Abnormal early warning method for thermal management system of air-cooled lithium-ion battery energy storage container

Abstract

The present invention relates to an abnormality early warning method for a thermal management system of an air-cooled lithium-ion battery energy storage container. The method comprises: 1) extracting data of each air-conditioning system in the energy storage container and preprocessing the data; establishing, based on the preprocessed data, probability density distribution models of the temperature differences between inlet and return air in the cooling state and the heating state of each air-conditioning system respectively based on a kernel density estimation (KDE) model; calculating, based on the probability density distribution models, the cooling and heating health indicators of each air-conditioning system respectively; judging the abnormality risk of the air-conditioning system and giving corresponding early warnings according to the cooling and heating health indicators of each air-conditioning system; 2) extracting data of a battery cluster fan system in the energy storage container and preprocessing the data; based on the preprocessed data, performing outlier analysis on the relevant data using a DBSCAN density clustering algorithm to obtain outlier identification results; judging the abnormality risk of the battery cluster fan system and giving corresponding early warnings according to the outlier identification results.

Description

Technical Field

[0001] The present invention relates to the field of battery cluster and air conditioning early warning in energy storage containers, and in particular to an abnormality early warning method for a thermal management system of an air-cooled lithium-ion battery energy storage container. Background Art 1.1 Background Technology

[0003] Temperature has a significant impact on the capacity, charge and discharge power, and safety of lithium-ion batteries. Energy storage systems integrate more batteries and have greater battery capacity. The batteries in energy storage systems are also arranged more closely, with smaller gaps. The battery modules have high energy density and complex, variable operating conditions, often switching between high and low charge and discharge rates. This can easily lead to heat accumulation between battery packs, uneven heat generation within the system, uneven temperature distribution, and large temperature differences between batteries. Therefore, the safe and stable operation of the thermal management system plays a vital role in ensuring that the temperature and humidity of the energy storage system remain within a reasonable range throughout its life cycle.

[0004] At present, a large amount of research has been carried out both domestically and internationally on the structural design, thermal management strategies, and control methods and devices of thermal management systems for lithium-ion battery energy storage containers. However, there is still a lack of exploration in the abnormal warning of air-conditioning systems based on online monitoring data analysis of energy storage containers. The present invention analyzes the distribution and changing trends of the inlet and outlet air temperatures of each air conditioner in the energy storage container and the temperature distribution patterns of the battery cells in each battery cluster, and proposes an abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container. This method can accurately identify early abnormalities of the air-conditioning system and battery cluster fans in the container and issue graded warning information, effectively avoiding problems such as battery cell overheating or uneven temperature distribution caused by long-term abnormal functions of the thermal management system, thereby improving the operational reliability of the energy storage container.

[0005] 1.2 Prior Art Related to the Present Invention

[0006] 1.2.1 Technical solution of existing technology 1

[0007] An air conditioning cooling early warning method (CN108826614A) and an air conditioning heating early warning method (CN108800422A): This technical solution uses two temperature sensors to respectively detect the temperature of the indoor point farthest from the air conditioner and the outdoor temperature, and calculates the temperature difference between indoor and outdoor through the air conditioner controller. When the indoor temperature is lower than the outdoor temperature by a value higher than the judgment threshold, an alarm is issued.

[0008] 1.2.2 Disadvantages of Existing Technology 1

[0009] This technology only determines the operating status of the air conditioner by comparing the indoor and outdoor temperature differences. However, for the air conditioners in air-cooled lithium-ion battery energy storage containers, their cooling and heating set points change dynamically with the operating conditions of the energy storage container. It is difficult to accurately identify its abnormal status by comparing the temperature difference at a single moment. In addition, this technology can only realize the alarm function and does not have the early warning function of air conditioner abnormalities.

[0010] 1.3 Prior Art II Related to the Present Invention

[0011] 1.3.1 Technical solution of existing technology 2

[0012] A refrigerant leakage early warning method for rail vehicle air-conditioning units (CN112696791A): This technical solution provides a refrigerant leakage early warning method for rail vehicle air-conditioning units. The parameter change patterns of various sensors when the air-conditioning unit is operating normally and stably under different ambient temperature conditions are simulated in the laboratory. Through big data modeling, a working model of the air-conditioning unit under normal operation is obtained. The big data model is used to realize refrigerant leakage early warning during train operation.

[0013] 1.3.2 Disadvantages of Existing Technology 2

[0014] This technology only proposes big data modeling using multi-sensor data, but the specific modeling process is not detailed. Furthermore, since the warning target is the rail vehicle air conditioning system, its operating conditions and control strategies are relatively simple compared to those of energy storage containers, and it does not involve warning strategies for coordinated control of multiple air conditioners. Therefore, it is not suitable for warning air conditioning systems in energy storage containers.

[0015] 1.4 Prior Art 3 Related to the Present Invention

[0016] 1.4.1 Technical solution of existing technology 3

[0017] A method and system for early warning of air conditioner faults (CN110440390A): This technical solution proposes a method and system for early warning of air conditioner faults based on the number of abnormal noises from the air conditioner outdoor unit, the number of abnormal currents from the air conditioner, and the number of abnormal noises from the indoor unit. Fault early warning is achieved by weighted summation and scoring of each index.

[0018] 1.4.2 Disadvantages of Existing Technology 3

[0019] This technology requires recording abnormal noise counts from the air conditioning system's outdoor unit. However, energy storage containers operate in an environment with numerous, complex, and variable noise sources, making anomaly identification based on noise counts difficult to apply. Furthermore, the evaluation metrics used in this technology don't account for inlet and outlet air temperature deviations caused by cooling / heating anomalies due to factors like compressor and condensation leaks, making it difficult to accurately identify cooling and heating anomalies.

[0020] 1.5 Prior Art Related to the Present Invention

[0021] 1.5.1 Technical solution of existing technology 4

[0022] A fan failure warning method and device for a power supply system (CN110594177A): This technical solution proposes a fan failure warning method for a power supply system. The method detects the fan speed signal of a PWM fan to obtain the fan detection speed frequency, calculates the fan speed frequency threshold based on the duty cycle of the PWM fan PWM signal, and generates fan warning information based on the fan detection speed frequency and the fan speed frequency threshold. A fan failure warning device and method thereof (CN102758787A): This technology proposes a fan failure warning device and method thereof. The method mainly implements fan failure warning by analyzing the speed and current peak of the fan motor power supply to set an abnormality judgment threshold.

[0023] 1.5.2 Disadvantages of Existing Technology 4

[0024] The above technology only realizes early warning analysis by setting thresholds for fan speed frequency, fan motor speed value or current peak value. However, due to the influence of many factors such as the operating status of the fan in the energy storage container, the fan connection status, fan power status, fan speed, and whether the fan vents are unobstructed, when the battery cluster fan is abnormal, it will cause the temperature distribution of the battery cells in the cluster to change significantly. The above technology is difficult to accurately and effectively identify energy storage battery cluster fan abnormalities by only analyzing the relevant parameters of the fan body. Summary of the Invention

[0025] The purpose of this invention is to achieve real-time early warning of the air conditioning system and battery cluster fan system in the container. A method for early warning of abnormalities in the thermal management system of an air-cooled lithium-ion battery energy storage container is proposed. This method realizes early warning of the air conditioning system and the battery cluster fan system:

[0026] Air conditioning system early warning: To provide early warning of abnormal cooling and heating functions of the air conditioner due to air conditioning condensation leakage, compressor failure, etc., the present invention analyzes the distribution and changing trends of the inlet and outlet air temperatures of each air conditioner in the energy storage container, and proposes an abnormality early warning method for the air conditioning system of an air-cooled lithium-ion battery energy storage container. This method can accurately identify early abnormalities of the air conditioning system under the coordinated control of multiple air conditioners in the container and issue graded early warning information, effectively avoiding problems such as battery cell overheating or uneven temperature distribution caused by long-term abnormal cooling / heating functions of the air conditioner, thereby improving the operational reliability of the energy storage container.

[0027] Battery cluster fan system early warning: In order to provide early warning of battery cluster fan malfunction caused by battery cluster fan failure and stoppage, abnormal fan power wiring, blocked fan vents, etc., the present invention analyzes the temperature distribution pattern of the battery cells in each battery cluster of the container, combines LOF and DBSCAN density clustering algorithms, and proposes a battery cluster fan abnormality early warning method for air-cooled lithium-ion battery energy storage containers. This method can accurately identify early abnormalities of battery cluster fans in the container and issue graded early warning information, effectively avoiding problems such as battery cell overheating or uneven temperature distribution caused by fan failure and stoppage, abnormal fan wiring, etc., thereby improving the operational reliability of the energy storage container.

[0028] Specifically, to achieve the above objectives, the present invention is implemented through the following technical solutions.

[0029] The present invention proposes an abnormality warning method for a thermal management system of an air-cooled lithium-ion battery energy storage container. The thermal management system of the energy storage container includes an air conditioning system and a battery cluster fan system. The method includes:

[0030] Step 1) Extracting and preprocessing data from the air conditioning system; Based on the preprocessed data, establish probability density distribution models for the temperature differences between the inlet and return air in the cooling and heating states of each air conditioner using the kernel density estimation (KDE) model; Calculate the cooling and heating health indicators of each air conditioner based on the probability density distribution model, and then determine abnormal risk and issue warnings for the air conditioning system;

[0031] Step 2) extracting data from the battery cluster fan system and performing preprocessing; based on the preprocessed data, using the DBSCAN density clustering algorithm to perform outlier analysis, obtain outlier identification results, and then perform abnormal risk assessment and early warning for the battery cluster fan system.

[0032] As one of the improvements to the above technical solution, step 1) specifically includes:

[0033] Step 1-1) Extract key data of each air conditioning system within a single day at a set sampling rate, including: time, air inlet temperature of each air conditioner, return air temperature of each air conditioner, cooling status of each air conditioner, and electric heating status of each air conditioner. Preprocess the data, including: removing data points where the air inlet / return air temperature is not within the set temperature range and removing null values ​​in the original data;

[0034] Step 1-2) determines whether the amount of pre-processed data meets the set warning requirement data volume. If not, no warning is required; if so, proceed to step 1-3);

[0035] Steps 1-3) Based on the preprocessed data, the daily inlet and return air temperature differences for each air conditioner in both cooling and heating modes are calculated. A probability density distribution model for the inlet and return air temperature differences for each air conditioner in both cooling and heating modes is established using a kernel density estimation (KDE) model. The KDE model uses an adaptive KDE algorithm to automatically select bandwidth.

[0036] Steps 1-4) Calculate the cooling and heating health indicators of each air conditioner based on the established probability density distribution model of the inlet and return air temperature differences in the cooling and heating states of each air conditioner;

[0037] Steps 1-5) Determine whether the cooling and heating health indicators of each air conditioner exceed the abnormality judgment threshold; if exceeded, perform abnormal risk judgment on the air conditioning system; if not exceeded, perform linear fitting based on the cooling and heating health indicators of each air conditioner over the past few days, and obtain the slope coef of the cooling and heating health indicators respectively. 制冷 、coef 制热 , and judge the abnormal risk of the air-conditioning system based on the slope, and then give corresponding early warnings.

[0038] As one of the improvements to the above technical solution, in step 1-3), the expression of the kernel density estimation KDE model f(y) is:

[0039]

[0040] Where h is the bandwidth; K(·) is the Gaussian kernel function; n is the number of sample points of the air conditioning inlet and return air temperature difference under cooling or heating conditions actually collected in a single day; y a is the ath sample point, that is, the inlet and return air temperature difference of a certain air conditioner in the cooling or heating state at a certain moment; y is the independent variable of the kernel density estimation KDE model;

[0041] The expression of Gaussian kernel function K(·) is:

[0042]

[0043] As one of the improvements to the above technical solution, in steps 1-4), the calculation formula for the cooling and heating health index of each air conditioner is:

[0044]

[0045]

[0046] Among them, HLI 制冷ACi 、HLI 制热ACi They are the health indicators of air conditioning cooling and heating respectively; max_ΔT 制冷ACiis the inlet and return air temperature difference value corresponding to the maximum probability density in the probability density distribution model of the inlet and return air temperature difference of the i-th air conditioner in the cooling state; max_ΔT 制热ACi is the inlet and return air temperature difference value corresponding to the maximum probability density in the probability density distribution model of the inlet and return air temperature difference of the i-th air conditioner in the heating state; max(·) represents the maximum function; N represents the total number of air conditioners.

[0047] As one of the improvements to the above technical solution, in step 1-5), the linear fitting expressions based on the cooling and heating health indicators of each air conditioner over the past few days are:

[0048] HLI 制冷ACi =coef 制冷 *x+b 制冷

[0049] HLI 制热ACi =coef 制热 *x+b 制热

[0050] Among them, x is the number of data points in the past few days; coef 制冷 is the slope of the linear fitting based on the cooling health index of each air conditioner in the past few days; coef 制热 is the slope of the linear fitting based on the heating health index of each air conditioner in the past few days; b 制冷 is the bias term of the linear fitting model of refrigeration health; b 制热 is the bias term of the linear fitting model of heating health.

[0051] As one of the improvements to the above technical solution, in step 1-5), abnormal risk judgment of the air-conditioning system includes:

[0052] If HLI 制冷ACi Exceeding the abnormality threshold, or HLI 制冷ACi When the abnormality judgment threshold is not exceeded, coef 制冷 <0, it is determined that the corresponding air conditioner No. i has a high-risk cooling anomaly;

[0053] If HLI 制冷ACi When the abnormality judgment threshold is not exceeded, coef 制冷 ≥0, it is determined that the corresponding air conditioner No. i has a low-risk cooling anomaly;

[0054] If HLI 制热ACi Exceeding the abnormality threshold, or HLI 制热ACi When the abnormality judgment threshold is exceeded, coef 制热 <0, it is determined that the corresponding air conditioner No. i has a high-risk heating anomaly;

[0055] If HLI制热ACi When the abnormality judgment threshold is not exceeded, coef 制热 ≥0, it is determined that the corresponding air conditioner No. i has a low-risk heating anomaly.

[0056] As one of the improvements to the above technical solution, step 2) specifically includes:

[0057] Step 2-1) Extract key data from the battery cluster fan system for a single day at a set sampling rate, including time, cooling status of each air conditioner, heating status of each air conditioner, battery cell temperature within the battery cluster, and fan relay status. Preprocess the data by removing data points where the cell temperature is outside the set temperature range and removing null values ​​from the raw data.

[0058] Step 2-2) Determine whether the amount of pre-processed data meets the set warning requirement data volume. If not, no warning is required; if so, proceed to step 2-3);

[0059] Step 2-3) Based on the preprocessed data, respectively calculate and extract the average temperature within each battery cluster and the set quantile of the temperature within the cluster, as well as the temperature standard deviation and coefficient of variation indicators under the two operating conditions of the air conditioning cooling state with the fan on and the air conditioning heating state with the fan on during the day, and perform outlier analysis on the average temperature and set quantile of each battery cluster based on the DBSCAN density clustering algorithm to obtain outlier identification results for the temperature within each battery cluster. Also, perform outlier analysis on the standard deviation and coefficient of variation of the temperature within each battery cluster based on the DBSCAN density clustering algorithm to obtain outlier identification results for the temperature dispersion indicators within each battery cluster;

[0060] Step 2-4) Based on the temperature outlier identification results between battery clusters and within each battery cluster and the temperature dispersion index outlier identification results, the abnormal risk of the battery cluster fan system is judged, and a corresponding warning is issued.

[0061] As one of the improvements to the above technical solution, in step 2-3), outlier analysis is performed on the temperature mean and set quantile within each battery cluster based on the DBSCAN density clustering algorithm to obtain temperature outlier identification results within each battery cluster, specifically including the following steps:

[0062] ① Input sample set D={x1,x2,...,x j ...,x m}, where x j represents the average intra-cluster temperature of battery cluster j within the day, sets the quantile intra-cluster temperature, sets the neighborhood parameters (ε, MinPts), and adopts the Euclidean distance as the sample distance metric; m is the total number of battery clusters in the container;

[0063] ② Initialize the core object collection Initialize the number of clusters k = 0, initialize the unvisited sample set Γ = D, and initialize cluster division

[0064] ③For j=1,2,...,m, find the cluster core object according to the following steps:

[0065] a. Find the sample x by distance measurement j The ε neighborhood subsample set N ε (x j );

[0066] b. Set the number of samples in the subsample set to satisfy |N ε (x j )|≥MinPts sample x j Add core object set Ω;

[0067] c. Repeat steps a and b to continuously update the core object sample set Ω;

[0068] ④ If the core object collection The algorithm ends and the result is output; if the core object set In the core object set Ω, randomly select a core object o and initialize the current core object queue Ω cur ={o}, initialize the category number k to k+1, and initialize the current cluster sample set Ω k ={o}, update the unvisited set Γ to Γ-{o};

[0069] ⑤ If the current cluster core object queue The current cluster C k After generation, update the cluster partition C={C1,C2,...,C k}, update the core object collection, update the current core object collection to the original collection and C k The intersection of , go to step ④; if the current cluster core object queue Update the current core object to the original collection and C k The intersection of

[0070] ⑥ In the current cluster core object queue Ω cur Take out a core object o' and find all the ε-neighborhood subset sample sets N by neighborhood distance ε (o'), let the set Δ=N ε (o')∩Γ, update the current cluster sample set C k C k ∪Δ, update the unvisited sample set Γ to Γ-Δ, update Ω cur Ω cur ∪(Δ∩Ω)-o', go to step ⑤;

[0071] ⑦The output result is: cluster partition C={C1,C2,...,C h ,...,C k}, C h represents the hth cluster in the cluster, k is the final total number of clusters; data points that do not belong to any cluster are defined as outliers.

[0072] As one of the improvements to the above technical solution, in step 2-4), risk assessment of the battery cluster fan system includes:

[0073] If two types of outliers exist simultaneously within a battery cluster, the fan corresponding to the battery cluster is considered high risk;

[0074] If there is only one type of outlier in the battery cluster, the fan corresponding to the battery cluster is of low risk.

[0075] The beneficial effects brought about by the technical solution of the present invention are:

[0076] 1. This invention provides an abnormal warning for the thermal management system of an air-cooled lithium-ion battery energy storage container (including the air conditioning system and battery cluster fan system) by analyzing the distribution and changing trends of the inlet and outlet air temperatures of each air conditioner in the energy storage container and the temperature distribution patterns of the battery cells in each battery cluster.

[0077] 2. The method of the present invention can accurately identify early abnormalities of the air-conditioning system and battery cluster fans in the container and issue graded warning information, effectively avoiding problems such as battery cell overheating or uneven temperature distribution caused by long-term abnormal functions of the thermal management system, and improving the operational reliability of the energy storage container. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a flowchart of the overall process of the method of the present invention;

[0079] Figure 2 This is a schematic diagram of the abnormal warning process of the air-cooled lithium-ion battery energy storage container air-conditioning system;

[0080] Figure 3 This is a schematic diagram of the abnormal warning process for the battery cluster fan of an air-cooled lithium-ion battery energy storage container. DETAILED DESCRIPTION

[0081] The technical solution provided by the present invention is further illustrated below with reference to embodiments.

[0082] like Figure 1 As shown in FIG, it is a flowchart of the overall process of the embodiment of the method of the present invention; specifically, as Figure 2 and Figure 3 The figures show the abnormal warning process flow diagrams of the air-cooled lithium-ion battery energy storage container air conditioning system and the battery cluster fan system according to embodiments of the present invention.

[0083] 1. Air conditioning system warning

[0084] Figure 2 This is a schematic diagram of the abnormality warning process for the air-cooled lithium-ion battery energy storage container air conditioning system. The overall design concept of this technology is: theoretically, the environmental conditions near the installation locations of each air conditioner in the container are similar (temperature, humidity, etc.). If each air conditioner is operating normally, the cooling / heating output of each air conditioner within a single day should be relatively small. Therefore, a statistical distribution model of the inlet and outlet air temperature difference ΔT in the cooling and heating states of the air conditioner within a single day is established. By comparing the statistical distribution differences of each air conditioner within a single day, abnormalities can be identified. The specific warning process is as follows:

[0085] 1) Single-day air conditioning data extraction: For the energy storage container to be analyzed, key data of each air conditioning system in the container within a single day is extracted (sampling rate is 1 minute), including time, air conditioning inlet temperature, air conditioning return temperature, air conditioning compressor status, and air conditioning electric heating status. Data preprocessing is performed to eliminate data points with air conditioning inlet / return air temperatures outside the range of [-35°C to 65°C] and to remove null values ​​in the original data.

[0086] 2) Data volume judgment: Determine whether the amount of pre-processed data meets the requirements of early warning modeling (>900 data points after elimination). If so, proceed to the next step;

[0087] 3) Statistical modeling for identifying abnormal cooling function: Extract the return air temperature and inlet air temperature data of each air conditioner compressor state ==2 (i.e., compressor working state) on a single day, and calculate the ΔT of each air conditioner on a single day 制冷ACi = Inlet air temperature ACi -Return air temperature ACi (ΔT 制冷ACi represents the temperature difference between the inlet and return air of the air conditioner No. i in the container under refrigeration state), and the kernel density estimation (KDE) is used to establish the ΔT 制冷ACi The probability density distribution model of , in which the KDE model uses the adaptive KDE algorithm to realize automatic bandwidth selection:

[0088]

[0089] Where: f(x) is the expression of the kernel density estimation KDE model, h is the bandwidth; K(·) is the kernel function, and this application uses the Gaussian kernel function; n is the number of sample points of the air conditioning inlet and return air temperature difference under cooling or heating conditions actually collected in a single day; x i is the i-th sample point, that is, the inlet and return air temperature difference of a certain air conditioner in the cooling or heating state at a certain moment. The expression of the Gaussian kernel function is:

[0090]

[0091] 4) Statistical modeling for identifying abnormal heating function: Extract the return air temperature and inlet air temperature data of each air conditioner in the electric heating state = = 1 (i.e., electric heating working state) on a single day, and calculate the ΔT of each air conditioner on a single day. 制热ACi =Return air temperature ACi -Inlet air temperature ACi (ΔT 制热ACi represents the return air and inlet air temperature difference of the air conditioner No. i in the container under heating state), and also use formulas (1) and (2) to establish each ΔT based on KDE ACi The probability density distribution model of .

[0092] 5) Calculate the cooling and heating health indicators of each air conditioner as follows:

[0093] HLI 制冷ACi =max_ΔT 制冷ACi / max(max_ΔT 制冷AC0 ,max_ΔT 制冷AC1 ,...,max_ΔT 制冷ACi ) (3)

[0094] HLI 制热ACi =max_ΔT 制热ACi / max(max_ΔT 制热AC0 ,max_ΔT 制热AC1 ,...,max_ΔT 制热ACi ) (4)

[0095] Where: HLI 制冷ACi 、HLI 制热ACi They are the health indicators of air conditioning cooling and heating respectively; max_ΔT 制冷ACi is the ΔT of the i-th air conditioner 制冷ACi The temperature difference between the inlet and return air corresponding to the maximum probability density in the KDE probability density model; max_ΔT 制热ACi is the ΔT of the i-th air conditioner 制热ACi The return air and inlet air temperature difference corresponding to the maximum probability density in the KDE probability density model.

[0096] 6) Determine HLI 制冷ACi 、HLI 制热ACi Whether it exceeds the abnormality judgment threshold; if it exceeds the judgment threshold, the abnormal risk of the air conditioner is judged; if it does not exceed the judgment threshold, it continues to be based on the HLI of the past five days 制冷ACi 、HLI 制热ACi Linear fitting (HLI 制冷ACi =coef 制冷 *x+b 制冷 、HLI 制热ACi =coef 制热 *x+b制热 , where x is the number of historical five-day data points), and the slope coef 制冷 、coef 制热 , and judge the abnormal risk of air conditioning based on the slope.

[0097] 7) Conduct air conditioning risk assessment:

[0098] ① Such as coef 制冷 Slope < 0 or HLI 制冷ACi If the abnormality determination threshold (0.45) is exceeded, the air conditioner is judged to have a high-risk cooling abnormality;

[0099] ② Such as coef 制冷 If the slope is ≥ 0, the air conditioner is judged to have a low-risk cooling anomaly;

[0100] ③ Such as coef 制热 Slope < 0 or HLI 制热ACi If the abnormality determination threshold (0.45) is exceeded, the air conditioner is judged to have a high-risk heating abnormality;

[0101] ④ Such as coef 制热 If the slope is ≥ 0, the air conditioner is judged to have a low-risk heating anomaly.

[0102] 2. Battery cluster fan system warning

[0103] Figure 3 This is a schematic diagram of the battery cluster fan anomaly warning process for air-cooled lithium-ion battery energy storage containers. The overall design concept of this technology is: Under normal circumstances, when the battery cluster fans and energy storage air conditioners are operating, the overall temperature consistency of the battery cells in each battery cluster in the container is relatively good. When a battery cluster fan is abnormal, the discreteness of the temperature distribution of the battery cells in the cluster increases, and the overall temperature of the battery cluster cells becomes significantly outliers. By using a clustering algorithm to identify the discreteness of the temperature within the cluster and the outlier characteristics of the temperature between clusters, it is possible to identify battery cluster fan anomalies. The specific warning process is as follows:

[0104] 1) Single-day data extraction: For the energy storage container to be analyzed, key data within the container within a single day is extracted (with a sampling rate of 1 minute), including time, status of each air conditioner compressor, status of each air conditioner electric heater, battery cell temperature within the battery cluster, and fan relay status. Data preprocessing is then performed to remove data points where the battery cell temperature is not within the range of [-35°C to 65°C] and to remove null values ​​in the original data.

[0105] 2) Data volume judgment: Determine whether the amount of pre-processed data meets the requirements of early warning modeling (>900 data points after elimination). If so, proceed to the next step;

[0106] 3) Identification of Outliers in Cell Temperature Clusters: The average temperature and 80% quantile of each cluster's temperature are extracted for both cooling and heating conditions. Outlier analysis is performed on these average and 80% quantiles based on DBSCAN density clustering to identify outlier battery clusters. The detailed process is as follows. (A battery cluster typically contains hundreds of cells. The quantile of the temperature within a cluster refers to the quantile of these cell temperatures (assuming there are 100 cell temperature values ​​within a cluster, the 80% quantile is the 80th value after these 100 temperature values ​​are arranged from smallest to largest. This quantile can be set as an adjustable variable based on actual needs.)

[0107] ① Input: Sample set D = {x1, x2, ..., x m}, where x1 represents the average temperature of battery cluster 1 within the day, the 80% quantile temperature of the cluster, the neighborhood parameters (ε, MinPts), and the sample distance metric adopts the Euclidean distance;

[0108] ②Output: cluster partition C.

[0109] 3.1 Initialize the core object collection Initialize the number of clusters k = 0, initialize the unvisited sample set Γ = D, and divide the clusters

[0110] 3.2 For j = 1, 2, ..., m, find the cluster core object according to the following steps:

[0111] a. Find the neighborhood subsample set N of sample xj by distance measurement ε (x j );

[0112] b. The number of samples in the subsample set satisfies |N ε (x j )|≥MinPts, the sample x j Add core object sample set: Ω=Ω∪{x j};

[0113] 3.3 Core Object Collection The algorithm ends, otherwise it goes to step 2.4;

[0114] 3.4 Randomly select a core object o from the core object set Ω and initialize the current core object queue Ω cur ={o}, initialize the category number k=k+1, initialize the current cluster sample set Ω k ={o}, update the unvisited set Γ=Γ-{o};

[0115] 3.5 If the current cluster core object The current cluster C k After generation, update the cluster partition C={C1,C2,...,C k}, update the core object set Ω = Ω - C k , go to step 3.3. Otherwise update the core object set Ω = Ω - C k ;

[0116] 3.6 In the current cluster core object queue Ω cur Take out a core object o' and find all the ε-neighborhood subset sample sets N by neighborhood distance ε (o'), let Δ=N ε (o')∩Γ, update the current cluster sample set C k =C k ∪Δ, update the unvisited sample set Γ=Γ-Δ, update Ω cur =Ω cur ∪(Δ∩Ω)-o', transfer to 3.5.

[0117] 3.7 The output result is: cluster partition C={C1,C2,...,C k}, data points that are not identified as clusters are defined as outliers.

[0118] 4) Outlier identification of inter-cluster battery cell temperature dispersion indicators: The standard deviation and coefficient of variation of the temperature within each cluster under two working conditions: air conditioning cooling state with fan on and air conditioning heating state with fan on, are extracted respectively. Outlier analysis is performed based on the DBSCAN clustering algorithm to identify outlier battery clusters.

[0119] 5) Battery cluster fan classification warning: Summarizes the outlier identification results of inter-cluster battery cell temperature and inter-cluster battery cell temperature dispersion index. For battery clusters with both types of outliers, the corresponding fans are at high risk. For battery clusters with only one type of outlier, the corresponding fans are at low risk.

[0120] As can be seen from the detailed description of the present invention above, the present invention implements real-time abnormality warnings for the air conditioning system and battery cluster fan system in an air-cooled lithium-ion battery energy storage container by analyzing the distribution and changing trends of the inlet and outlet air temperatures of each air conditioner in the energy storage container and the temperature distribution patterns of the battery cells in each battery cluster.

[0121] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention and are intended to be encompassed by the claims of the present invention.

Claims

1. A method for anomaly warning of a thermal management system of an air-cooled lithium-ion battery energy storage container, wherein the thermal management system of the energy storage container includes an air conditioning system and a battery cluster fan system. The method comprises: Step 1) Extract the data of the air conditioning system and perform preprocessing; According to the preprocessed data, the probability density distribution model of the inlet and return air temperature difference under the cooling and heating states of each air conditioner is established based on the kernel density estimation KDE model; Based on the probability density distribution model, the cooling and heating health indicators of each air conditioner are calculated separately, and then the abnormal risk of the air conditioning system is judged and early warning is issued; Step 2) Extract data from the battery cluster fan system and perform preprocessing. Based on the preprocessed data, use the DBSCAN density clustering algorithm to perform outlier analysis and obtain outlier identification results. This allows for abnormal risk assessment and early warning of the battery cluster fan system. The step 2) includes: Step 2-1): Calculate and extract the average temperature within each battery cluster and the set quantile of the temperature within the cluster, as well as the temperature standard deviation and coefficient of variation indicators under the two operating conditions of the air conditioning cooling state with the fan on and the air conditioning heating state with the fan on during the day, and perform outlier analysis on the average temperature and set quantile of each battery cluster based on the DBSCAN density clustering algorithm to obtain the temperature outlier identification results within each battery cluster. Perform outlier analysis on the standard deviation and coefficient of variation of the temperature within each battery cluster based on the DBSCAN density clustering algorithm to obtain the temperature dispersion indicator outlier identification results within each battery cluster; Step 2-2) Based on the temperature outlier identification results between battery clusters and within each battery cluster and the temperature dispersion index outlier identification results, the abnormal risk of the battery cluster fan system is judged, and a corresponding warning is issued.

2. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 1 is characterized in that: The step 1) specifically includes: Step 1-1) Extract key data from each air conditioning system within a single day at a set sampling rate, including time, air inlet temperature, return air temperature, cooling status, and electric heating status. Preprocess the data, including removing data points where the air inlet / return air temperature is outside the set temperature range and removing null values ​​from the raw data. Step 1-2) Determine whether the amount of pre-processed data meets the set warning requirement. If not, no warning is required; if so, proceed to step 1-3); Steps 1-3) Based on the preprocessed data, calculate the daily inlet and return air temperature differences for each air conditioner in both cooling and heating modes. Then, establish probability density distribution models for these differences using a kernel density estimation (KDE) model. The KDE model uses an adaptive KDE algorithm to automatically select bandwidth. Steps 1-4) Calculate the cooling and heating health indicators of each air conditioner based on the established probability density distribution model of the temperature difference between the inlet and return air in the cooling and heating states of each air conditioner; Steps 1-5) Determine whether the cooling and heating health indicators of each air conditioner exceed the abnormality judgment threshold; if exceeded, the air conditioning system is judged to have abnormal risk; if not, a linear fit is performed based on the cooling and heating health indicators of each air conditioner over the past few days to obtain the slope coef of the cooling and heating health indicators respectively. 制冷 、coef 制热 , and judge the abnormal risk of the air-conditioning system based on the slope, and then give corresponding early warnings.

3. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 2 is characterized in that: In steps 1-3), the kernel density estimation KDE model The expression is: ； in h is bandwidth; K (·) is the Gaussian kernel function; n The number of sample points of the air-conditioning inlet and return air temperature difference actually collected in a single day under cooling or heating status; y a For the a A sample point, that is, the inlet and return air temperature difference of a certain air conditioner in the cooling or heating state at a certain moment; y It is the independent variable of the kernel density estimation KDE model; Gaussian kernel function The expression is: 。 4. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 2 is characterized in that: In steps 1-4), the calculation formula for the cooling and heating health index of each air conditioner is: ； ； Among them, HLI 制冷ACi 、HLI 制热ACi They are the health indicators of air conditioning cooling and heating respectively; max_ΔT 制冷ACi is the inlet and return air temperature difference value corresponding to the maximum probability density in the probability density distribution model of the inlet and return air temperature difference of the i-th air conditioner in the cooling state; max_ΔT 制热ACi is the inlet and return air temperature difference value corresponding to the maximum probability density in the probability density distribution model of the inlet and return air temperature difference of the i-th air conditioner in the heating state; max(·) represents the maximum function; N represents the total number of air conditioners.

5. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 4 is characterized in that: In the steps 1-5), the linear fitting expressions based on the cooling and heating health indicators of each air conditioner over the past few days are: HLI 制冷ACi = coef 制冷* x + b 制冷； HLI 制热ACi = coef 制热* x + b 制热； in, x is the number of historical data points for several days; coef 制冷 is the slope of the linear fitting based on the cooling health index of each air conditioner in the past few days; coef 制热 is the slope of the linear fitting based on the heating health index of each air conditioner over several days of history; b 制冷 is the bias term of the linear fitting model of refrigeration health; b 制热 is the bias term of the linear fitting model of heating health.

6. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 5 is characterized in that: In the steps 1-5), abnormal risk assessment of the air conditioning system is performed, including: If HLI 制冷ACi Exceeding the abnormality threshold, or HLI 制冷ACi When the abnormality judgment threshold is not exceeded, coef 制冷 <0, it is determined that the corresponding air conditioner No. i has a high-risk cooling anomaly; If HLI 制冷ACi When the abnormality judgment threshold is not exceeded, coef 制冷 ≥0, it is determined that the corresponding air conditioner No. i has a low-risk cooling anomaly; If HLI 制热ACi Exceeding the abnormality threshold, or HLI 制热ACi When the abnormality judgment threshold is not exceeded, coef 制热 <0, it is determined that the corresponding air conditioner No. i has a high-risk heating anomaly; If HLI 制热ACi When the abnormality judgment threshold is not exceeded, coef 制热 ≥0, it is determined that the corresponding air conditioner No. i has a low-risk heating anomaly.

7. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 1 is characterized in that: The step 2) further includes: Step 2-1-1) Extract key data from the battery cluster fan system for a single day at the set sampling rate, including time, cooling status of each air conditioner, heating status of each air conditioner, battery cell temperature within the battery cluster, and fan relay status. Preprocess the data by removing data points where the cell temperature is outside the set temperature range and removing null values ​​from the original data. Step 2-1-2) determines whether the amount of pre-processed data meets the set warning requirement data amount. If not, no warning is required; if so, proceed to step 2-1). In step 2-1), step 2-1) is performed according to the pre-processed data.

8. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 1 is characterized in that: In step 2-1), outlier analysis is performed on the temperature mean and set quantile within each battery cluster based on the DBSCAN density clustering algorithm to obtain temperature outlier identification results within each battery cluster, which specifically includes the following steps: ① Input sample set D = { x 1, x 2, ..., x j ..., x m },in x j Representative day j The average temperature within the cluster of battery cluster No. is set, the quantile temperature within the cluster is set, the neighborhood parameters ε and MinPts are set, and the sample distance measurement method adopts the Euclidean distance; m is the total number of battery clusters in the container; ② Initialize the core object set Ω = Ø and the number of clusters k = 0, initialize the unvisited sample set Γ = D, initialize the cluster partition C = Ø; ③For j = 1,2,..., m , follow the steps below to find the cluster core objects: a. Find samples by distance measurement x j The ɛ neighborhood subsample set N ε ( x j ); b. Set the number of samples in the subsample set to satisfy |N ε ( x j )|≥MinPts samples x j Add core object set Ω; c. Repeat steps a and b to continuously update the core object sample set Ω; ④ If the core object set Ω≠Ø, the algorithm ends and outputs the result; if the core object set Ω=Ø, randomly select a core object in the core object set Ω o , initialize the current core object queue Ω cur ={ o }, initialize the category number k for k +1, initialize the current cluster sample set Ω k = { o }, update the unvisited set Γ to Γ- { o }; ⑤ If the current cluster core object queue Ω cur =Ø, then the current cluster C k After generation, update the cluster partition C = {C1, C2,..., C k }, update the core object collection, update the current core object collection to the original collection and C k The intersection of Ω and Ω, go to step ④; if the current cluster core object queue Ω cur =Ø, update the current core object to the original collection and C k The intersection of ⑥ In the current cluster core object queue Ω cur Take out a core object o' and find all the Neighborhood subset sample set N ε (o'), let the set Δ = N ε (o')∩Γ, update the current cluster sample set C k C k ∪ Δ, update the unvisited sample set Γ to Γ- Δ, update Ω cur Ω cur ∪ (Δ∩Ω) - o', go to step ⑤; ⑦The output result is: cluster partition C = {C1, C2, ...,C h ,..., C k }, C h Indicates the first h clusters, k is the final total number of clusters; data points that do not belong to any cluster are defined as outliers.

9. The abnormal warning method for the thermal management system of an air-cooled lithium-ion battery energy storage container according to claim 1 is characterized in that: In step 2-2), risk assessment of the battery cluster fan system includes: If two types of outliers exist simultaneously within a battery cluster, the fan corresponding to the battery cluster is considered high risk; If there is only one type of outlier in the battery cluster, the fan corresponding to the battery cluster is of low risk.

Images