model of the electric bus based on the longitudinal dynamics of the vehicle to reflect the relationship between motor energy consumption and the parameters of the vehicle physical model, vehicle speed and acceleration; use the speed, acceleration and motor energy consumption data corresponding to bus trip segments to perform the least squares method to identify the parameters of the vehicle physical model.

[0030] Step 3: Statistically analyze the speed distribution in each segment and calculate the state transition matrix of the electric bus speed at different times and road segments. Generate the first half of a driving segment using the Markov Monte Carlo method. Then, select a suitable segment from the corresponding deceleration segments in the segments between stations and splice it together with the first half to form a complete driving segment and obtain the speed curve corresponding to the complete driving segment.

[0031] Step 4: Using the physical model identified in Step 2, calculate the motor energy consumption characteristics of different trips based on the speed curve obtained in Step 3. Input the characteristic parameters related to the energy consumption of electric buses and the trip energy consumption obtained in Step 1 into the LightGBM model for training. After training, based on the real-time acquired vehicle driving data and corresponding speed curves, the LightGBM model predicts the energy consumption of the electric bus for the corresponding trip.

[0032] In the preferred embodiment of the present application, the data cleaning process in step one includes filling the missing values in the original vehicle driving data by using linear interpolation method; when extracting the feature parameters from the divided segments, the 3σ rule is used to eliminate outliers; the energy consumption of each trip of the electric bus and the motor energy consumption are respectively calculated by the following formulas:

[0033] EC trip =∑U b (t)×I b (t)×Δt

[0034] EC motor =∑U m (t)×I m (t)×Δt

[0035] Wherein, EC trip represents the energy consumption of the bus in a trip, U b (t), I b (t) and U m (t), I m (t) represent the voltage and current of the battery and the motor at t time, and Δt represents the time interval of data acquisition.

[0036] Since the electric bus is mainly affected by air resistance, rolling resistance, slope resistance and inertia force during acceleration or deceleration in operation. According to Newton's second law of motion, the traction force of the electric bus can be calculated according to the following equation:

[0037]

[0038] Therefore, in the preferred embodiment of the present application, the above influencing factors are considered when constructing the physical model of the electric bus in step two, and the following relationship between the motor energy consumption and the vehicle physical model parameters, vehicle speed and acceleration is established:

[0039]

[0040] In the formula, ρ represents the air density, C d is the air resistance coefficient, A is the vehicle frontal area, M is the vehicle weight, g is the gravitational acceleration, f is the rolling resistance coefficient, θ is the road slope, δ is the rotational inertia coefficient, v(t) and a(t) are the speed and acceleration at t time, η b is the battery discharge efficiency, η r is the energy recovery rate of the electric bus, and k1, k2, k3,..., k6 are physical model parameters to be fitted.

[0041] In the preferred embodiment of the present application, in step three, when calculating the state transition matrix, the maximum speed of the electric bus is considered to be limited to 55km / h, the speed is first discretized into 12 speed intervals: 0km / h, (0, 5]km / h, (5, 10]km / h, (10, 15]km / h, …, (45, 50]km / h and (50, 55]km / h); and a state value of 0 to 11 is assigned to each interval; the process of generating the speed curve includes:

[0042] The target driving distance L, the state transition matrix TPM and the deceleration segment set Decs are input into the Markov Monte Carlo algorithm, and the following steps are executed in sequence:

[0043] ① Initialize the algorithm, set the cumulative driving distance = 0, the current speed = 0, the current state = 0, and the speed list = [0];

[0044] ② Randomly generate a number s in the range of (0, 1];

[0045] ③ Obtain the next state i according to the state transition matrix and s;

[0046] ④ Randomly generate a speed v in the corresponding speed range according to the state i;

[0047] ⑤ Add the speed v to the end of the speed list;

[0048] ⑥ Update the cumulative driving distance = cumulative driving distance + v x Δt;

[0049] ⑦ Determine whether the following conditions are met: |deceleration segment initial speed-v|<5km / h, and |deceleration segment driving distance-remaining driving distance|<5m; if yes, place the deceleration segment at the end of the generated segment, and exit the loop; if not, return to step ③.

[0050] It should be understood that the size of the serial number of each step in the embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0051] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for electric bus energy consumption prediction combined with Markov state prediction, characterized in that: Specifically comprising the following steps; Step one, collect the original vehicle driving data including running time, speed, temperature inside and outside the vehicle, air conditioning state, battery SOC, current and voltage of the battery in the historical operation of the electric bus, and perform data cleaning processing; the original vehicle driving data after data cleaning is divided into multiple trip segments according to different trips of vehicle operation, and inter-station segments are divided according to the positions of each station of the bus; the characteristic parameters related to the energy consumption of the electric bus are extracted from each segment, and the energy consumption and motor energy consumption of each trip of the electric bus are calculated; Step two, a physical model of the electric bus is constructed based on vehicle longitudinal dynamics, which is used to reflect the relationship between motor energy consumption and vehicle physical model parameters, vehicle speed and acceleration; the vehicle physical model parameters are obtained by using the least square method to identify the speed, acceleration and motor energy consumption data corresponding to the bus trip segment; Step three, the speed distribution in each segment is counted and the state transition matrix of the electric bus speed at different times and road sections is calculated, a first half of a driving segment is generated by Markov Monte Carlo method, and a suitable segment is selected from the corresponding deceleration segment in each inter-station segment, which is spliced with the first half to form a complete driving segment and obtain the speed curve corresponding to the complete driving segment; Step four, the physical model identified in step two is used to calculate the motor energy consumption characteristics of different trips based on the speed curve obtained in step three, and the characteristic parameters related to the energy consumption of the electric bus and the trip energy consumption obtained in step one are input into the LightGBM model for training; after training, the energy consumption of the corresponding trip of the electric bus is predicted by the LightGBM model based on the real-time acquired vehicle driving data and the corresponding speed curve.

2. The method of claim 1, wherein: The data cleaning process in step one includes filling the missing values in the original vehicle driving data by linear interpolation; when extracting the characteristic parameters from the divided segments, the 3σ rule is used to remove outliers; the trip energy consumption and motor energy consumption of the electric bus are calculated by the following formulas respectively: EC trip =∑U b (t)×I b (t)×Δt EC motor =∑U m (t)×I m (t)×Δt where EC trip represents the energy consumption of a bus in a trip, U b (t), I b (t) and U m (t), I m (t) respectively represent the voltage and current of the battery and motor at time t, and Δt represents the time interval of data collection.

3. The method of claim 1, wherein: In step two, when constructing the physical model of the electric bus, the influence of air resistance, rolling resistance, slope resistance and inertia force on the vehicle is considered, and the following relationship between motor energy consumption and vehicle physical model parameters, vehicle speed and acceleration is established: where p represents the air density, C d is the air resistance coefficient, A is the vehicle frontal area, M is the vehicle weight, g is the gravitational acceleration, f is the rolling resistance coefficient, θ is the road slope, δ is the rotational inertia coefficient, v(t) and a(t) are the speed and acceleration at time t, η b is the battery discharge efficiency, η r is the electric bus energy recovery rate, k1, k2, k3,..., k6 are physical model parameters to be fitted.

4. The method of claim 1, wherein: In step three, when calculating the state transition matrix, the speed is first discretized into different speed intervals and multiple state values are assigned to each interval; The process of generating the speed curve includes: Input the target driving distance L, the state transition matrix TPM and the deceleration segment set Decs into the Markov Monte Carlo algorithm, and execute the following steps in turn: ① Initialize the algorithm, set cumulative driving distance = 0, current speed = 0, current state = 0, and speed list = [0]; ② Randomly generate a number s in the range of (0, 1]; ③ Get the next state i according to the state transition matrix and s; ④ Randomly generate a speed v in the corresponding speed range according to the state i; ⑤ Add the speed v to the end of the speed list; ⑥ Update the cumulative driving distance = cumulative driving distance + v x Δt; ⑦Judge whether the following condition is satisfied: |initial speed of deceleration segment-v|<5km / h, and |driving distance of deceleration segment-remaining driving distance|<5m; if yes, put the deceleration segment at the end of the generated segments, and exit the loop; if not, return to step ③.

Images