Real-time heat dissipation monitoring, regulating and controlling method for vehicle-mounted starting inverter power supply battery
Through real-time temperature acquisition and LSTM model prediction, the cooling device of the inverter power battery is dynamically regulated, which solves the problem of single regulation strategy in the existing technology, realizes the stable operation of the battery within the safe temperature range, and extends the battery life.
Patent Information
- Application Number
- CN202411115289.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing methods of on-board inverter power battery monitoring and regulation cannot flexibly regulate based on real-time temperature changes and predicted results, resulting in the battery being easily overheated or overcooled, affecting battery life and performance.
By periodically collecting data based on the temperature sensor, filtering processing, predicting temperature change trends using the LSTM model, and building a safe temperature range based on real-time and predicted results, the heat dissipation device is dynamically regulated.
It effectively avoids performance degradation or safety risks caused by overheating or overcooling of the battery, and extends the service life and stability of the battery.
Smart Images

Figure CN119029411A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of inverter power supply battery system, and more specifically, relates to a real-time heat dissipation monitoring and control method for a vehicle-mounted starting inverter power supply battery. Background Art
[0002] Inverter batteries are widely used in electric vehicles, hybrid vehicles and other on-board devices that require efficient power conversion. With the rapid development of the electric vehicle market, the demand for inverter batteries is also increasing. These batteries not only need to provide efficient power conversion, but also need to have safe and stable working performance to cope with various complex on-board environments.
[0003] The current on-board inverter power battery monitoring and control methods mainly rely on the following means:
[0004] Temperature sensor monitoring: Temperature sensors are installed inside and outside the battery to collect temperature data in real time to monitor the battery's operating status.
[0005] Simple control strategy: Based on the preset temperature threshold, when the temperature exceeds the safe range, simple switch control or cooling equipment startup and shutdown are performed.
[0006] Data logging and analysis: Temperature data is logged and analyzed offline to assess battery performance and health.
[0007] However, the current control strategy is single and lacks the ability to dynamically adjust. It cannot be flexibly controlled according to real-time temperature changes and prediction results, which can easily cause the battery to overheat or overcool, affecting battery life and performance. Summary of the invention
[0008] The present invention provides a real-time heat dissipation monitoring and control method for a vehicle-mounted starting inverter power supply battery. The current control strategy is single and lacks the ability of dynamic adjustment. It cannot be flexibly controlled according to real-time temperature changes and prediction results, which easily leads to technical problems such as battery overheating or overcooling.
[0009] A real-time heat dissipation monitoring and control method for a vehicle-mounted starting inverter power supply battery comprises the following steps:
[0010] Step 1: Periodically collect temperature data of the inverter power supply battery based on the temperature sensor;
[0011] Step 2: Filter the collected temperature data to remove noise;
[0012] Step 3: Calculate the temperature change trend of the inverter battery according to the filtered temperature data, and use the machine learning algorithm to predict the change trend of the inverter battery in the future based on the temperature change trend;
[0013] Step 4: Build a safe temperature range for the inverter battery based on the real-time temperature and prediction results, and dynamically adjust the heat dissipation device.
[0014] The present invention constructs a safe temperature range for the inverter power battery based on the real-time temperature and prediction results, and dynamically controls the heat dissipation device, which can effectively solve the technical problem of a single control strategy; uses filtering technology and machine learning prediction to achieve efficient monitoring and accurate prediction of the battery, and dynamically adjusts the working state of the heat dissipation device according to the real-time and predicted temperature to ensure that the battery operates within a safe temperature range, effectively avoiding performance degradation or safety risks caused by overheating or overcooling of the battery, and extending the service life and stability of the battery.
[0015] Preferably, the temperature variation trend of the inverter power supply battery is calculated as follows:
[0016] Calculate the temperature change rate at the current moment relative to the previous moment:
[0017] ΔT(t)=T(t)-T(t-1);
[0018] Where: ΔT(t) represents the temperature change rate at time t; T(t) represents the temperature data at time t; T(t-1) represents the temperature data at time t-1;
[0019] The moving average method is used to smooth the temperature change rate and obtain the trend curve:
[0020]
[0021] Where: ΔT(t)′ represents the moving average of the temperature change rate at time t; N represents the window size of the moving average; ΔT(ti) represents the temperature change rate at time ti; w i Represents weight, satisfying
[0022] Preferably, the machine learning algorithm adopts an LSTM model, and the structure of the LSTM model is as follows:
[0023] Input layer: used to receive time series feature data;
[0024] LSTM layer: used to capture the temporal dependency of temperature changes;
[0025] Attention layer: highlights important time step features based on a weighted mechanism;
[0026] Fully connected layer: maps the outputs of the LSTM layer and the Attention layer to obtain the prediction results;
[0027] Output layer: Output the prediction results, that is, output the predicted value of the temperature change rate in the future time step.
[0028] Preferably, the loss function of the LSTM model is as follows:
[0029]
[0030] Where: n represents the number of samples; ΔT(t)i represents the true temperature change rate of the i-th sample; represents the predicted temperature change rate of the i-th sample; λ represents the penalty coefficient; ExtremePenalty represents the penalty term for the extreme temperature prediction value.
[0031] Preferably, the LSTM layer includes:
[0032] Forget Gate ft:
[0033] f t =σ(W f ·[h t-1 , X t ]+b f );
[0034] Where: W f represents the weight matrix of the forget gate; b f represents the bias vector of the forget gate; σ represents the Sigmoid activation function; X t represents the input features of the current time step; h t-1 represents the hidden state at time step t-1;
[0035] Input Gate i t :
[0036] i t =σ(W i ·[h t-1 , X t ]+b i );
[0037] Where: W i represents the weight matrix of the input gate; b i represents the bias vector of the input gate;
[0038] Candidate memory element
[0039]
[0040] Where: tanh represents the tanh activation function; W C represents the weight matrix of candidate memory elements; b C The bias vector representing the candidate memory element;
[0041] Memory element update C t :
[0042]
[0043] Where: C t-1 Represents the memory element of the previous time step;
[0044] Output gate o t :
[0045] o t =σ(W o ·[h t-1 , X t ]+b o );
[0046] Where: W o represents the weight matrix of the output gate; b o Represents the bias vector of the output gate;
[0047] Hidden state h t :
[0048] h t =o t tanh(C t );
[0049] Output H1 = [h1, h2, ..., h 10 ].
[0050] Preferably, the specific steps of the Attention layer highlighting important time step features based on the weighted mechanism are as follows:
[0051] The attention score e is calculated based on the demerits obtained in the last LSTM layer. t :
[0052] e t =tanh(W a ·h′ t +b a );
[0053] Where: W a The weight matrix representing the attention score; b a The bias vector representing the attention score; h′ t represents the hidden state of the last LSTM layer at time step t; tanh represents the tanh activation function;
[0054] Calculate the attention weight a t :
[0055]
[0056] Where: a t represents the attention weight at time step t; exp represents the exponential function, which is used to calculate the soft maximization of the attention score;
[0057] Calculate the context vector c t :
[0058]
[0059] Where: a k represents the attention weight at time step k; h′ k Represents the hidden state of the last LSTM layer at time step t; combines the context vector and LSTM output:
[0060] H combined =[H n , c t ];
[0061] Where: H combined H represents the combined vector of the output of the last LSTM layer and the context vector; n Represents the output of the last LSTM layer.
[0062] Preferably, step 4 comprises the following steps:
[0063] Use the LSTM model to predict the temperature change rate ΔT(t) in the future and accumulate the temperature prediction value at the future moment:
[0064]
[0065] Where: T pred (t+k) represents the predicted temperature value at the kth time step in the future; ΔT(t+i) represents the predicted temperature change rate at the t+i time step;
[0066] Based on T pred (t+k) Dynamically adjust the heat sink:
[0067]
[0068] Where: T min Indicates the lowest safe temperature for battery operation; T max Indicates the maximum safe temperature for battery operation.
[0069] The beneficial effects of the present invention include:
[0070] The present invention constructs a safe temperature range for the inverter power battery based on the real-time temperature and prediction results, and dynamically controls the heat dissipation device, which can effectively solve the technical problem of a single control strategy; uses filtering technology and machine learning prediction to achieve efficient monitoring and accurate prediction of the battery, and dynamically adjusts the working state of the heat dissipation device according to the real-time and predicted temperature to ensure that the battery operates within a safe temperature range, effectively avoiding performance degradation or safety risks caused by overheating or overcooling of the battery, and extending the service life and stability of the battery. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0072] Figure 1 The overall system architecture diagram provided for the embodiment of the present invention. DETAILED DESCRIPTION
[0073] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0074] It should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element.
[0075] It should be understood that the orientation or position relationship indicated by terms such as "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside" and "outside" are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.
[0076] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0077] See also Figure 1 As shown, the best embodiment of the present invention is further described;
[0078] A real-time heat dissipation monitoring and control method for a vehicle-mounted starting inverter power supply battery comprises the following steps:
[0079] Step 1: Periodically collect temperature data of the inverter power supply battery based on the temperature sensor;
[0080] Step 2: Filter the collected temperature data to remove noise; the filtering adopts Kalman filtering.
[0081] Step 3: Calculate the temperature change trend of the inverter battery according to the filtered temperature data, and use the machine learning algorithm to predict the change trend of the inverter battery in the future based on the temperature change trend;
[0082] The temperature variation trend of the inverter power supply battery is calculated as follows:
[0083] Calculate the temperature change rate at the current moment relative to the previous moment:
[0084] ΔT(t)=T(t)-T(t-1);
[0085] Where: ΔT(t) represents the temperature change rate at time t; T(t) represents the temperature data at time t; T(t-1) represents the temperature data at time t-1;
[0086] The moving average method is used to smooth the temperature change rate and obtain the trend curve:
[0087]
[0088] Where: ΔT(t)′ represents the moving average of the temperature change rate at time t; N represents the window size of the moving average; ΔT(ti) represents the temperature change rate at time ti; w i Represents weight, satisfying
[0089] In this embodiment, multiple data points are averaged by moving average, which can reduce the impact of noise and improve data accuracy. In the process of moving average, weights are assigned to each data point to further optimize the smoothing effect. By assigning larger weights to newer data points in the window (for example, data close to the current moment), the smoothed results can better reflect the current temperature change trend.
[0090] The machine learning algorithm adopts the LSTM model, and the structure of the LSTM model is as follows:
[0091] Input layer: used to receive time series feature data;
[0092] LSTM layer: used to capture the temporal dependency of temperature changes;
[0093] Attention layer: highlights important time step features based on a weighted mechanism;
[0094] Fully connected layer: maps the outputs of the LSTM layer and the Attention layer to obtain the prediction results;
[0095] Output layer: Output the prediction results, that is, output the predicted value of the temperature change rate in the future time step.
[0096] In this embodiment, the Attention layer can dynamically highlight important time step features by calculating the attention weight of each time step, so the model can adaptively focus on the time most relevant to the current prediction, rather than treating the data of all time steps equally; and in long time series data, the contribution of different time steps to the final prediction is uneven, so the Attention layer helps the model identify and focus on time steps with significant information, thereby improving the prediction accuracy and the overall performance of the model.
[0097] Preferably, the loss function of the LSTM model is as follows:
[0098]
[0099] Where: n represents the number of samples; ΔT(t)i represents the true temperature change rate of the i-th sample; represents the predicted temperature change rate of the i-th sample; λ represents the penalty coefficient; ExtremePenalty represents the penalty term for the extreme temperature prediction value.
[0100] In this embodiment, a penalty for extreme predicted values is added to the loss function. Since extreme temperatures may cause performance degradation or safety risks of the battery system, it is used to limit the extreme values output by the model, ensure that the predicted values are within a reasonable range, and enhance the stability of the model. The intensity of the penalty term is adjusted through the penalty coefficient to balance the relationship between the model's pursuit of prediction accuracy and the constraint on extreme values.
[0101] Preferably, the LSTM layer includes:
[0102] Forget Gate t :
[0103] f t =σ(W f ·[h t-1 , X t ]+b f );
[0104] Where: W f represents the weight matrix of the forget gate; b f represents the bias vector of the forget gate; σ represents the Sigmoid activation function; X t represents the input features of the current time step; h t-1 represents the hidden state at time step t-1;
[0105] Input Gate i t :
[0106] i t =σ(W i ·[h t-1 , X t ]+b i );
[0107] Where: W i represents the weight matrix of the input gate; b i represents the bias vector of the input gate;
[0108] Candidate memory element
[0109]
[0110] Where: tanh represents the tanh activation function; W C represents the weight matrix of candidate memory elements; b C The bias vector representing the candidate memory element;
[0111] Memory element update C t :
[0112]
[0113] Where: Ct-1 Represents the memory element of the previous time step;
[0114] Output gate o t :
[0115] o t =σ(W o ·[h t-1 , X t ]+b o );
[0116] Where: W o represents the weight matrix of the output gate; b o Represents the bias vector of the output gate;
[0117] Hidden state h t :
[0118] h t =o t tanh(C t );
[0119] Output H1 = [h1, h2, ..., h 10 ].
[0120] Preferably, the specific steps of the Attention layer highlighting important time step features based on the weighted mechanism are as follows:
[0121] The attention score e is calculated based on the demerits obtained in the last LSTM layer. t :
[0122] e t =tanh(W a ·h ′ t+b a );
[0123] Where: W a The weight matrix representing the attention score; b a The bias vector representing the attention score; h′ t represents the hidden state of the last LSTM layer at time step t; tanh represents the tanh activation function;
[0124] Calculate the attention weight a t :
[0125]
[0126] Where: a t represents the attention weight at time step t; exp represents the exponential function, which is used to calculate the soft maximization of the attention score;
[0127] Calculate the context vector ct :
[0128]
[0129] Where: a k represents the attention weight at time step k; h′ k Represents the hidden state of the last LSTM layer at time step t; combines the context vector and LSTM output:
[0130] H combined =[H n , c t ];
[0131] Where: H combined H represents the combined vector of the output of the last LSTM layer and the context vector; n Represents the output of the last LSTM layer.
[0132] Step 4: Build a safe temperature range for the inverter battery based on the real-time temperature and prediction results, and dynamically adjust the heat dissipation device.
[0133] The step 4 comprises the following steps:
[0134] Use the LSTM model to predict the temperature change rate ΔT(t) in the future and accumulate the temperature prediction value at the future moment:
[0135]
[0136] Where: T pred (t+k) represents the predicted temperature value at the kth time step in the future; ΔT(t+i) represents the predicted temperature change rate at the t+i time step;
[0137] Based on T pred (t+k) Dynamically adjust the heat sink:
[0138]
[0139] Where: T min Indicates the lowest safe temperature for battery operation; T max Indicates the maximum safe temperature for battery operation.
[0140] The present invention constructs a safe temperature range for the inverter power battery based on the real-time temperature and prediction results, and dynamically controls the heat dissipation device, which can effectively solve the technical problem of a single control strategy; uses filtering technology and machine learning prediction to achieve efficient monitoring and accurate prediction of the battery, and dynamically adjusts the working state of the heat dissipation device according to the real-time and predicted temperature to ensure that the battery operates within a safe temperature range, effectively avoiding performance degradation or safety risks caused by overheating or overcooling of the battery, and extending the service life and stability of the battery.
[0141] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.
Claims
1. A real-time heat dissipation monitoring and control method for a vehicle-mounted starting inverter power supply battery, characterized in that: The following steps are involved: Step 1: Periodically collect temperature data of the inverter power supply battery based on the temperature sensor; Step 2: Filter the collected temperature data to remove noise; Step 3: Calculate the temperature change trend of the inverter battery according to the filtered temperature data, and use the machine learning algorithm to predict the change trend of the inverter battery in the future based on the temperature change trend; Step 4: Build a safe temperature range for the inverter battery based on the real-time temperature and prediction results, and dynamically adjust the heat dissipation device.
2. The real-time heat dissipation monitoring and control method of a vehicle-mounted inverter power battery according to claim 1 is characterized in that: The temperature variation trend of the inverter power supply battery is calculated as follows: Calculate the temperature change rate at the current moment relative to the previous moment: ΔT(t)=T(t)-T(t-1); Where: ΔT(t) represents the temperature change rate at time t; T(t) represents the temperature data at time t; T(t-1) represents the temperature data at time t-1; The moving average method is used to smooth the temperature change rate and obtain the trend curve: Where: ΔT(t)′ represents the moving average of the temperature change rate at time t; N represents the window size of the moving average; ΔT(ti) represents the temperature change rate at time ti; w i Represents weight, satisfying 3. The real-time heat dissipation monitoring and control method of a vehicle-mounted inverter power battery according to claim 1 is characterized in that: The machine learning algorithm adopts the LSTM model, and the structure of the LSTM model is as follows: Input layer: used to receive time series feature data; LSTM layer: used to capture the temporal dependency of temperature changes; Attention layer: highlights important time step features based on a weighted mechanism; Fully connected layer: maps the outputs of the LSTM layer and the Attention layer to obtain the prediction results; Output layer: Output the prediction results, that is, output the predicted value of the temperature change rate in the future time step.
4. The real-time heat dissipation monitoring and control method of a vehicle-mounted inverter power battery according to claim 3 is characterized in that: The loss function of the LSTM model is as follows: Where: n represents the number of samples; ΔT(t) i represents the true temperature change rate of the i-th sample; represents the predicted temperature change rate of the i-th sample; λ represents the penalty coefficient; ExtremePenalty represents the penalty term for the extreme temperature prediction value.
5. The real-time heat dissipation monitoring and control method of a vehicle-mounted inverter power battery according to claim 3 is characterized in that: The LSTM layer includes: Forget Gate t : f t =σ(W f ·[h t-1 ,X t ]+b f ); Where: W f represents the weight matrix of the forget gate; b f represents the bias vector of the forget gate; σ represents the Sigmoid activation function; X t represents the input features of the current time step; h t-1 represents the hidden state at time step t-1; Input Gate i t : i t =σ(W i ·[h t-1 ,X t ]+b i ); Where: W i represents the weight matrix of the input gate; b i represents the bias vector of the input gate; Candidate memory element Where: tanh represents the tanh activation function; W C represents the weight matrix of candidate memory elements; b C The bias vector representing the candidate memory element; Memory element update C t : Where: C t-1 Represents the memory element of the previous time step; Output gate o t : the t =σ(W o ·[h t-1 ,X t ]+b o ); Where: W o represents the weight matrix of the output gate; b o represents the bias vector of the output gate; Hidden state h t : h t =o t ·tanh(C t ); Output H1 = [h1, h2, ..., h 10 ].
6. The real-time heat dissipation monitoring and control method of a vehicle-mounted inverter power battery according to claim 3 is characterized in that: The specific steps of the Attention layer to highlight important time step features based on the weighted mechanism are as follows: Calculate the attention score e based on the demerits obtained in the last LSTM layer t : e t =tanh(W a ·h′ t +b a ); Where: W a The weight matrix representing the attention score; b a The bias vector representing the attention score; h′ t represents the hidden state of the last LSTM layer at time step t; tanh represents the tanh activation function; Calculate the attention weight a t : Where: a t represents the attention weight at time step t; exp represents the exponential function, which is used to calculate the soft maximization of the attention score; Calculate the context vector c t : Where: a k represents the attention weight at time step k; h′ k represents the hidden state of the last LSTM layer at time step t; Combine the context vector and LSTM output: H combined =[H n ,c t ]; Where: H combined H represents the combined vector of the output of the last LSTM layer and the context vector; n Represents the output of the last LSTM layer.
7. The real-time heat dissipation monitoring and control method of a vehicle-mounted inverter power battery according to claim 1 is characterized in that: The step 4 comprises the following steps: Use the LSTM model to predict the temperature change rate ΔT(t) in the future and accumulate the temperature prediction value at the future moment: Where: T pred (t+k) represents the predicted temperature value at the kth time step in the future; ΔT(t+i) represents the predicted temperature change rate at the t+i time step; Based on T pred (t+k) Dynamically adjust the heat sink: Where: T min Indicates the lowest safe temperature for battery operation; T max Indicates the maximum safe temperature for battery operation.
Citation Information
Cited By
Over-temperature control method for boiler wall temperature
CN121091928A
