Artificial neural network based battery liquid cooling system and control method

Abstract

The invention relates to the battery liquid cooling system and control method based on an artificial neural network, which predicts the temperature change of the battery pack before it reaches the critical temperature by making temperature predictions based on an artificial neural network of battery packs used in large-scale and high-power structures such as electric vehicles (EV) and energy storage systems (ESS) and optimizing the cooling capacity of the cooling system Thanks to the system and method in the invention, the battery is kept within the temperature range recommended by the battery pack manufacturer, its aging is reduced and energy saving is achieved in the battery thermal management system.

Description

[0001] DESCRIPTION

[0002] ARTIFICIAL NEURAL NETWORK BASED BATTERY LIQUID COOLING SYSTEM AND CONTROL METHOD

[0003] TECHNICAL FIELD

[0004] The invention relates to the battery liquid cooling system and control method based on an artificial neural network, which predicts the temperature change of the battery pack before it reaches the critical temperature by making temperature predictions based on an artificial neural network of battery packs used in large-scale and high-power structures such as electric vehicles (EV) and energy storage systems (ESS) and optimizing the cooling capacity of the cooling system. Thanks to the system and method in the invention, the battery is kept within the temperature range recommended by the battery pack manufacturer, its aging is reduced and energy saving is achieved in the battery thermal management system.

[0005] BACKGROUND ART

[0006] Nowadays, temperature control of battery packs is provided by liquid, air, thermoelectric, or phase change materials, especially in large-scale energy storage systems (ESS) and electric vehicles (EV). Liquid or air cooling systems are widely preferred among these systems. The basic components of liquid cooling systems consist of one or more cooling plates positioned in contact with the battery pack to provide heat transfer, a circulation pump that circulates the liquid fluid, valves that control the flow rate of the liquid fluid, temperature sensors connected to the tabs or different areas of the battery, the main Electronic Control Unit (ECU) that processes the data from these sensors, and the radiator to which the liquid fluid is sent to compensate for the heat absorbed by the liquid fluid as a result of the heat transfer with the battery pack. Liquid cooling systems operating in a closed loop are integrated with Battery Management Systems (BMS) and Battery Thermal Management Systems (BTMS) via the main ECU. Similarly, temperature sensors and the ECU are used in air cooling systems. In addition, instead of a liquid cooling plate, there are one or more fans and air ducts that cover the surface of the battery. The fan operating speed is adjusted using Pulse Width Modulation (PWM) or a similar control method, depending on the battery temperature and / or ambient temperature.

[0007] The general problem in the temperature control part of the systems described above is that the cooling process starts after the battery reaches a certain temperature limit, i.e. , the fan speed control or the temperature and flow rate of the liquid fluid are adjusted. Since this problem cannot provide instantaneous temperature control and improvement, it seriously reduces the battery life in the long term. According to the literature, every 5°C temperature increase in the battery during 2000 charge-discharge cycles reduces the battery capacity by approximately 5%. Capacity loss increases more rapidly at high outdoor temperatures. Another consequence of the inadequacy of instantaneous temperature control is that the amount of energy spent to keep the battery pack within the recommended temperature range, especially at very hot outdoor temperatures (40°C and above), can be equivalent to 30% or more of the battery capacity. This negatively affects the end-user experience and system stability.

[0008] BRIEF DESCRIPTION OF THE INVENTION

[0009] The invention relates to the battery liquid cooling system and control method based on an artificial neural network, which predicts the temperature change of the battery pack before it reaches the critical temperature by predicting temperature based on an artificial neural network of battery packs used in large-scale and high- power structures such as electric vehicles (EV) and energy storage systems (ESS) and optimizing the cooling capacity of the cooling system. In addition, with the artificial intelligence algorithm integrated into the battery liquid cooling system, the temperature of the battery pack is predicted depending on the external environment temperature, instantaneous battery temperature, current intensity, operating voltage, state of charge (SOC), and parameters obtained from battery management systems (BMS), and accordingly, the coolant temperature and flow rate in liquid cooling systems are controlled instantly.

[0010] With the artificial intelligence algorithm in our invention, the cooling process is started before the critical temperature values are reached for the battery. Considering that the guaranteed usage period of a battery pack is between 10 and 20 years, the artificial intelligence controlled cooling system significantly increases the usage period by slowing down the battery aging, and energy saving is provided in the battery thermal management system. In addition to slowing down aging, by increasing the thermal stability of each cell in the battery pack, overall battery performance is also increased.

[0011] LIST OF FIGURES

[0012] Figure 1. Schematic View of the System

[0013] Figure 2. Artificial Intelligence Diagram

[0014] Correspondence of the Numbers Shown in the Figures

[0015] 1. Battery pack with liquid cooling plate

[0016] 2. Battery Management System

[0017] 3. Gateway electronic control unit (ECU)

[0018] 4. Artificial intelligence algorithm

[0019] 5. Cooling control algorithm

[0020] 6. Main electronic control unit (ECU)

[0021] 7. Liquid cooling system

[0022] DETAILED DESCRIPTION OF THE INVENTION

[0023] The invention is characterized by the components of the battery pack with liquid cooling plate (1 ), battery management system (BMS) (2), gateway electronic control unit (ECU) (3), artificial intelligence algorithm (4), cooling control algorithm (5), main electronic control unit (ECU) (6) and liquid cooling system (7).

[0024] The invention is characterized by the components of the gateway electronic control unit (ECU) (3) integrated between the battery management system (BMS) (2) and the main electronic control unit (ECU) (6) in the energy storage system, and the artificial intelligence algorithm (4) and cooling control algorithm (5) in the gateway ECU (3). The artificial intelligence algorithm (4) predicts future temperature with the sensor data received as input from the battery management system (BMS) (2) and the main ECU (6) and provides this data as input to the cooling control algorithm (5). The cooling control algorithm (5) increases the stability of the battery temperature and its performance by controlling the liquid coolant temperature, speed, and / or flow rate in liquid cooling systems (7) and ensures that the cooling process is carried out before the critical temperature is reached. It receives direct input from the artificial intelligence algorithm (4) for control. The artificial intelligence algorithm (4) and the cooling control algorithm (5) are placed in a new Gateway ECU (3) to be added between the BMS (2) and the main ECU (6). Liquid coolant inlet temperature and flow control commands generated by the artificial intelligence algorithm (4) and cooling control algorithm (5) are sent to the main ECU (6), allowing the cooling system to be controlled.

[0025] The steps of the method that controls the battery liquid cooling system (7) based on the artificial neural network in our invention are given below.

[0026] - The battery temperature values obtained from the temperature sensors on the battery pack with liquid cooling plate (1 ) and the voltage, current, state of charge (SOC), and operating time received from the circuit are collected and recorded by the BMS (2). With our invention, only data is received from the BMS (2), the invention does not have direct control over the BMS (2).

[0027] - BMS (2); sends the battery temperature values and the voltage, current, state of charge (SOC), and operating time data received from the circuit to the artificial intelligence algorithm (4) in the Gateway ECU (3).

[0028] - The main electronic control unit (ECU) (6) sends the data on the inlet and outlet temperature of the coolant in the liquid cooling system (7) to the cooling plate, the coolant speed and / or flow rate, the outdoor temperature and humidity to the artificial intelligence algorithm (4) in the gateway ECU (3).

[0029] - The artificial intelligence algorithm (4) receives the parameters sent from the BMS (2) and the main ECU (6) as input and makes the next temperature predictions at least every three minutes. Then, the artificial intelligence algorithm (4) sends these predictions to the cooling control algorithm (5).

[0030] - The cooling control algorithm (5) evaluates the short and long-term temperature prediction and sends new optimized values of the parameters coolant inlet temperature to the cooling plate, coolant speed, and / or flow rate to the main ECU (6) to control the battery temperature.

[0031] - The main ECU (6) applies the parameters of the coolant inlet temperature to the cooling plate, coolant speed, and / or flow rate to the liquid cooling system (7). - The liquid cooling system (7) sends the changed inlet temperature of the coolant to the cooling plate, the coolant speed and / or flow rate, and the outside temperature and humidity to the main ECU (6) via feedback to control the changed cooling parameters.

[0032] - The liquid cooling plate connected to the liquid cooling system (7) operates according to the parameters of the inlet temperature of the optimized cooler to the cooling plate, the cooling speed and / or flow rate, and continues to cool the battery pack. During the battery operating period, the battery temperature values obtained from the temperature sensors and the voltage, current, charge status, and operating time received from the circuit are collected and recorded by the BMS (2), and the cycle is completed.

[0033] In our invention; the current intensity applied to the battery pack, battery charge status, voltage, cooler flow rate and average temperature, outdoor temperature, and humidity are the parameters in the input layer. The input layer is sent to the 1st hidden layer of at least one of the 50-cell LSTM and RNN artificial neural network models to calculate the layer weights. Then, it is sent to the 2nd hidden layer and the same process is repeated. The average battery temperature is calculated in the output layer according to the weights obtained. Finally, with this output layer, the next average battery temperature predictions are obtained at least three minutes apart and sent to the cooling control algorithm (5), allowing the cooler circulating the cooling plate connected to the cooling system to be controlled.

[0034] The battery pack, average battery temperature, voltage, charge status, applied current intensity, and operating time are provided by the configuration of the battery management system (BMS) (2).

[0035] The battery pack, the inlet and outlet temperature of the coolant in the liquid cooling plate, the speed and / or flow rate of the coolant, and the outside temperature and humidity are provided by the configuration of the main electrical control unit (ECU) (6).

[0036] The artificial intelligence algorithm (4) has previously been tested at a current intensity of at least 0.1 C (optimum 0.1 , 0.2, 0.5, 1 , 2, 3, 4, 5, 10C), with at least one of the constant and / or dynamic current cycles; for at least 0.10 hours (optimum 10, 5, 2, 1 , 0.5, 0.33, 0.25, 0.20, 0.10 hours) of the operation time, at an outdoor temperature of at least 0°C (optimum 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60°C), in charging and discharging states; It has been trained at a coolant inlet temperature of at least 5°C (optimum 5, 10, 15, 20, 25, 30°C) and a coolant flow rate of at least 2 L / min (optimum 2, 4, 6, 8, 10, 12, 14, 16 L / min). Naturally, the training of the artificial intelligence algorithm (4) according to these values; For the battery pack with liquid cooling plate (1 ), at a current intensity of at least 0.1 C (optimum 0.1 , 0.2, 0.5, 1 , 2, 3, 4, 5, 10C), with at least one of the constant and / or dynamic current cycles; It provides a recording of parameters through experiments carried out at least 0.10 hours (optimum 10, 5, 2, 1 , 0.5, 0.33, 0.25, 0.20, 0.10 hours) operating time, at least 0°C (optimum 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60°C) outdoor temperature, in charge and discharge states; at least 5°C (optimum 5, 10, 15, 20, 25, 30°C) coolant inlet temperature and at least 2 L / min (optimum 2, 4, 6, 8, 10, 12, 14, 16 L / min) coolant flow rate.

[0037] The average battery temperature output is created with the input layer of the LSTM and RNN artificial neural network models, where the remaining parameters (voltage, state of charge, applied current intensity, operating time, inlet and outlet temperature of the cooler in the liquid cooling plate, speed and flow rate of the cooler, outdoor temperature, and humidity) are inputs; the input layer with the parameters, 2 hidden layers with 50 cells and the average temperature output layer. The data distribution selected for training, validation, and testing of the artificial neural network model is 70%, 15%, and 15%, respectively.

[0038] The data collected for current intensities, operating times, outdoor temperatures, charge and discharge states, cooler inlet temperatures, cooler speeds, and / or flow rates are organized and normalized; LSTM and RNN models are trained based on the mean square error metric and Adam optimizer. In case of underfitting in the model, at least one of the layer number and cell number is increased. In the case of overfitting in the model, at least one of the layer numbers and cell numbers is reduced. These LSTM and RNN artificial neural network models are tested with experimental test data after training and validation. The most accurate model with the lowest error is selected using MSE, RMSE, and MAE metrics, and its weights are recorded.

[0039] This neural network algorithm, the resulting neural network weights, and the cooling control algorithm (5) are placed in the gateway ECU (3) integrated between the battery management system and the main electrical control unit. The gateway ECU (3) receives the average battery temperature, voltage, state of charge, applied current, operating time, cooler inlet and outlet temperature, cooler speed, flow rate, outdoor temperature, and humidity from the battery management system and the main electrical control unit as inputs. The neural network algorithm predicts the next average battery temperature at least three minutes after the instantaneous operating time. The obtained temperature predictions are sent as input to the cooling control algorithm, where they are compared with the battery manufacturer's recommended battery operating temperature range. Based on these predictions, the cooling control algorithm (5) determines the coolant temperature, speed, and / or flow rate in the liquid cooling system (7) and sends it as a command to the main electrical control unit.

[0040] Depending on the instantaneous cooling performed by the liquid cooler control command sent to the main ECU (6), the average temperature of the battery system continues to be taken as input by the cooling control algorithm (5) in the gateway ECU (3). As the error margin between the actual average battery temperature and the predicted one increases, the cooling control algorithm (5) is instantly optimized, updated commands are sent to the main ECU (6) and the cooling control cycle is completed.

Claims

CLAIMS1. It is the method that controls the battery liquid cooling system based on the artificial intelligence algorithm (4), it is characterized by;- Collecting and recording the battery temperature values obtained from the temperature sensors on the battery pack with liquid cooling plate (1) and the voltage, current, charge status, and operating time obtained from the circuit using battery management systems (BMS) (2),- sending battery temperature values and voltage, current, charge status, and operating time data received from the circuit to the artificial intelligence algorithm (4) in the Gateway electronic control unit (ECU) (3),- the main electronic control unit (ECU) (6) sends the inlet and outlet temperature of the coolant in the liquid cooling system (7) to the cooling plate, the coolant speed and / or flow rate, the outdoor temperature and humidity data to the artificial intelligence algorithm (4) in the Gateway ECU (3),- the artificial intelligence algorithm (4) takes the parameters sent from the BMS (2) and the main ECU (6) as input and predicts future temperature at least every three minutes,- then the artificial intelligence algorithm (4) sends these predictions to the cooling control algorithm (5),- the cooling control algorithm (5) sends new optimized values of the parameters coolant inlet temperature to the cooling plate, coolant speed, and / or flow rate to the main ECU (6) to control the battery temperature by evaluating the short and long-term temperature predictions,- the main ECU (6) applies the parameters of the coolant inlet temperature to the cooling plate, coolant speed and / or flow rate to the liquid cooling system (7),- the liquid cooling system (7) checks the changed cooling parameters such as the inlet temperature of the cooling plate, the coolant speedand / or flow rate, the outside temperature and humidity, and sends them to the main ECU (6) via feedback,- the liquid cooling plate connected to the liquid cooling system (7) operates according to the cooling plate inlet temperature, cooling speed, and / or flow rate parameters of the optimized cooler and continues to cool the battery pack,- the stages of collecting and recording the battery temperature values obtained from the temperature sensors and the voltage, current, charge status, and operating time received from the circuit with the BMS (2) to continue the temperature and cooling controls during the battery operating period.

2. The method that controls the battery liquid cooling system mentioned in claim 1 , is characterized by; obtaining the parameters through experiments carried out for the battery pack with liquid cooling plate (1 ) with at least 0.1 C current intensity, at least one of the constant and / or dynamic current cycles; at least 0.10 hours of operation, at least 0°C outdoor temperature, in charging and discharging states, at least 5°C cooler inlet temperature and at least 2 L / min cooler flow rate.

3. Battery liquid cooling system based on the artificial intelligence algorithm (4) operating with the method mentioned in claims 1 and 2, it is characterized by;- gateway electronic control unit (ECU) (3), integrated between the battery management systems (BMS) (2) and the main electronic control unit (ECU) (6) in the energy storage system,- artificial intelligence algorithm (4) that is previously trained at different current intensities, at least one of the constant and / or dynamic current cycles, different operating times, different outdoor temperatures, different coolant inlet temperatures and flow rates, located in the gateway ECU (3), makes future temperature predictions using sensor data received as input from the BMS (2) andthe main ECU (6) and provides this data as input to the cooling control algorithm (5),- the cooling control algorithm (5) located within the gateway ECU (3), which controls the temperature, speed, and / or flow rate of the liquid coolant in liquid cooling systems (7), ensuring that the cooling process is carried out before the critical temperature is reached, thus increasing the stability and performance of the battery temperature.

Images