Lithium battery equalization method based on improved fuzzy control

Abstract

The invention discloses a lithium battery equalization method based on improved fuzzy control, which aims at solving the problems of low single variable control precision and poor fixed universe adaptability in the existing equalization method, and comprises the following steps: firstly, dividing SOC into different intervals through battery OCV-SOC characteristic analysis, and preferentially carrying out voltage equalization in an extreme SOC interval to avoid overcharge and overdischarge; sOC equalization is preferentially carried out in the middle SOC interval so as to improve the energy transfer efficiency. A voltage and SOC dual fuzzy controller is designed, a variable universe fuzzy controller is introduced to dynamically adjust an input variable universe, the universe is expanded at the initial stage of equalization to accelerate response, and the universe is shrunk at the later stage of equalization to improve precision. And further designing a self-adaptive fuzzy controller, taking the SOC mean value and the equalizing current as input, and dynamically outputting a weight coefficient of a double-fuzzy controller, thereby realizing self-adaptive switching of the equalizing strategy under different SOC intervals. According to the invention, the equalization precision and efficiency are effectively improved, and the robustness and safety of the system are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of battery balancing technology, and in particular relates to a lithium battery balancing method based on improved fuzzy control. Background Technology

[0002] With the rapid development of the new energy vehicle market in recent years, lithium-ion batteries, due to their high energy density and long lifespan, have been widely used as energy storage units in electric vehicles. However, due to differences in manufacturing processes, material properties, and operating environments, individual lithium-ion battery cells exhibit a certain degree of inconsistency in voltage and capacity. If operated under this unbalanced state for extended periods, the battery pack will face problems such as overcharging and over-discharging, capacity decay, reduced operating efficiency, and decreased safety, thereby shortening the lifespan of the lithium-ion battery and potentially leading to safety hazards. To address the inconsistency between individual battery cells, research on lithium-ion battery balancing technology has become a hot topic recently.

[0003] Currently, common equalization strategies include the mean difference method and fuzzy control. The former becomes extremely cumbersome in designing complex control flows and cannot dynamically adjust the current, resulting in a long equalization time. The latter can reduce equalization energy loss and time while achieving dynamic duty cycle adjustment, but its fixed domain of discourse leads to a decrease in control accuracy as the equalization error decreases. Furthermore, the selection of a single equalization variable introduces a certain degree of equalization error, further reducing control accuracy. Therefore, a lithium battery equalization method based on improved fuzzy control is needed to address these issues. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a lithium battery equalization method based on improved fuzzy control. It realizes variable universe of discourse on the traditional fixed universe of discourse fuzzy control and establishes a multivariate adaptive fuzzy controller that can perform staged hybrid control according to the battery operating state, which enhances the flexibility of the equalizer and improves the equalization efficiency.

[0005] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is as follows: A lithium battery equalization method based on improved fuzzy control includes the following steps: Obtain the OCV-SOC curve of the battery pack, and divide the SOC into a low SOC range, a middle SOC range and a high SOC range according to the OCV-SOC curve. In the low SOC range and the high SOC range, the OCV changes drastically with the SOC, while in the middle SOC range, the OCV changes slowly with the SOC. Design voltage fuzzy controller and SOC fuzzy controller: The voltage fuzzy controller uses the average voltage of adjacent battery cells or battery packs. and voltage difference The input variable is the duty cycle that adjusts the magnitude of the balancing current, and the output variable is the duty cycle that adjusts the magnitude of the balancing current. The SOC fuzzy controller uses the average SOC of adjacent battery cells or battery packs. Difference between SOC and The input variable is the duty cycle that adjusts the magnitude of the balancing current, and the output variable is the duty cycle that adjusts the magnitude of the balancing current. Design a variable universe of discourse fuzzy controller for dynamically adjusting voltage difference. Difference between SOC and The universe of discourse, wherein the variable universe of discourse fuzzy controller uses voltage difference or SOC difference The voltage difference is adjusted by taking its dispersion as the input variable and the scaling factor as the output variable. or SOC difference The initial domain; Design an adaptive fuzzy controller based on the average SOC of adjacent battery cells or battery packs. Using the equalization current as input variables, and the weighting coefficients of the SOC fuzzy controller output duty cycle and the voltage fuzzy controller output duty cycle as output variables, the desired output duty cycle is obtained by weighted fusion of the SOC fuzzy controller output duty cycle and the voltage fuzzy controller output duty cycle, which is used to dynamically adjust the magnitude of the equalization current.

[0006] Preferably, the low SOC range is 0% ≤ SOC ≤ 15%, the high SOC range is 95% ≤ SOC ≤ 100%, and the intermediate SOC range is 15% ≤ SOC ≤ 95%.

[0007] Preferably, the input and output variables of the voltage fuzzy controller and the SOC fuzzy controller are divided into 5 fuzzy sets, which include SS, S, M, B, and BB, representing minimal, small, moderate, large, and maximal, respectively.

[0008] Preferably, the SOC mean value is the input variable of the SOC fuzzy controller. The universe of discourse is [0,1], and the SOC difference is... The universe of discourse for the output duty cycle is [0, 0.5], and the universe of discourse for the output duty cycle is [0, 0.5]; the average voltage value is the input variable of the voltage fuzzy controller. The universe of discourse is [2.4, 3.6], and the voltage difference is... The universe of discourse is [0, 0.5], and the universe of discourse for the output duty cycle is [0, 0.5].

[0009] Preferably, the variable universe of discourse fuzzy controller uses the SOC difference. Or its dispersion as input, or voltage difference Or its discreteness is the input.

[0010] Preferably, the dispersion is the standard deviation of the SOC difference or voltage difference, calculated using the following formula: ; In the formula, This represents the SOC value of monomer i; is the average SOC value of the battery pack; N is the total number of individual cells.

[0011] Preferably, the formula for weighted fusion is: ; in, To determine the desired output duty cycle, the output of the SOC fuzzy controller is... Weighting coefficients for control For the output of the voltage fuzzy controller Weighting coefficients for control , and The magnitudes are all distributed between 0 and 1. After weighted fusion, the duty cycle of the desired output is obtained. .

[0012] Preferably, the fuzzy rules of the SOC fuzzy controller and the voltage fuzzy controller are as follows: when the mean of the input variable is extremely small or extremely large, the output duty cycle is small; when the mean of the input variable is in the middle value, the output duty cycle is appropriately increased according to the magnitude of the difference between the input variables.

[0013] Preferably, the fuzzy rule of the variable universe fuzzy controller is: when the difference between the input variables is extremely small and the dispersion is extremely small, the output scaling factor is a minimum value; when the difference between the input variables is extremely large and the dispersion is extremely large, the output scaling factor is a maximum value.

[0014] Preferably, the fuzzy rule of the adaptive fuzzy controller is: when the mean of the input variable is extremely small or extremely large, the output weight coefficient... Increase Decrease; when the mean of the input variable is at the median, the output weight coefficient is reduced. Increase Decrease.

[0015] The beneficial effects of this invention are as follows: 1. This invention employs multi-variable balancing of voltage and SOC to reduce the errors of traditional single-variable balancing. Voltage balancing is prioritized in the 0%≤SOC≤15% and 95%≤SOC≤100% ranges to avoid overcharging and over-discharging of cells at extreme SOC values, thus improving the safety of cell balancing within these extreme ranges. In the intermediate range of 15%≤SOC≤95%, SOC variable balancing is prioritized, and energy transfer efficiency is improved by increasing the duty cycle while ensuring safety.

[0016] 2. This invention addresses the nonlinear and time-varying parameter characteristics of battery systems by constructing a highly robust system through multi-layer fuzzy control. The essential advantage of fuzzy control lies in its independence from precise mathematical models. By constructing fuzzy sets to process fuzzy information, it effectively reduces interference such as SOC estimation errors and voltage measurement noise, thereby improving equalization accuracy to a certain extent.

[0017] 3. This invention introduces the principle of variable universe of discourse to adjust the initial universe of discourse of the variable, thereby reducing the equilibrium error caused by traditional fixed universe of discourse fuzzy control. In the early stage of equilibrium, when the difference is large and the dispersion is high, the scaling factor appropriately expands the universe of discourse, accelerating the convergence speed of large errors; in the later stage of equilibrium, when the difference is small and the dispersion is low, the universe of discourse shrinks, making the membership function more refined in its division of small differences and the output more accurate.

[0018] 4. This invention achieves phased equalization control of SOC and voltage through an adaptive fuzzy controller, dynamically outputting the corresponding weight coefficients of the SOC fuzzy controller and the voltage fuzzy controller to meet the control requirements of the battery under different states, prioritizing voltage equalization in the extreme value range and prioritizing SOC equalization in the intermediate range. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the OCV-SOC curve of the present invention; Figure 2 This is a schematic diagram of the equalization control process in an embodiment of the present invention; Figure 3 In the embodiments of the present invention, The membership function diagram of the input and output variables of the SOC fuzzy controller is shown. Figure 4 In the embodiments of the present invention, For input quantity The membership function diagram of the input and output variables of the SOC fuzzy controller is used to represent the output quantity. Figure 5 In the embodiments of the present invention, The membership function diagram of the input and output variables of the voltage fuzzy controller is shown. Figure 6 In the embodiments of the present invention, The membership function diagram of the input and output variables of the voltage fuzzy controller is shown. Figure 7 In the embodiments of the present invention, Membership function graph for input variables; Figure 8 In the embodiments of the present invention, Membership function graph of the output quantity; Figure 9 In the embodiments of the present invention, The membership function graph of the input and output variables of the adaptive fuzzy controller for the input quantity; Figure 10 In the embodiments of the present invention, The membership function graph of the input and output variables of the adaptive fuzzy controller for the input quantity; Figure 11 In the embodiments of the present invention, , The membership function graph of the input and output variables of the adaptive fuzzy controller for the output quantity; Figure 12 This is an embodiment of the present invention. , , 3D relationship diagram of input and output of adaptive fuzzy controller; Figure 13 This is an embodiment of the present invention. , , Three-dimensional relationship diagram of input and output of adaptive fuzzy controller. Detailed Implementation

[0020] Example 1: A lithium battery equalization method based on improved fuzzy control includes the following steps: Obtain the OCV-SOC curve of the battery pack, and divide the SOC into a low SOC range, a middle SOC range and a high SOC range according to the OCV-SOC curve. In the low SOC range and the high SOC range, the OCV changes drastically with the SOC, while in the middle SOC range, the OCV changes slowly with the SOC. Design voltage fuzzy controller and SOC fuzzy controller: The voltage fuzzy controller uses the average voltage of adjacent battery cells or battery packs. and voltage difference The input variable is the duty cycle that adjusts the magnitude of the balancing current, and the output variable is the duty cycle that adjusts the magnitude of the balancing current. The SOC fuzzy controller uses the average SOC of adjacent battery cells or battery packs. Difference between SOC and The input variable is the duty cycle that adjusts the magnitude of the balancing current, and the output variable is the duty cycle that adjusts the magnitude of the balancing current. Design a variable universe of discourse fuzzy controller for dynamically adjusting voltage difference. Difference between SOC and The universe of discourse, wherein the variable universe of discourse fuzzy controller uses voltage difference or SOC difference The voltage difference is adjusted by taking its dispersion as the input variable and the scaling factor as the output variable. or SOC difference The initial domain; Design an adaptive fuzzy controller based on the average SOC of adjacent battery cells or battery packs. Using the equalization current as input variables, and the weighting coefficients of the SOC fuzzy controller output duty cycle and the voltage fuzzy controller output duty cycle as output variables, the desired output duty cycle is obtained by weighted fusion of the SOC fuzzy controller output duty cycle and the voltage fuzzy controller output duty cycle, which is used to dynamically adjust the magnitude of the equalization current.

[0021] Preferably, the low SOC range is 0% ≤ SOC ≤ 15%, the high SOC range is 95% ≤ SOC ≤ 100%, and the intermediate SOC range is 15% ≤ SOC ≤ 95%.

[0022] Preferably, the input and output variables of the voltage fuzzy controller and the SOC fuzzy controller are divided into 5 fuzzy sets, which include SS, S, M, B, and BB, representing minimal, small, moderate, large, and maximal, respectively.

[0023] Preferably, the SOC mean value is the input variable of the SOC fuzzy controller. The universe of discourse is [0,1], and the SOC difference is... The universe of discourse for the output duty cycle is [0, 0.5], and the universe of discourse for the output duty cycle is [0, 0.5]; the average voltage value is the input variable of the voltage fuzzy controller. The universe of discourse is [2.4, 3.6], and the voltage difference is... The universe of discourse is [0, 0.5], and the universe of discourse for the output duty cycle is [0, 0.5].

[0024] Preferably, the variable universe of discourse fuzzy controller uses the SOC difference. Or its dispersion as input, or voltage difference Or its discreteness is the input.

[0025] Preferably, the dispersion is the standard deviation of the SOC difference or voltage difference, calculated using the following formula: ; In the formula, This represents the SOC value of monomer i; is the average SOC value of the battery pack; N is the total number of individual cells.

[0026] Preferably, the formula for weighted fusion is: ; in, To determine the desired output duty cycle, the output of the SOC fuzzy controller is... Weighting coefficients for control For the output of the voltage fuzzy controller Weighting coefficients for control , and The magnitudes are all distributed between 0 and 1. After weighted fusion, the duty cycle of the desired output is obtained. .

[0027] Preferably, the fuzzy rules of the SOC fuzzy controller and the voltage fuzzy controller are as follows: when the mean of the input variable is extremely small or extremely large, the output duty cycle is small; when the mean of the input variable is in the middle value, the output duty cycle is appropriately increased according to the magnitude of the difference between the input variables.

[0028] Preferably, the fuzzy rule of the variable universe fuzzy controller is: when the difference between the input variables is extremely small and the dispersion is extremely small, the output scaling factor is a minimum value; when the difference between the input variables is extremely large and the dispersion is extremely large, the output scaling factor is a maximum value.

[0029] Preferably, the fuzzy rule of the adaptive fuzzy controller is: when the mean of the input variable is extremely small or extremely large, the output weight coefficient... Increase Decrease; when the mean of the input variable is at the median, the output weight coefficient is reduced. Increase Decrease.

[0030] Example 2: This embodiment provides a novel battery balancing strategy, which is analyzed from five aspects: the establishment of multiple balancing variables, the design of balancing control strategy, the design of voltage and SOC fuzzy controllers, the design of variable universe fuzzy controllers, and the design of adaptive fuzzy controllers.

[0031] First, using a 26650LiFePO4 battery as the subject, an HPPC hybrid pulse experiment was conducted. The experimental data obtained were then fitted using a polynomial to obtain the following results: Figure 1 The OCV-SOC curve is shown below.

[0032] Depend on Figure 1 It can be seen that small changes in SOC in the ranges of 0%≤SOC≤15% and 95%≤SOC≤100% can cause large fluctuations in OCV. If SOC is used as the equilibrium variable only in these ranges, the voltage approaching the limit value may be ignored, leading to overcharging and over-discharging of the battery.

[0033] When the State of Charge (SOC) is between 15% and 95%, the OVC changes very gradually. However, when the voltage differences between battery packs are small but the SOC differences are large, using only voltage as the balancing variable cannot promptly complete the energy transfer between high-SOC and low-SOC cells, affecting the overall balancing performance of the battery pack. Therefore, using a single balancing variable for balancing control has limitations and affects the balancing effect.

[0034] like Figure 2 As shown, this embodiment proposes a balanced strategy for coordinated control of SOC and voltage; the design process of the voltage and SOC fuzzy controller is as follows: (1) The two input variables of the voltage and SOC fuzzy controller are the mean and difference of SOC or voltage of adjacent battery cells or battery packs, and the one output variable is the duty cycle that adjusts the magnitude of the equalization current.

[0035] Five fuzzy variables are formed into a fuzzy set {SS, S, M, B, BB}, representing minimum, small, moderate, large, and maximum respectively.

[0036] The domain of discourse is

[01] . The fundamental universe of discourse is [00.5], and the duty cycle of the controller output is... The domain of discourse is [0, 0, 5]. The domain of discourse is [2.43.6]. The universe of discourse is [0, 0, 5], and the output duty cycle is... The domain of discourse is [0, 0, 5].

[0037] The membership functions of both input and output variables are composed of triangles and trapezoids. The membership functions of the input and output variables of the SOC fuzzy controller are as follows: Figures 3-4 As shown, the membership functions of the input and output variables of the voltage fuzzy controller are as follows: Figures 5-6 As shown.

[0038] Output variables for SOC fuzzy controller and voltage fuzzy controller output variables The established fuzzy rules are similar, as shown in Table 1 below.

[0039] Table 1: Fuzzy rules;

[0040] The core of rule-making lies in improving balancing speed and efficiency while ensuring charging and discharging safety and avoiding overcharging and over-discharging of batteries.

[0041] When the battery When smaller, regardless Regardless of the differences, to prevent over-discharge of the battery, and prioritizing safety, a smaller balancing current is used. The value should not be too large, and similarly... This is also to avoid overcharging when the charge is large.

[0042] when When it is in the middle value, according to The magnitude of the difference can be appropriately increased. A larger balancing current is used to improve the speed and efficiency of balancing.

[0043] (2) The design process of variable universe fuzzy control is as follows: Battery and The initial universe of discourse will gradually decrease as the equilibrium is reached, resulting in an excessively large universe of discourse that affects the control effect. This embodiment introduces the principle of variable universe of discourse and optimizes the control accuracy by dynamically adjusting the fuzzy universe of discourse.

[0044] Traditional function modeling methods for selecting scaling factors are easily affected by the model and parameters, while fuzzy logic control methods can more easily describe the changes in scaling factors.

[0045] or and its dispersion The scaling factor serves as the input to the variable universe fuzzy controller. After introducing a scaling factor into its output and The domain of discourse is adjusted to [0 ].

[0046] Taking SOC as an example, its dispersion expression is: ; In the formula, This represents the SOC value of monomer i; is the average SOC value of the battery pack; N is the total number of individual cells.

[0047] variable and Membership function graphs are respectively composed of Figure 3 and Figure 5 As shown, variables and Membership function such as Figure 7 and Figure 8 As shown, The rules for setting the voltage and SOC are shown in Table 2. The rules for setting the voltage and SOC are similar.

[0048] Table 2: Fuzzy rules;

[0049] (3) The design process of adaptive fuzzy control is as follows: The entire equilibrium process is controlled collaboratively by two variables: SOC and voltage. The adaptive fuzzy circuit selects the corresponding weight values ​​for the two variables based on the current equilibrium state.

[0050] Adjacent cells or battery packs For one of the input variables of the adaptive fuzzy controller, select the equalization current. As another equilibrium variable, Reflects the current state of battery energy, current The intensity of the current equilibrium is reflected, and the effects of both achieve multi-dimensional integrated control.

[0051] The output variable is the output of the SOC fuzzy controller. Weighting coefficients for control and the output of the voltage fuzzy controller Weighting coefficients for control , and The magnitudes are all distributed between 0 and 1. After weighted fusion, the duty cycle of the desired output is obtained. The expression is as follows: ; Input variables The fuzzy domain is

[01] , and the input variables are... The fuzzy universe of discourse is

[010] , and the weight coefficients are... and The domain of discourse is [0, 1]. The input variables use triangular and trapezoidal membership functions, and the output variables all use triangular functions. Figures 9-11 As shown.

[0052] right and The two output variables are defined with the fuzzy rules shown in Tables 3 and 4 below.

[0053] Table 3: Fuzzy rules

[0054] Table 4: Fuzzy rules

[0055] The core requirements for rule formulation are as follows: when input variables When the mean is extremely high or extremely low, according to the OVC-SOC curve, voltage equalization should be prioritized, and small current equalization should be performed to avoid overcharging and over-discharging of the battery.

[0056] when When the value is in the middle, SOC balancing should be performed first, and the weighting coefficients should be increased appropriately. To accelerate the balancing process during this stage, a high-current balancing method can be appropriately adopted.

[0057] Three fuzzy controllers using the centroid method were employed for model decomposition, with the three-dimensional relationship between the input and output of the adaptive fuzzy controller as follows: Figure 12 and Figure 13 As shown.

Claims

1. A lithium battery equalization method based on improved fuzzy control, characterized in that, Includes the following steps: Obtain the OCV-SOC curve of the battery pack, and divide the SOC into a low SOC range, a middle SOC range and a high SOC range according to the OCV-SOC curve. In the low SOC range and the high SOC range, the OCV changes drastically with the SOC, while in the middle SOC range, the OCV changes slowly with the SOC. Design voltage fuzzy controller and SOC fuzzy controller: The voltage fuzzy controller uses the average voltage of adjacent battery cells or battery packs. and voltage difference The input variable is the duty cycle that adjusts the magnitude of the balancing current, and the output variable is the duty cycle that adjusts the magnitude of the balancing current. The SOC fuzzy controller uses the average SOC of adjacent battery cells or battery packs. Difference between SOC and The input variable is the duty cycle that adjusts the magnitude of the balancing current, and the output variable is the duty cycle that adjusts the magnitude of the balancing current. Design a variable universe of discourse fuzzy controller for dynamically adjusting voltage difference. Difference between SOC and The universe of discourse, wherein the variable universe of discourse fuzzy controller uses voltage difference or SOC difference The voltage difference is adjusted by taking its dispersion as the input variable and the scaling factor as the output variable. or SOC difference The initial domain of discourse; Design an adaptive fuzzy controller based on the average SOC of adjacent battery cells or battery packs. Using the equalization current as input variables, and the weighting coefficients of the SOC fuzzy controller output duty cycle and the voltage fuzzy controller output duty cycle as output variables, the desired output duty cycle is obtained by weighted fusion of the SOC fuzzy controller output duty cycle and the voltage fuzzy controller output duty cycle, which is used to dynamically adjust the magnitude of the equalization current.

2. The lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The low SOC range is 0% ≤ SOC ≤ 15%, the high SOC range is 95% ≤ SOC ≤ 100%, and the intermediate SOC range is 15% ≤ SOC ≤ 95%.

3. The lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The input and output variables of the voltage fuzzy controller and the SOC fuzzy controller are divided into 5 fuzzy sets, which include SS, S, M, B, and BB, representing minimal, small, moderate, large, and maximal, respectively.

4. The lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The SOC mean, the input variable of the SOC fuzzy controller The universe of discourse is [0,1], and the SOC difference is... The universe of discourse for the output duty cycle is [0, 0.5], and the universe of discourse for the output duty cycle is [0, 0.5]; the average voltage value is the input variable of the voltage fuzzy controller. The universe of discourse is [2.4, 3.6], and the voltage difference is... The universe of discourse is [0, 0.5], and the universe of discourse for the output duty cycle is [0, 0.5].

5. A lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The variable universe fuzzy controller uses the SOC difference. Or its dispersion as input, or voltage difference Or its discreteness is the input.

6. The lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The dispersion is the standard deviation of the SOC difference or voltage difference, and the calculation formula is: ； In the formula, This represents the SOC value of monomer i; This represents the average SOC value of the battery pack. N represents the total number of individuals.

7. The lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The formula for the weighted fusion is: ； in, To determine the desired output duty cycle, the output of the SOC fuzzy controller is... Weighting coefficients for control For the output of the voltage fuzzy controller Weighting coefficients for control , and The magnitudes are all distributed between 0 and 1. After weighted fusion, the duty cycle of the desired output is obtained. .

8. A lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The fuzzy rules of the SOC fuzzy controller and the voltage fuzzy controller are as follows: when the mean of the input variable is extremely small or extremely large, the output duty cycle is small; when the mean of the input variable is in the middle value, the output duty cycle is appropriately increased according to the magnitude of the difference between the input variables.

9. A lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The fuzzy rule of the variable universe fuzzy controller is as follows: when the difference between the input variables is extremely small and the dispersion is extremely small, the output scaling factor is a minimum value; when the difference between the input variables is extremely large and the dispersion is extremely large, the output scaling factor is a maximum value.

10. A lithium battery equalization method based on improved fuzzy control according to claim 1, characterized in that, The fuzzy rule of the adaptive fuzzy controller is: when the mean of the input variable is extremely small or extremely large, the output weight coefficient is... Increase Decrease; when the mean of the input variable is at the median, the output weight coefficient is reduced. Increase Decrease.

Images