CN1246704C - Estimation Method of Remaining Capacity of Electric Vehicle Battery - Google Patents

Abstract

The invention relates to a method for estimating the residual capacity of a storage battery of an electric vehicle by applying a neural network model, belonging to the technical field of storage battery capacity estimation and electric vehicles. The invention adopts the distribution of the discharging and regenerating charging capacities of the storage battery to estimate the residual capacity of the storage battery, and the implementation steps are as follows: 1. testing the discharging and regenerative charging current and temperature of the storage battery of the electric vehicle; 2. generating discharge and regeneration charge capacity distribution by accumulating instantaneous current of the storage battery according to a plurality of selected current ranges and corresponding upper and lower limit currents; 3. forming a vector consisting of the battery discharge and regenerative charge capacity distribution and temperature; 4. normalizing this vector to meet the requirements as input to the neural network model; 5. and estimating the residual capacity of the storage battery by applying a neural network model. The method has the main characteristic that the influence of the discharge current mode of the electric vehicle on the available total capacity of the storage battery can be considered, so that the residual capacity of the storage battery of the electric vehicle can be accurately estimated.

Description

Estimation Method of Remaining Capacity of Electric Vehicle Battery

technical field

The invention relates to the technical field of storage battery capacity estimation and electric vehicles.

Background technique

As people pay more and more attention to environmental and energy issues, the research and development of electric vehicle technology has been paid more and more attention. At present, among the many electric vehicle technologies, accurately estimating the remaining capacity of the battery is both a difficult problem and one of the most critical technologies, because it is the core technology of the electric vehicle energy management system.

Generally, the following three methods are used to estimate the remaining capacity of the electric vehicle battery. First, the state of charge estimation method. This method estimates the ratio of the remaining effective substances in the battery to the total effective substances, that is, the so-called state of charge, so the value estimated by this method actually refers to the state of the battery rather than the driving distance of the electric vehicle. The remaining capacity. Although in the case of the same discharge current, the higher the state of charge of the battery, the more the remaining capacity, but there is no clear quantitative relationship between them. For example: under the same state of charge, increasing the temperature and reducing the discharge current will increase the total available capacity of the battery, and its effect is equivalent to increasing the remaining capacity of the battery.

Figure 1 shows the effect of discharge current and temperature on the total available capacity of the battery. The test condition is that the battery is fully charged before each discharge, so the state of charge of the battery before each test is the same and equal to 1. Second, the timekeeping method. This approach is derived from the following most basic equation:

C _r ＝C _a -q(t) (1)

$\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; t</span>}}{<span\; class="notranslate">\; \&Integral;</span>}}{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; dt</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Among them, C _r represents the remaining capacity of the battery, C _a represents the total available capacity of the battery in a certain discharge mode, q(t) represents the discharged capacity of the battery, and I _d (t) represents the discharge current of the battery. Since the total available capacity of electric vehicle batteries varies greatly with the discharge current mode, before estimating the remaining capacity of the battery, the total available capacity of the battery generally takes an appropriate value based on the average discharge current or the reference discharge current. If the value is taken according to the average discharge current, the remaining capacity of the battery can be estimated by the following equation:

C _r ＝C _ave (t)-q(t) (3)

Wherein, C _ave (t) represents the total battery capacity available relative to the average discharge current. Using (3) to estimate the remaining battery capacity can lead to large errors unless the battery discharge current does not change much. Table 1 compares the total usable capacity of the battery under various discharge current patterns (shown in Figure 2) based on U.S. urban driving style, U.S. highway driving style, European standard driving style, and Japanese driving style . It can be seen from Table 1 that although the average discharge current of these modes is approximately equal to 13 amperes, the total available capacity of the batteries varies greatly.

Table 1 Comparison of available total capacity of batteries under various discharge current modes Discharge current mode Average discharge current (ampere) Available Total Capacity (Ampere Hours) Discharge current pattern based on U.S. urban driving style Discharge current pattern based on U.S. highway driving style Discharge current pattern based on European standard driving style Discharge current pattern based on Japanese driving style 13.0813.1113.2113.12 15.9625.0513.0515.43

If the value is taken according to the reference discharge current, then the remaining capacity of the battery can be estimated according to the following equation:

C _r ＝C _ref -α(I _d )q(t) (4)

Wherein, C _ref represents the total available capacity of the battery relative to the reference discharge current, for example: the total available capacity of the battery corresponding to the discharge rate of 3 hours or 5 hours corresponding to the reference discharge current. α(I _d ) represents the conversion factor, which is used to calculate the equivalent capacity discharged when the discharge current is higher or lower than the reference discharge current. In order to obtain this conversion factor (that is, the ratio of the available total capacity relative to the reference discharge current to the available total capacity relative to the discharge current to be converted), the discharge current to be converted is either alone or together with the reference discharge current. Discharge to measure their corresponding usable total capacity. Obviously, using such test results to calculate the conversion factor will ignore the impact of the discharge current mode on the total battery capacity available. Table 2 compares the available total capacity of the battery under a simple two-stage discharge current mode, indicating that the calculation of the conversion coefficient using the aforementioned test method will indeed produce a large error. In addition, another disadvantage of this method is that the impact of temperature on the total available capacity of the battery cannot be considered in the conversion factor, because the relationship between the total available capacity and temperature is nonlinear, as shown in Figure 1.

Table 2 Comparison of the total available capacity of the battery under the two-stage discharge current mode Discharge current mode Available Total Capacity (Ampere Hours) First 8 amps for 3 hours, then 20 amps for 0.38 hours first 20 amps for 0.38 hours, then 8 amps for 3.17 hours first 8 amps for 3.17 hours, then 20 amps for 0.1 hours 31.6633.0027.33

Third, neural network model estimation method. This method uses a three-layer (namely input layer, hidden layer and output layer) neural network model to estimate the remaining capacity of the battery. At present, there are two models: one model has four input units in the input layer, which respectively represent the battery terminal voltage, discharge current, temperature and internal resistance; the other model also has four input units in the input layer, but They represent the battery terminal voltage, discharge current, temperature and discharged capacity respectively. At the output level, both models have only one output unit indicating the remaining capacity of the battery. However, it is found through observation that the input of these two models does not describe the discharge current mode of the battery, so they have the same problem as the second method - the estimation of the remaining capacity of the battery cannot consider the influence of the discharge current mode.

Contents of the invention

The invention provides a new method for estimating the remaining capacity of the accumulator by applying the neuron network model. The input of the method model is the battery temperature and the capacity distribution of electric and regenerative charging, and the output is the state of the total available capacity of the battery. The capacity distribution of discharge and regenerative charge describes the discharge current pattern of the battery, and the state of the total capacity (p _a (t)) can be used to represent the remaining capacity of the battery. Here, the state of the total available capacity of the battery is defined as the percentage of the total available capacity of the battery in a certain discharge current mode, and its mathematical expression is:

p _a (t)=1-q(t)/C _a (5)

It can be seen that, unlike the aforementioned state of charge, the state of the total available capacity of the battery is indeed the remaining capacity of the battery that is related to the driving distance of the electric vehicle.

A method for estimating the remaining capacity of an electric vehicle storage battery using a neuron network model of the present invention, wherein the neuron network model used is composed of three layers:

The first layer is the input layer, which contains several processing units representing the battery discharge and regeneration charging capacity distribution and battery temperature;

The second layer is the hidden layer, which contains several nonlinear processing units;

The third layer is the output layer, which contains a linear processing unit representing the remaining capacity of the battery.

The invention is a general estimation method for the remaining capacity of the storage battery of an electric vehicle. By properly selecting several current ranges and the upper and lower limit currents of each range, the capacity distribution of discharge and regenerative charge can be generated very flexibly to adapt to describe various discharge current modes, so that the present invention can be applied to different storage batteries It can also be applied to the very complex electric vehicle battery discharge current mode. Moreover, by implementing the present invention, the electric vehicle battery remaining capacity indicator can become a commercial product.

Description of drawings

The features of the present invention will be more easily understood with reference to the following detailed description of the embodiments of the present invention and the accompanying drawings. These drawings include:

Figure 1 shows the effect of discharge current and temperature on the total available capacity of the battery;

Figure 2A shows the discharge current pattern based on driving style in urban areas of the United States;

Figure 2B shows the discharge current pattern based on the US highway driving style;

Figure 2C shows the discharge current pattern based on the European standard driving style;

Figure 2D shows the discharge current pattern based on the Japanese driving style;

Figure 3A shows an example of a discharge current pattern based on driving patterns in urban areas of the United States:

Figure 3B gives an example of the discharge current pattern based on the European standard driving style:

Fig. 4 is the neural network model of battery remaining capacity estimation;

Figure 5 presents a comparison of the actual and estimated states of available total capacity using the training sample comparison;

Figure 6A presents a comparison of actual and estimated total usable capacity states in the case of a test sample calibration based on the discharge current pattern of the U.S. urban driving style;

Figure 6B presents a comparison of the state of the actual and estimated available total capacity in the case of verification using test samples in the discharge current mode based on the European standard driving style;

Figure 7 presents the average relative error for all 29 tests.

Detailed ways

The driving distance of an electric vehicle is closely related to the total available capacity of the battery under different discharge current modes. In order to make the established neural network model for battery remaining capacity estimation suitable for use in electric vehicles, the discharge current patterns of different driving modes of electric vehicles were simulated to test the total cycle capacity of the battery.

Figure 2A. 2D shows electric battery discharge current patterns based on U.S. urban driving style, U.S. highway driving style, European standard driving style, and Japanese driving style. In order to obtain comparable test results, the definition of the total usable capacity of the battery is given here, that is, the capacity released when the fully charged battery is discharged to the preset discharge stop voltage under a certain discharge current mode and temperature. Mathematically, it can be expressed as:

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>{<span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; off</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

Wherein, V(t) represents the terminal voltage of the battery, T(t) represents the temperature of the battery, and V _off represents the preset discharge stop voltage of the battery. According to this definition, different combinations of discharge current patterns and temperatures are used to test the battery. The test conditions are: the battery is first fully charged (p _a (t) = 1) and then discharged to a preset discharge stop voltage (p _a (t )=0). In this example, a total of 29 such tests were performed and the results of each test were recorded.

Figures 3A and 3B show two examples of electric vehicle battery discharge current patterns, which are based on the American highway driving style and the European standard driving style respectively, and the battery temperature is 25°C. Then, the state of the available total capacity of the battery can be calculated by using the released capacity of the battery in each test result and the formula (5), so as to obtain the state of the total available capacity of the battery, the discharge current mode and the temperature of the battery expressed by the test data The relationship between. In order to use the neural network model to describe this relationship and achieve the purpose of estimating the remaining capacity of the battery using the state of the total available capacity of the battery, the capacity distribution of discharge and regenerative charging is proposed to describe the discharge current mode. Five current ranges and corresponding upper current limits (shown in Table 3), namely I _i ^l and I _i ^u (i=1, . . . , 5), are used to generate capacity distributions for discharge and regenerative charge.

Table 3 Generation of upper and lower limit currents for discharge and regenerative charge capacity distribution i I _i ^l (A)I _i ^u (A) 10C _N /5 2C _N /5C _N /3 3C _N /3C _N /2 4C _N /2C _N /1 5C _N /1100

Among them, _CN represents the rated capacity of the battery to be tested. On this basis, the present invention proposes a three-layer neuron network model for estimating the remaining capacity of the battery using the state of the total available capacity of the battery, as shown in FIG. 4 . The first layer, the input layer, has 7 neurons, which represent:

· X ₁ (t) - discharged capacity, $I_1^1\&le;I_d\left(t\right)<I_1^u;$

· X ₂ (t) - discharged capacity, $I_2^1\&le;I_d\left(t\right)<I_2^u;$

· X ₃ (t) - discharged capacity, $I_3^1\&le;I_d\left(t\right)<I_3^u;$

· X ₄ (t) - discharged capacity, $I_4^1\&le;I_d\left(t\right)<I_1^u;$

· X ₅ (t) - discharged capacity, $I_5^1\&le;I_d\left(t\right)<I_5^u;$

· X ₆ (t) - regenerative charging capacity,

• X ₇ (t) - battery temperature. Consider the battery discharge and regenerative charge capacity distribution [X ₁ (t)X ₂ (t)X ₃ (t)X ₄ (t)X ₅ (t)X ₆ (t)] and temperature together to form a vector X( t)=[X ₁ (t)X ₂ (t)X ₃ (t)X ₄ (t)X ₅ (t)X ₆ (t)X ₇ (t)], then this neuron network model can be seen As a mapping from the input vector X(t) to the output vector p _a (t), that is, the state of the total available capacity of the battery. Mathematically, it can be expressed as:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

是指蓄电池可用总容量的状态的估计值，n是指隐含层的神经元个数，W_I(i＝1，…，n)是指隐含层和输出层之间的权值，b_l ^o是指输出层的阈值，F(y_i)是Tangent-Sigmoid函数，y_i(i＝1，…，n)是指在隐含层中第i个神经元的输入，它可以表示为：in,

refers to the estimated value of the state of the available total capacity of the storage battery, n refers to the number of neurons in the hidden layer, W _I (i=1,...,n) refers to the weight between the hidden layer and the output layer, b _l ^o refers to the threshold of the output layer, F(y _i ) is the Tangent-Sigmoid function, y _i (i=1,...,n) refers to the input of the i-th neuron in the hidden layer, which can be expressed as :

${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$

Among them, W _ij (i=1,...,n, j=1,...,7) refers to the weight between the input layer and the hidden layer, b _i ^h (i=1,...,n) refers to the hidden layer Contains layer thresholds. Eight neuron network models (n=8-15) were tested to determine the number of neurons in the hidden layer, and finally the neuron network model with the number of neurons in the hidden layer equal to 11 was selected. Because the simulation test found that even if the number of neurons in the hidden layer is greater than 11 under the considered discharge current mode, the estimation accuracy of the model cannot be significantly improved.

The weights between the layers of the neuron network model and the threshold of each neuron are obtained through appropriate learning. The learning process includes a test sample set to improve the universality of the neural network model, so when the error is less than the preset allowable value (set to 10 ^-5 ) or when the error starts to increase on the verification sample set, learning The process stops. Here, the error function E is defined as:

Among them, m represents the number of training samples, p _a (k) represents the kth training sample of the state of the actual available total capacity, An estimate representing the status of the corresponding total available capacity. The learning algorithm uses an improved error backpropagation algorithm, that is, the Levenberg-Marquardt algorithm, which is especially suitable for optimizing the function composed of the sum of squares of nonlinear functions, such as the error function defined in (10). Using this algorithm, E can be expressed as a function of the parameters of the neural network model:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 11</span>}<span\; class="notranslate">\; )</span>$

Therefore, the optimal values of these parameters can be obtained by the following iterative process:

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span>$

in, $A_r\&equiv;{\&dtri;}^2\mathrm{E.}\left(h\right){|}_{h=h_r}$ and $g_r\&equiv;\&dtri;\mathrm{E.}\left(h\right){|}_{h=h_r}$ They are respectively called the Hessian matrix and the gradient matrix after the rth iteration.

In order to effectively use the error backpropagation algorithm, the following equation is usually used to normalize the input of the neural network model:

$x_{\mathrm{jn}}\left(t\right)=\frac{x_j\left(t\right)-x_{j\min }}{x_{j\max }-x_{j\min }}$ (j=1,...,7) (13)

Wherein, X _jn (t) is the normalized value, and X _jmax and X _jmin are the maximum and minimum values of X _j (t) respectively. After normalizing all raw data, these data are put together to form a complete dataset. Then, they are evenly divided into training sample set, validation sample set and test sample set. The training sample set is used for the learning process of the neural network model, and the test sample set is used to confirm the accuracy and effectiveness of the neural network model.

In order to compare the actual and estimated states of total battery capacity available, the concept of an average relative error is used, which is defined as:

$\mathrm{<span\; class="notranslate">\; ARPE</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; ae</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; ac</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; ac</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}\mathrm{<span\; class="notranslate">\; 100</span>}<span\; class="notranslate">\; \%</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 14</span>}<span\; class="notranslate">\; )</span>$

Among them, N refers to the number of training samples or the number of testing samples for each test, p _ae and p _ac respectively refer to the state of the available total capacity estimated by the neural network model and the actual capacity calculated by the test data The state of the total available capacity. Substituting the training samples and the testing samples for each test into (14), their average relative errors can be calculated respectively.

Figure 5 shows the comparison of the actual and estimated states of the available total capacity in the case of using the training samples for comparison. It can be seen from the figure that the estimation of the state of the available total capacity of the battery has a high accuracy and its corresponding average relative error Just 1.27%.

Figures 6A and 6B show the comparison of the actual and estimated available total capacity states in the case of using the test sample comparison. The two discharge current patterns in the figure are based on the US highway driving style and the European standard driving style, corresponding to The average relative errors are 1.22% and 1.28%, respectively, which also have high precision, thus confirming the validity of the neural network model.

Figure 7 presents the average relative error for all 29 tests. In fact, it is worth mentioning that for all 29 tests, the average relative error of the state estimation of the available total capacity of the storage battery is within 2%, thus confirming that the present invention can indeed be used in a very complicated discharge current mode. Accurately estimate the state of the total battery capacity available. Therefore, the method of estimating the remaining capacity of the battery according to the state of the available total capacity of the battery and applying the neural network model can be realized by the following steps. Test electric vehicle battery discharge and regenerative charging current. According to the selected current ranges and the corresponding upper and lower limits of the current (as shown in Table 3), the measured currents are accumulated to generate the capacity distribution of discharge and regenerative charge. This capacity distribution forms a vector together with the battery temperature. Use (13) to normalize the original data of this vector and substitute this normalized vector into (7)-(9) as the input of the neural network model, then the remaining capacity of the battery can be obtained in terms of the state of the total available capacity of the battery.

Although the method for estimating the remaining capacity of the storage battery in the present invention is described in detail through examples above, such description is illustrative rather than limiting. According to the idea and spirit of the present invention, those skilled in the art can make various modifications, and these modifications all belong to the protection scope of the present invention. The protection scope of the present invention is defined by the appended claims.

Claims (4)

1. use the method that neural network model is estimated the electromobile battery residual capacity for one kind, wherein the neural network model that is adopted is formed by three layers:

Ground floor is an input layer, and it comprises several processing units and represents battery discharging and refresh charging capacity to distribute and battery temp respectively;

The second layer is a hidden layer, and it comprises several Nonlinear Processing unit;

The 3rd layer is output layer, and it comprises the residual capacity that accumulator is represented in a linear process unit.

2. the method for estimation of electromobile battery residual capacity according to claim 1, the value of its residual capacity will change between 0 to 1.

3. the method for estimation of electromobile battery residual capacity according to claim 1, its discharge and the distribution of refresh charging capacity are to generate by the accumulator momentary current that adds up according to selected several range of current and corresponding electric current bound.

4. the method for estimation of electromobile battery residual capacity according to claim 1, can realize as follows:

Three layers of neural network model that comprise input layer, non-linear hidden layer and linear output layer are provided;

Measure momentary current and the temperature of accumulator under the electric motor car service condition;

Add up the accumulator momentary current to generate the distribution of battery discharging and refresh charging capacity according to several selected range of current and corresponding electric current bound;

Calculating is corresponding remaining battery capacity under different batteries discharge and refresh charging capacity distribution situation;

Form the training sample set of battery discharging and the distribution of refresh charging capacity, temperature and remaining battery capacity;

Three layers of neural network model that adopt above-mentioned training sample set training to be provided;

Use the neural network model of having trained, estimate the residual capacity of accumulator accumulator under the situation of different temperatures and discharge and the distribution of refresh charging capacity.

Images