4.2. Mode Division Based on Logical Thresholds
An ECMS control strategy considering mode switching has been proposed in the literature [
27]. By calculating the equivalent fuel consumption corresponding to each operating point in all selectable modes, a particle swarm algorithm is used to search for the operating point with the lowest equivalent fuel consumption, and the operating mode and torque distribution corresponding to this optimal operating point is used as the output, thus achieving the integration of operating mode switching and torque distribution. However, this method tends to lead to frequent switching of operating modes and is relatively computationally intensive. For this reason, this paper adopts a more mature and simple method, which is based on the engine optimized operating curve to divide the engine torque interval, as shown in
Figure 9, where
Te_max is the maximum engine torque,
Te_opt is the engine economically optimal torque curve, and
Te_low is the minimum torque in the engine efficiency interval.
Based on the optimal operating curve of the engine, the operating mode division rules are developed as shown in
Figure 10. Firstly, the demand torque
Treq is calculated from the tractor speed
v and pedal opening
α, and the maximum torque
Te_max, the optimal torque
Te_opt, and the efficient interval torque threshold
Te_low of the engine at the current speed are obtained from the engine universal characteristic curve. The overcharge and overdischarge of the battery have a great impact on their working performance and service life. Therefore, the battery maintenance target value
SOCobj is set as 0.6, and the highest and lowest fluctuation ranges,
SOCh and
SOCl, are set as 0.8 and 0.4, respectively. Combining the demand torque
Treq and the battery SOC signal, the logic threshold is set to judge the working mode during driving. Finally, the working mode division rules of the parallel hybrid tractor are obtained.
4.3. Fuzzy Adaptive Equivalent Fuel Consumption Minimization Strategy
For the hybrid tractor studied in this paper, the battery cannot be actively charged through the external power grid, so it needs to consume fuel to maintain a stable battery SOC at the end of the working cycle. In a given operating cycle, the power consumed by the battery will be replenished in the subsequent driving and operation process through the driving charging mode or regenerative braking mode, and the final power consumption is actually converted to fuel consumption. For this reason, it is necessary to formulate an appropriate power source torque distribution strategy to reduce fuel consumption as much as possible, while maintaining the balance of battery SOC.
The main idea of the equivalent fuel consumption minimization strategy (ECMS) is to equate the electric consumption of the hybrid vehicle to the fuel consumption by introducing an oil-electric equivalence factor, followed by the equivalent total fuel consumption as a unified indicator for optimal control [
28]. The sum of engine fuel consumption and battery equivalent fuel consumption under each instant is minimized as the objective function, and the optimal solution is found under the condition that the relevant constraints are satisfied. The mathematical model is as follows:
where
is the instantaneous equivalent total fuel consumption,
is the instantaneous engine fuel consumption,
is the instantaneous equivalent fuel consumption of the battery,
Pbat is the battery output power,
s(t) is the oil-electric equivalence factor,
LHVfuel is the low heating value of fuel.
Taking the minimum total fuel consumption of the tractor in the time domain [
t0,
tf] as the control objective, the system objective function
J is defined as:
where the state variable
x(
t) is defined as the battery SOC value and the control variable
u(
t) is the torque distributed between the engine and motor at each moment.
where
is the battery charge or discharge change rate,
ηbat is the battery charge or discharge efficiency,
Ibat is the current, and
Qbat is the rated capacity of the battery.
The objective function should meet the following constraints:
where
Te_max and
Te_min are the upper and lower limits of engine torque, respectively;
ωe_max and
ωe_min are the upper and lower limits of engine speed, respectively;
Tm_max and
Tm_min are the upper and lower limits of motor torque, respectively;
ωm_max and
ωm_min are the upper and lower limits of motor speed, respectively; and
SOCh and
SOCl are the upper and lower limits of battery SOC, respectively; where
SOCl = 0.4 and
SOCh = 0.8.
According to the Pontryagin principle of minimal values, the instantaneous Hamiltonian function is constructed as:
where
H is the Hamiltonian function and
λ is the Lagrangian operator, the optimal sequence of control variables that minimizes the Hamiltonian function can be expressed as:
By taking Equations (4) and (10) into Equation (13), we can obtain:
Combine Equation (8) with Equation (15) to define the oil-electric equivalence factor.
When the set oil-electric equivalent factor is small, the equivalent fuel consumption of the battery is relatively low, and the energy management system tends to use electricity, which can easily lead to overdischarge of battery SOC. On the contrary, when the equivalent factor is large, the energy management system is more inclined to use oil, resulting in the rise of battery SOC deviating from the target value. In order to improve the fuel economy of the tractor and maintain the stability of battery SOC, it is necessary to adjust the oil-electric equivalence factor online. In this paper, the equivalence factor is corrected in real time by introducing a PI controller, and the mathematical model of the equivalence factor can be expressed as:
where
s0 is the initial equivalent factor,
SOCref is the reference SOC,
SOC(t) is the actual SOC,
KP is the proportionality coefficient, and
KI is the integration coefficient.
The setting of the proportional coefficient
KP and the integral coefficient
KI also has an impact on the value of the equivalent factor. In order to improve the robustness of the system, an optimization model of the equivalent factor based on fuzzy control is established, and a fuzzy controller is introduced to modify the proportional and integral coefficient. A two-dimensional fuzzy controller is set. The deviation ∆SOC and the deviation change rate dSOC between the SOC reference value and the actual value are taken as inputs of the fuzzy inference system, and the correction amounts ∆
KP and ∆
KI of the proportional coefficient
KP and the integral coefficient
KI are taken as outputs. The final values of
KP and
KI can be expressed by the following equation:
Define “positive large (PB)”, “positive small (PS)”, “zero” (ZR), “negative small (NS)”, and “negative large (NB)” as five linguistic variables to describe the variation of ∆SOC, dSOC, ∆
KP, and ∆
KI. Using the triangular function as the affiliation function, the theoretical domain of the input quantity is set to [−0.1, 0.1], and the theoretical domains of the output quantities ∆
KP and ∆
KI are set to [–6, 6] and [–1, 1], respectively. The membership functions of the input and output quantities are shown in
Figure 11.
When ∆SOC > 0, the actual battery power is less than the reference battery power, at which time the proportional coefficient
KP and the integral coefficient
KI are proportional to the equivalent factor. Depending on the magnitude of the power deviation ∆SOC and the deviation change rate dSOC,
KP and
KI are corrected to increase the equivalent factor, so that the energy management system tends to directly slow power consumption or charge. Similarly, when ∆SOC < 0, the power consumption trend can be adjusted by modifying the coefficient. Based on this, the fuzzy rules shown in
Table 3 and
Table 4 are developed for the proportional coefficient correction Δ
KP and the integral coefficient correction Δ
KI, respectively, achieving a more accurate adjustment of the equivalent factor.
The fuzzy rule surfaces of the correction quantities Δ
KP and Δ
KI are shown in
Figure 12. The equivalent factor is further adjusted by the correction of the proportional and integral coefficients, so as to construct a fuzzy adaptive model of the equivalence factor based on the battery SOC feedback.