An energy management method based on data rule mining

Abstract

The application discloses a kind of energy management methods based on data rule mining, it is related to train energy management method technical field, and its technical solution points are: the application proposes a kind of ground type energy storage system energy management method based on data rule mining, according to the working condition identification method based on K-means clustering algorithm, the overall working condition of train in power supply section can be accurately identified, and by working condition identification result realizes the dynamic adjustment of energy storage system charge-discharge voltage threshold value.Compared with the traditional fixed threshold or energy management method based on no-load voltage identification, the method can effectively absorb the residual regenerative braking energy of train, and fully utilize the absorbed regenerative braking energy in traction working condition, reduce the output energy of substation to achieve the goal of energy saving and consumption reduction.

Description

Technical Field

[0001] This invention relates to the field of train energy management methods, and more specifically, to an energy management method based on data pattern mining. Background Technology

[0002] Urban rail trains widely employ regenerative braking technology. During braking, the traction motor operates as a generator, generating electricity for auxiliary power supply to the train itself, with the remainder fed back into the traction power grid for interaction with nearby trains. Due to the unidirectional conductivity of the substation traction rectifier units, if the regenerative braking energy generated by the train cannot be fully absorbed, the traction grid voltage will rise. When this voltage reaches the activation voltage of the train's braking resistor, the resistor activates to dissipate energy. If this still fails to effectively control the grid voltage, regenerative braking may fail as the voltage continues to rise, resulting in the activation of the air brake. If this occurs, the train's regenerative braking energy cannot be fully utilized.

[0003] The urban rail ground-based energy storage system connects the energy storage device to the traction network via a bidirectional converter. Through effective energy management (EMS) control, it absorbs the remaining regenerative braking energy of the braking train. By setting the charging / discharging voltage thresholds of the energy storage system, the energy is released when the train is in traction. This effectively utilizes the train's regenerative braking energy while suppressing grid voltage fluctuations, significantly reducing the energy consumption of the urban rail traction power supply system.

[0004] Currently, urban rail ground-based energy storage systems mainly employ fixed-threshold or dynamic-threshold energy management methods based on traction grid voltage. This is because it's generally believed that changes in traction grid voltage reflect train operating conditions—that is, the grid voltage decreases when the train is traction-driven and increases when the train is braking. Therefore, the operating state of the energy storage system is determined by setting charging and discharging thresholds and comparing them with the traction grid voltage. This method is currently the most widely used in practice. However, when considering the fluctuations in the no-load voltage of substations and the differences in no-load voltage between multiple substations, relying solely on the traction grid voltage is insufficient to accurately reflect the train operating conditions within a section. In such cases, energy storage systems controlled by this energy management method struggle to achieve good energy-saving performance, leading to abnormal situations such as "charging without discharging" or "discharging without charging," thus failing to achieve optimal energy-saving effects.

[0005] Therefore, this invention proposes an energy management method based on data pattern mining, which takes into account the fluctuation of the no-load voltage of the substation itself and the difference in no-load voltage between multiple substations. The charging / discharging voltage threshold of the energy storage system will be adaptively adjusted according to the train operating condition information obtained through data mining. Summary of the Invention

[0006] The purpose of this invention is to solve the above-mentioned problems and provide an energy management method based on data pattern mining.

[0007] The above-mentioned technical objective of the present invention is achieved through the following technical solution: an energy management method based on data pattern mining, including a train condition identification method based on K-means clustering algorithm and an energy management method. The train condition identification method based on K-means clustering algorithm includes offline processing and online processing. Offline, the K-means clustering algorithm is used to mine the patterns of the collected historical data to establish the relationship between train conditions and ground features. Online, only relevant ground information needs to be collected in real time, and the train conditions within the section are identified online based on the mining results of historical data.

[0008] The clusterer is trained based on the novel train condition identification method, and the condition is identified by the online collected ground data. The condition is determined and the condition coefficient M is output.

[0009] The energy management method, as a higher-level control method for converter control, sets the charging / discharging voltage threshold of the energy storage system to ensure that the system can operate in the discharging and charging states during train traction and braking conditions.

[0010] The present invention is further configured such that: the working condition identification is performed on the ground data collected online, including the Euclidean distance calculation formula between the newly collected online sample X and the mean vector obtained by the trained clusterer:

[0011]

[0012] Where dis_tra is the distance between the new sample and the mean vector representing the traction condition, and dis_bra is the distance between the new sample and the mean vector representing the braking condition. Based on the relationship between these two distances, it is determined whether the new sample belongs to the traction condition or the braking condition, and the final condition coefficient M is output. Under the traction condition, M = -1, and under the braking condition, M = 1, as shown in equation (4):

[0013]

[0014] The present invention is further configured to: adjust the reference value U of the charging and discharging voltage threshold of the energy storage system according to the output operating condition coefficient M and the current operating state of the energy storage system. ref To ensure the relative stability of the energy storage system's operating state, i.e., to prevent frequent switching between charging and discharging modes, if the energy storage system is already in discharging mode during traction or charging mode during braking, then U ref No adjustment is made; otherwise, the baseline value of the threshold is adjusted accordingly according to equation (5), ΔUref The adjustment amount of each benchmark value can be flexibly adjusted according to the actual project needs;

[0015] U ref =U ref -M·ΔU ref (5)

[0016] The reference value U for the charge / discharge threshold is obtained by calculation using the above formula. ref Then, the charging and discharging voltage threshold of the energy storage system is calculated according to equation (6), thereby realizing the dynamic adjustment of the threshold. In the equation, Δu c The adjustment amount for the charging voltage threshold, Δu d The adjustment range of the discharge voltage threshold can be flexibly adjusted according to actual engineering needs. At the same time, in order to avoid the substation charging the energy storage system itself due to the charging threshold being too low, and to avoid the grid voltage rising rapidly when the energy storage system discharges due to the discharge threshold being too high, causing the braking resistor on the train to start, a limit is set on the adjustment range of the reference value.

[0017]

[0018] In summary, the present invention has the following beneficial effects:

[0019] This invention proposes an energy management method for ground-based energy storage systems based on data pattern mining. According to the operating condition identification method based on the K-means clustering algorithm proposed in this invention, the overall operating condition of trains within the power supply section can be accurately identified, and the charging and discharging voltage thresholds of the energy storage system can be dynamically adjusted based on the operating condition identification results.

[0020] Compared to traditional energy management methods based on fixed thresholds or no-load voltage identification, this method can effectively absorb the train's remaining regenerative braking energy and make full use of the absorbed regenerative braking energy during traction conditions, thereby reducing the substation's output energy and achieving the goal of energy saving and consumption reduction. Attached Figure Description

[0021] Figure 1 This is a control architecture diagram of a ground-based energy storage system in an embodiment of the present invention;

[0022] Figure 2 This is a schematic diagram of the 24-pulse rectifier unit in the traction substation of this invention.

[0023] Figure 3 This is a flowchart of the K-means clustering algorithm in an embodiment of the present invention;

[0024] Figure 4 This is a clustering result diagram in an embodiment of the present invention;

[0025] Figure 5This is a comparison chart of the clustering results and the sum of the power of the trains in the interval in an embodiment of the present invention;

[0026] Figure 6 This is a control flowchart of the energy management method based on data pattern mining in an embodiment of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be described in further detail below with reference to the embodiments and accompanying drawings. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the embodiments.

[0029] Example:

[0030] 1. Control Architecture of Urban Rail Ground-Based Energy Storage System

[0031] Figure 1 The diagram shows the control architecture of a ground-mounted energy storage system for urban rail transit. The Energy Management System (EMS), as a higher-level control method above the converter control, sets the charging / discharging voltage thresholds for the energy storage system to ensure that the system operates in both discharging and charging states during train traction and braking. Therefore, the train's status has a crucial impact on the design of the energy storage system's charging and discharging control and the EMS method. Accurately acquiring the train's traction or braking conditions ensures that the energy storage system absorbs as much of the train's remaining regenerative braking energy as possible and releases it during train traction, achieving better energy-saving effects.

[0032] 2. Working Condition Identification Method Based on K-means Clustering Algorithm

[0033] Currently, most energy management methods for urban rail ground-based energy storage systems based on traction grid voltage determine whether a train in the power supply section is in traction or braking mode by using the relationship between the traction grid voltage and the substation's no-load voltage, thereby controlling the charging and discharging of the ground-based energy storage system. Figure 2 As shown. When the traction network voltage is lower than the substation's no-load voltage, it is considered that a train traction condition has occurred; when the traction network voltage is higher than the substation's no-load voltage, it is considered that a train braking condition has occurred.

[0034] In the diagram, Usub represents the traction network voltage, and Unoload represents the substation's own no-load voltage.

[0035] However, when considering multiple substations and energy storage systems, especially when there are differences in the no-load voltage of each substation, the relationship between the traction network voltage and the no-load voltage of a particular substation is insufficient to accurately reflect the traction and braking characteristics of trains within the power supply section due to the influence of the no-load voltage of other substations. Furthermore, real-time communication between the ground-based energy storage system and the train is difficult to achieve due to requirements for high-performance communication and constraints related to information security. Therefore, it is necessary to propose a method that determines whether a train within the power supply section is in traction or braking mode solely by collecting available ground-based information online, thereby providing reference information for the charging and discharging control of the energy storage system.

[0036] K-means clustering, as an unsupervised algorithm, primarily functions to automatically group similar data in a dataset into the same cluster. Therefore, it is often used for autonomous data pattern discovery. The K-means algorithm measures the similarity between data samples by calculating Euclidean distance, dividing the dataset into clusters with K mean vectors, where the mean vector μ... k Indicates clustering C k The centroid of the mean vector. The expression for the mean vector is as follows:

[0037]

[0038] The sum of squared distances from each sample to the cluster center in a cluster is denoted as J. The ultimate goal of the K-means algorithm is to find K cluster centers that minimize J through an iterative process. The objective function J describes the proximity of samples within a cluster. The expression for J is as follows:

[0039]

[0040] Therefore, the K-means clustering algorithm is essentially a simple iterative process of numerical computation. The algorithm itself has low complexity, thus requiring low processor performance. Generally, DSP-based control systems used for energy storage system control can execute this algorithm. The K-means clustering algorithm flow is as follows: Figure 3 As shown.

[0041] A novel train condition identification method is proposed based on the K-means clustering algorithm. Offline, K-means clustering is used to mine patterns in collected historical data to establish the relationship between train conditions and ground features. Online, only real-time collection of relevant ground features is required, and train conditions within the section are identified based on the mining results of historical data. Ground information can be sampled using the substation's own sampling devices and the sampling devices of the energy storage system installed within the substation. The collectable ground information includes: substation voltage, current, and no-load voltage.

[0042] 2.1 Case Analysis

[0043] Taking a subway line in Beijing as an example, five substation sections numbered 1-5 were selected. Substations 2 and 4 have ground-based energy storage systems installed. Therefore, historical ground data from these two stations was collected and clustered using the K-means clustering algorithm. Following the classic "elbow method," the cluster size was determined to be K=2, meaning the collected historical ground data will ultimately be clustered into two categories. The original data has 6 dimensions. Figure 4 The image shows the clustering results after standardization of the 3-dimensional data.

[0044] Based on the final clustering results, the original ground data were labeled as 1 and 2, respectively. Since historical train information can also be obtained offline, train information passing through these five stations at the same scale was collected, and the sum of train power was obtained. The labeling results of the ground data were compared with the sum of train power within the interval. Figure 5 The results shown are the comparison results over a certain period of time.

[0045] As shown in the figure above, the clustering results of historical ground data accurately distinguished between train traction conditions (sum of power greater than 0) and braking conditions (sum of power less than 0) within the interval. Therefore, based on the patterns between ground information and train information obtained from the mining results of historical data, in online scenarios, it is only necessary to collect ground information of the same category and identify the corresponding train conditions according to the clustering results. Table 1 shows the statistical results of train condition identification accuracy based on this method and traditional methods.

[0046] Table 1 Accuracy of Train Operating Condition Identification

[0047] Identification Method - Test Conditions accuracy This method involves collecting ground information from substations 2 and 4. 90.67％ This method involves collecting ground information from Substation No. 2. 87.65％ Traditional method - Determining the relationship between voltage and no-load voltage at station No. 2 60.35％

[0048] The test results show that even if only relevant ground data from a single substation is collected and a working condition identification method based on data pattern mining is adopted, the identification accuracy is far higher than that of the traditional method of judging train working conditions based on the relationship between traction network voltage and no-load voltage.

[0049] 3. Energy Management Methods Based on Data Pattern Mining

[0050] Based on the aforementioned K-means clustering-based operating condition identification method, this invention proposes an energy management method based on data pattern mining. The clusterer trained using the above method identifies operating conditions from online-collected ground data. The clusterer, specifically, refers to a method based on the K-means clustering algorithm that mines data from offline-collected ground information to establish a correlation between ground information and train operating conditions (also collected offline), forming the mean vector described below (different mean vectors represent different train operating conditions). In the online identification phase, based on the offline training results, the train operating condition is determined by the relationship between the online-collected ground information and the mean vector.

[0051] Equation (3) is the formula for calculating the Euclidean distance between the newly collected online sample X and the mean vector (representing different working conditions) obtained by the trained clusterer.

[0052]

[0053] Where dis_tra is the distance between the new sample and the mean vector representing the traction condition, and dis_bra is the distance between the new sample and the mean vector representing the braking condition. Based on the relationship between these two distances, it is determined whether the new sample belongs to the traction condition or the braking condition, and the final condition coefficient M is output. Under the traction condition, M = -1, and under the braking condition, M = 1, as shown in equation (4).

[0054]

[0055] The reference value U for adjusting the charging and discharging voltage threshold of the energy storage system is adjusted based on the output operating condition coefficient M and the current operating status of the energy storage system. ref To ensure the relative stability of the energy storage system's operating state, i.e., to avoid switching back and forth between charging and discharging modes, if the energy storage system is already in discharging mode under traction conditions, or in charging mode under braking conditions, then U ref No adjustment is made; otherwise, the baseline value of the threshold is adjusted accordingly according to equation (5), ΔU ref The adjustment amount for each benchmark value can be flexibly adjusted according to the actual project needs.

[0056] U ref =U ref -M·ΔU ref (5)

[0057] The reference value U for the charge / discharge threshold is obtained by calculation using the above formula. ref Then, the charging and discharging voltage threshold of the energy storage system is calculated according to equation (6), thereby realizing the dynamic adjustment of the threshold. Where Δu c The adjustment amount for the charging voltage threshold, Δu dThe adjustment range for the discharge voltage threshold can be flexibly adjusted according to actual engineering needs. Meanwhile, to prevent the substation from charging the energy storage system itself due to an excessively low charging threshold, and to avoid the braking resistor on the train being activated due to a rapid rise in grid voltage caused by the energy storage system discharging when the discharge threshold is too high, a limit is set on the adjustment range of the reference value.

[0058]

[0059] Figure 6 The diagram shows the overall control flow of the energy management method based on data pattern mining proposed in this invention.

[0060] The energy management method based on data pattern mining proposed in this invention changes the reference value setting basis for the charging and discharging voltage threshold of the energy storage system from the traditional substation no-load voltage to the operating condition identification result. This effectively absorbs the regenerative braking energy of the train and releases it during train traction, thereby achieving better energy-saving effect. Table 2 shows the simulation verification results of the total output energy of all substations in a certain section of a Beijing subway line, the energy loss of train braking resistors, and the energy saving rate of substations when using different energy management methods in the above example analysis.

[0061] Table 2 Simulation verification results of different energy management methods (departure interval: 420s)

[0062]

[0063] It can be seen that when the energy management method proposed in this invention is adopted, the overall operating conditions of the train within the power supply section are accurately identified, the regenerative braking energy of the train is effectively absorbed and fully released during traction, thereby reducing the output energy of the substation within the power supply section and simultaneously reducing the train braking resistor loss. Compared with energy management methods based on no-load voltage, the energy saving rate of the substation is greatly improved after adopting this method, and the ground-based energy storage system achieves better energy-saving effects.

[0064] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. An energy management method based on data pattern mining, characterized in that, Includes the following steps: S1. Real-time acquisition of relevant ground feature information; S2. Based on the mining structure of historical data of the ground feature information, the train operating conditions in the section are identified online; S3. Establish the relationship between train operating conditions and ground characteristics; S4. Energy management is then carried out based on the relationship between train operating conditions and ground characteristics. In step S2, the online identification of train operating conditions is based on the K-means clustering algorithm. The clusterer is obtained by processing and training the K-means clustering algorithm, and the operating conditions are identified by the online collected ground data to determine the operating conditions and output the operating condition coefficient M. The online-collected ground data is used for working condition identification, including the Euclidean distance calculation formula between the newly collected online sample X and the mean vector obtained from the trained clusterer: （3） Based on the relationship between these two distances, determine whether the new sample belongs to the traction condition or the braking condition, and output the final condition coefficient M. Under the traction condition, M=-1, and under the braking condition, M=1, as shown in equation (4): （4） In equations (3) and (4), dis_tra is the distance between the new sample and the mean vector representing the traction condition, and dis_bra is the distance between the new sample and the mean vector representing the braking condition. The reference value U for adjusting the charging and discharging voltage threshold of the energy storage system is adjusted based on the output operating condition coefficient M and the current operating status of the energy storage system. ref To ensure the relative stability of the energy storage system's operating state, i.e., to prevent frequent switching between charging and discharging modes, if the energy storage system is already in discharging mode during traction or charging mode during braking, then U ref No adjustment is made; otherwise, the baseline value of the threshold is adjusted accordingly according to equation (5), ΔU ref The adjustment amount of each benchmark value can be flexibly adjusted according to the actual project needs; （5） The reference value U for the charge / discharge threshold is obtained by calculation using the above formula. ref Then, the charging and discharging voltage thresholds of the energy storage system are calculated according to equation (6), thereby realizing dynamic adjustment of the thresholds. They can be flexibly adjusted according to actual engineering needs. At the same time, in order to avoid the substation charging the energy storage system itself due to the charging threshold being too low, and the grid voltage rising rapidly when the energy storage system discharges due to the discharging threshold being too high, causing the braking resistor on the train to start, a limit is set on the adjustment range of the reference value. （6） In the formula, Δu c The adjustment amount for the charging voltage threshold, Δu d This is the adjustment amount for the discharge voltage threshold.

2. The energy management method based on data pattern mining according to claim 1, characterized in that: The energy management is an upper-level control method for converter control. By setting the charging / discharging voltage threshold of the energy storage system, it ensures that the system can operate in the discharging and charging states during train traction and braking conditions.

3. The energy management method based on data pattern mining according to claim 1, characterized in that: The clusterer that has been trained refers to the K-means clustering algorithm, which uses data mining to collect ground feature information offline, establishes the correlation between ground information and train operating conditions, and forms a mean vector. Different mean vectors represent different train operating conditions.

4. The energy management method based on data pattern mining according to claim 1, characterized in that: The ground feature information includes: substation voltage, current, and no-load voltage.

Images