CN117473408A - An energy management method based on train status awareness - Google Patents
An energy management method based on train status awareness Download PDFInfo
- Publication number
- CN117473408A CN117473408A CN202311643801.5A CN202311643801A CN117473408A CN 117473408 A CN117473408 A CN 117473408A CN 202311643801 A CN202311643801 A CN 202311643801A CN 117473408 A CN117473408 A CN 117473408A
- Authority
- CN
- China
- Prior art keywords
- train
- energy storage
- storage system
- power
- traction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
- B60L7/00—Electrodynamic brake systems for vehicles in general
- B60L7/10—Dynamic electric regenerative braking
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
- B60L2200/00—Type of vehicles
- B60L2200/26—Rail vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Transportation (AREA)
- Power Engineering (AREA)
- Probability & Statistics with Applications (AREA)
- Electric Propulsion And Braking For Vehicles (AREA)
Abstract
The invention discloses an energy management method based on train state sensing, which is used for clustering collected original features based on a clustering algorithm, and taking the collected original features as a primary feature extraction result; merging the primary feature extraction result with the original feature to serve as a final feature of the input deep neural network; and carrying out self-adaptive adjustment on the reference value of the voltage outer ring instruction of the energy storage converter according to the train state sensing result so as to realize the working condition that the energy storage system can automatically charge or discharge along with the train state, and meanwhile, the original control structure of the converter is not required to be changed. According to the train state sensing-based energy management method, on-line identification of the whole train state in the traction power supply section can be realized by only collecting part of ground information based on the deep learning method, so that the self-adaptive charge and discharge control of the energy storage system is realized according to the sensing result.
Description
Technical Field
The invention relates to the technical field of train energy management, in particular to an energy management method based on train state sensing.
Background
The urban rail train adopts a regenerative braking technology widely, namely, a traction motor works in a generator state in the braking process of the train, and the generated electric energy is fed back to a traction power supply network to interact with a similar traction train except for auxiliary power supply of the train. Due to the unidirectional conductivity of the traction rectifier unit of the transformer substation, when regenerative braking energy generated by a braking train cannot be fully absorbed, the traction network pressure will be increased. When the starting voltage of the train brake resistor is reached, the brake resistor is started to consume energy, and if the network voltage is still not controlled effectively, the regeneration failure, namely the air brake input, can occur along with the continuous lifting of the network voltage. If this occurs, the regenerative braking energy of the train cannot be fully utilized.
And the urban rail ground energy storage system is used for connecting the energy storage device with the traction network through the bidirectional converter. The remaining regenerative braking energy of the brake train is absorbed through active control of an Energy Management Strategy (EMS). By setting the charge/discharge voltage threshold of the energy storage system, the energy in the energy storage system is released when the train is pulled. The regenerative braking energy of the train can be effectively utilized, the effect of inhibiting the fluctuation of the network voltage is achieved, and the energy consumption of the urban rail traction power supply system is effectively reduced.
At present, a urban rail ground energy storage system mainly adopts a fixed threshold value or dynamic threshold value energy management strategy based on traction network voltage, and as the change of the traction network voltage is widely considered in the past, the train working condition can be reflected, namely, the network voltage is reduced when a train is in traction, and the network voltage is increased when the train is braked. The operating state of the energy storage system is thus determined by setting a charge threshold value and a discharge threshold value and comparing them with the traction network voltage. This method is also one of the most widely used in practice at present. However, when the fluctuation of the no-load voltage of the transformer substation and the difference of no-load voltages among a plurality of transformer substations are considered, the train working condition in the section is difficult to accurately reflect only through the traction network voltage. In this case, it is difficult for the energy storage system under the control of such an energy management policy to achieve better energy saving performance, resulting in an abnormal condition of "charge-only-discharge-free" or "discharge-only-discharge-non-charge" of the energy storage system, and thus, an optimal energy saving effect cannot be achieved. In addition, methods of directly acquiring train conditions require high performance communication between the train and the energy storage system.
Disclosure of Invention
The invention aims to provide an energy management method based on train state sensing, which can realize on-line identification of the whole train state in a traction power supply section by only collecting part of ground information based on a deep learning method, thereby realizing self-adaptive charge and discharge control of an energy storage system according to a sensing result.
In order to achieve the above object, the present invention provides an energy management method based on train state sensing, comprising the steps of:
s1, train state sensing method based on clustering combined deep learning algorithm
S11, train state features based on a K-means clustering algorithm are extracted once;
s12, train state feature secondary extraction based on deep learning is a train state sensing method;
s2, an energy management strategy based on train state sensing.
Preferably, in step S1, the train states include a train traction state and a train braking state, and based on the sampling device of the substation itself and the sampling device of the energy storage system installed in the substation, the ground information, that is, the original characteristics, which can be collected, including the voltage, the current and the no-load voltage of the substation, can be sampled.
Preferably, in step S11, the collected original features are clustered based on a clustering algorithm, and the original features can be divided into clusters and corresponding cluster centers, so that the distribution situation of the original data, i.e. which cluster the original data belongs to and the euclidean distance corresponding to the cluster center can be obtained, and the distribution situation is used as a result of primary feature extraction;
and combining the primary feature extraction result and the original feature as final features of the input deep neural network, taking the overall power condition of the train in the power supply interval obtained offline under the same time scale as a tag, and training the deep neural network to obtain the relationship between the features and the tag, namely the inherent relationship between the ground features and the working conditions of the train.
Preferably, in step S11, the following steps are included:
s111, collecting the traction and braking characteristics as original characteristics, performing clustering operation, and outputting a working condition coefficient M according to the following mode
S112, outputting Euclidean distance between the original characteristic and each mean value vector according to the formula (2).
S k ＝||x i -μ k || 2 (2)
S113, taking the working condition coefficient M as positive and negative relation, taking Euclidean distance as numerical value, and jointly forming power characteristics, wherein the power characteristics can be described as
Power feature= [ clustering original feature, m×s1/S2] (3)
It should be noted that the division operation is performed on the euclidean distance between the original feature and each mean vector, so as to better reflect the different distribution situations in the feature space.
Preferably, in step S2, the following steps are included:
s21, the working condition coefficient M represents braking or traction characteristics obtained by train state sensing, namely + -1 is respectively taken, the adjusting direction of a reference value is determined by the working condition coefficient, the adjusting quantity delta Uref of the reference value is determined by the difference value between the total power of the train and the input power of an energy storage system in a power supply interval obtained by a state sensing algorithm, the reference value adjusting rule is as follows, and the charging power of the energy storage system is positive and the discharging power is negative; the traction power of the train is positive, and the braking power is negative;
s22, correspondingly setting a charge and discharge threshold of the energy storage system according to the adjusted reference value, wherein Deltau is hysteresis voltage between the charge threshold and the discharge threshold, the hysteresis voltage and the adjustment coefficient kp can take the energy saving effect as an optimization target, and a specific optimization value is obtained by using a corresponding optimization algorithm;
s23, according to the adjustment rule, the reference value is reduced to reduce the corresponding charging threshold value when the brake working condition is met, and the energy storage system is controlled to enter a charging mode; when the energy storage system is in a traction working condition, the reference value is increased to correspond to the discharge threshold value, the energy storage system is controlled to enter a discharge mode, the boundary conditions of the reference value, the charge threshold value and the discharge threshold value are shown in a formula (6), the charge threshold value is larger than the no-load voltage of the transformer substation, the transformer substation is prevented from charging the ESS as far as possible, the discharge threshold value is lower than the starting voltage of the train brake resistor, and the train brake resistor is prevented from being started by mistake due to overlarge discharge power of the energy storage system;
based on the control logic, the adjustment degree of the charge and discharge of the energy storage system is based on the total power of the train in the power supply interval obtained through sensing, so that the ground energy storage system is controlled to effectively absorb the regenerative braking energy of the train and release the regenerative braking energy when the train is pulled, and the self-adaptive adjustment process is realized.
Therefore, the energy management strategy based on train state sensing is adopted, and on-line identification of the whole train state in the traction power supply section can be realized only by collecting part of ground information based on a deep learning method, so that the self-adaptive charge and discharge control of the energy storage system is realized according to the sensing result.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
FIG. 1 is a ground-based energy storage system control architecture;
FIG. 2 is a flow of a train state sensing method based on a cluster-combined deep learning algorithm;
FIG. 3 illustrates a simplified circuit model of a dual-sided power supply system under traction conditions;
the spatial distribution of the vectors of FIG. 4 versus operating conditions;
FIG. 5 clustering results;
FIG. 6 is a comparison of the cluster result and the sum of the section train powers;
FIG. 7 is a deep neural network training process;
FIG. 8 is a graph showing the result of identifying the sum of the power of the trains in the power supply section in a departure interval;
FIG. 9 is a graph showing the result of identifying the sum of the power of the trains in the power supply section in a departure interval;
Detailed Description
The technical scheme of the invention is further described below through the attached drawings and the embodiments.
Unless defined otherwise, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this invention belongs.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
Furthermore, it should be understood that although the present disclosure describes embodiments, not every embodiment is provided with a separate embodiment, and that this description is provided for clarity only, and that the disclosure is not limited to the embodiments described in detail below, and that the embodiments described in the examples may be combined as appropriate to form other embodiments that will be apparent to those skilled in the art. Such other embodiments are also within the scope of the present invention.
It should also be understood that the above-mentioned embodiments are only for explaining the present invention, the protection scope of the present invention is not limited thereto, and any person skilled in the art should be able to cover the protection scope of the present invention by equally replacing or changing the technical scheme and the inventive concept thereof within the scope of the present invention.
Techniques, methods, and apparatus known to one of ordinary skill in the relevant art may not be discussed in detail, but are intended to be considered part of the specification where appropriate.
The disclosures of the prior art documents cited in the present specification are incorporated by reference in their entirety into the present invention and are therefore part of the present disclosure.
Example 1
As shown in the figure, the invention provides an energy management method based on train state sensing, wherein an Energy Management Strategy (EMS) is used as an upper control strategy for converter control, and a charge/discharge voltage threshold of an energy storage system is set so as to ensure that the system can work in a discharge state and a charge state under the working conditions of train traction and braking. Therefore, the train state plays a vital role in the charge and discharge control of the energy storage system and the design of an energy management strategy, and the traction braking working condition and the power of the train are accurately acquired, so that the energy storage system is ensured to absorb the residual regenerative braking energy of the train as much as possible, and the energy storage system is released under the traction working condition of the train, so that a better energy saving effect is achieved.
The patent provides an energy management strategy based on train state perception, only needs to collect partial ground information based on a clustering combination deep learning algorithm, and can realize on-line recognition of the overall state of the train in a traction power supply section, including traction braking working conditions and power thereof, so as to realize self-adaptive charge and discharge control of an energy storage system according to a perception result.
2.1 train state sensing method based on clustering combined deep learning algorithm
At present, most urban rail ground energy storage system energy management strategies based on traction network voltage judge whether a train in a power supply section is in traction or braking working conditions according to the relation between the traction network voltage and no-load voltage of a transformer substation, so as to control the charging and discharging of the ground energy storage system. When the traction network voltage is lower than the idle voltage of the substation, the energy storage system is considered to discharge under the traction working condition of the train; when the traction network voltage is higher than the no-load voltage of the transformer substation, the train braking working condition is considered to occur, and the energy storage system should be charged. However, when a plurality of substations and a plurality of energy storage systems are considered, particularly when there is a difference in the no-load voltage of each substation, it is difficult to accurately reflect the traction braking characteristics of the train in the power supply section by only the relationship between the traction network voltage and the no-load voltage of a certain station due to the influence of the no-load voltage of other substations. And the traction braking characteristic of the train only determines the charge and discharge states of the energy storage system, and more accurate judgment on the residual traction braking power of the train is required to control the energy storage system more accurately.
The existing method is basically based on the combination of a single train power curve and offline information such as a train running chart, and the like, and the whole train power curve in an interval is drawn in advance and stored in a ground energy storage system controller. However, due to factors such as late train, inter-station residence time errors and the like, the offline information often has errors with actual conditions. Meanwhile, due to the requirements on high-performance communication conditions and the restrictions on information safety and other problems, real-time communication between the ground energy storage system and the train is difficult to realize. It is therefore necessary to consider a method for providing reference information for real-time control of the energy storage system by merely collecting ground-available information to identify online whether the overall train condition within the power supply section includes traction or braking and power conditions.
The key of on-line perception of the train state is to extract the relevant characteristics reflecting the train state, and the deep learning method has outstanding advantages in the fields of characteristic extraction, pattern recognition and the like, so that the deep neural network is adopted to automatically extract the characteristics reflecting the train state. Meanwhile, in order to improve the efficiency and accuracy of train state sensing, the patent proposes that a K-means clustering algorithm is firstly applied to perform primary extraction before the original features are input into a deep neural network, and the feature result of the primary extraction is combined with the original features to serve as input of secondary extraction. The principle flow chart of the train state sensing method provided by the patent is shown in the following chart.
2.1.1 train state characteristics once extraction based on K-means clustering algorithm
The K-means clustering algorithm is used as an unsupervised algorithm, and the main function of the algorithm is to automatically group similar data in a data set into the same class. The K-means algorithm measures similarity between data samples by calculating Euclidean distance, dividing the data set into clusters with K mean vectors, where the mean vectors μ k Representing cluster C k Is a centroid of (c). The expression of the mean vector is as follows
The sum of the squares of the distances of each sample in the cluster to the cluster center is denoted J. The final result of the K-means algorithm is that K cluster centers are found through an iterative process to minimize J, the objective function J describing the proximity of the samples within the cluster. The expression of J is as follows
(1) Train traction status feature
Taking a typical urban rail traction power supply system bilateral power supply topology as an example. Fig. 3 is a simplified circuit model comprising two substations and a train, wherein the two substations differ in no-load voltage and the train is in traction.
The power of the train is provided by two substations on two sides, and can be described as the power of the train since the train is equivalent to a constant power load at any time under simplified conditions
According to Kirchhoff's Voltage Law (KVL), a voltage loop equation in the circuit can be obtained, and the train node voltage can be deduced as follows
According to Kirchhoff's Current Law (KCL), a current node equation in the circuit can be obtained, and the current of the train can be deduced into the following formula
Substituting the train current expression in the formula (5) into the train voltage expression in the formula (4) and substituting the two into the formula (3), the functional relationship of the train power and the related quantity can be expressed as
The line impedance in the formula (4) is related to the number of trains and the positions thereof, and is uniformly expressed by the train position St in the above formula.
(2) Train braking status feature
Similarly, the functional relationship between train power and related quantities in a braking state consistent with the method can be expressed as
From equations (6) and (7), the traction network voltage is no longer the only factor reflecting the train operating characteristics when considering the polytropic plant. The position of the train, the current and voltage change rates of the substations at the two sides are also affected by the running conditions of the train. The variables of the equations except the train position of the blue mark are all information which can be acquired from the ground.
In summary, all the ground information contained in the deduction can be used to judge the train traction braking condition based on the clustering algorithm, and the primary characteristics required by train state sensing can be extracted.
(3) Train state feature one-time extraction
Based on the traction and braking characteristics, a K-means clustering algorithm is adopted to perform data clustering, different clusters representing different operation conditions can be formed, and mean value vectors representing centers of the different clusters are obtained. The different mean vectors represent the traction or braking conditions, i.e. the total power positive and negative relationship of the train in the power supply interval. The euclidean distance between the feature vector and the mean vector that are clustered therefore represents, to some extent, the value of the total power of the train in the power supply interval. As shown in fig. 4.
According to the analysis, the power characteristics in the power supply interval are extracted according to the clustering result of the original characteristics, and the characteristic extraction process is as follows
Step 1, collecting the traction and braking characteristics as original characteristics, clustering, and outputting a working condition coefficient M according to the following mode
And 2, outputting Euclidean distance between the original characteristic and each mean value vector according to the formula (9).
S k ＝||x i -μ k || 2 (9)
And 3, taking the working condition coefficient M as a positive-negative relation and taking the Euclidean distance as a numerical value to jointly form the power characteristic. The power characteristics may be described as
Power feature= [ clustering original feature, m×s1/S2] (10)
It should be noted that the division operation is performed on the euclidean distance between the original feature and each mean vector, so as to better reflect the different distribution situations in the feature space. As shown in fig. 4, due to the mean vector μ 2 Is less than mu 1 Distance of (x) to be characteristic x i And x j Classified as a braking condition. But at the same time from x i And x j To mu 2 The euclidean distances S12 and S22 are substantially equal, but their distribution in the feature space is quite different.
2.1.1 train state feature secondary extraction, namely train state sensing method based on deep learning
Due to the great advantage of deep learning in the aspect of feature learning, a primary feature extraction result and an original feature are combined to be used as final features of the input deep neural network, and the whole power condition of the train in the power supply interval obtained offline under the same time scale is used as a tag, and the deep neural network is trained to obtain the relationship between the features and the tag, namely the inherent relationship between the ground features and the train working condition. The following table is a training process of the deep neural network, wherein x is an input feature (original feature+primary feature extraction result), and y is a tag (sum of train power in a power supply interval at the same time).
TABLE 1 deep neural network training process
And realizing on-line real-time perception of the train state based on the trained deep neural network.
Taking a certain subway line in Beijing as an example, selecting five transformer substation intervals with numbers 1-5 as an example, wherein the ground energy storage systems are installed in the transformer substations 2 and 4. And therefore, the historical ground information acquired in the two stations is collected and clustered based on a K-means clustering algorithm. The clustering class k=2 is determined according to the classical "elbow method", i.e. the above collected historical ground data will eventually be clustered into two classes. The original data dimension is 6 dimensions, and fig. 5 shows the clustering result after the 3-dimensional data is normalized.
And marking the original ground data as 1 and 2 respectively according to the final clustering result. Because the history information of the train can be obtained on line, the information of the train passing through the five stations under the same scale is collected, and the sum of the power of the train is obtained. Comparing the marking result of the ground data with the sum of the train power in the section, FIG. 6 shows the comparison result in a certain period of time,
From the above graph, it can be seen that the clustering result of the historical ground data accurately distinguishes the train traction condition (the sum of the power is greater than 0) and the braking condition (the sum of the power is less than 0) in the section. And according to the extraction result of the clustering algorithm on the original features, taking the distribution condition of the original data, namely the information such as the relative distance between each data point and the cluster center, and the like as a primary feature extraction result, and combining the primary feature with the original features to obtain a new feature vector. A deep neural network containing 6 hidden layers is built, and network parameter optimization in the training process is performed based on an Adam optimizer. And taking the combined feature vector as input of the deep neural network, taking the sum of train power in a power supply interval as a tag, and training the deep neural network. The lower graph is a training iterative process, and it can be seen that the entire training process exhibits good convergence characteristics, and that no significant over-fit or under-fit problems have occurred, based on the performance of the loss function on the validation set.
The training-completed deep neural network is used for on-line sensing of the train states of the section, the ground available information is collected on line based on the whole flow, the feature extraction is carried out once by a trained clustering device, and the extraction result and the original information are combined to be used as final feature vectors and input into the training-completed deep neural network. Fig. 8 shows the recognition result of the sum of the power of the trains in the section by the train state sensing method provided by the patent in a certain departure interval.
Based on the results, the method provided by the patent can realize the on-line identification of the overall state of the train in the urban rail traction power supply section, including the overall traction braking characteristic and the corresponding total power of the train.
2.2 energy management policy based on train Condition awareness
Conventional traction network voltage-based energy management strategies typically indirectly determine train conditions based on the relationship of traction network voltage to no-load voltage, and set charging and discharging voltage thresholds based on substation no-load voltage. However, this method is not reliable based on the above analysis. Therefore, according to the train state sensing method based on the clustering combination deep learning algorithm, the patent provides an energy management strategy based on train state sensing. And carrying out self-adaptive adjustment on the reference value of the voltage outer ring instruction of the energy storage converter according to the train state sensing result so as to realize the working condition that the energy storage system can automatically charge or discharge along with the train state, and meanwhile, the original control structure of the converter is not required to be changed. The overall control structure is shown in fig. 9.
The working condition coefficient M represents braking or traction characteristics obtained by train state sensing, the braking or traction characteristics are respectively + -1, the adjusting direction of the reference value is determined by the working condition coefficient, and the adjusting quantity delta Uref of the reference value is determined by the difference value between the total power of the train and the input power of the energy storage system in the power supply interval obtained by the state sensing algorithm. The reference value adjustment rule is as follows, and it should be noted that the charging power of the energy storage system is positive and the discharging power is negative; the train traction power is positive and the braking power is negative.
And correspondingly setting the charge and discharge threshold of the energy storage system according to the adjusted reference value, as shown in a formula (11). Where Δu is a hysteresis voltage between the charge threshold and the discharge threshold. The hysteresis voltage and the adjustment coefficient kp can take the energy-saving effect as an optimization target, and a specific optimization value is obtained by using a corresponding optimization algorithm.
According to the regulation rule, when the brake working condition is met, the reference value is reduced, the corresponding charging threshold is reduced, and the energy storage system is controlled to enter a charging mode; when the traction working condition is adopted, the reference value will rise and the corresponding discharge threshold value will rise, and the energy storage system will be controlled to enter a discharge mode. In order to ensure the normal and reliable operation of the system, the boundary conditions of the reference value, the charge threshold value and the discharge threshold value are shown as a formula (12). The charging threshold is larger than the no-load voltage of the transformer substation, and the transformer substation is prevented from charging the ESS as much as possible. The discharge threshold value is lower than the starting voltage of the train braking resistor, so that the train braking resistor is prevented from being started by mistake due to overlarge discharge power of the energy storage system.
Based on the control logic, the adjustment degree of the charge and discharge of the energy storage system is based on the total power of the train in the power supply interval obtained through sensing, so that the ground energy storage system is controlled to effectively absorb the regenerative braking energy of the train and release the regenerative braking energy when the train is pulled, and the self-adaptive adjustment process is realized. Table 2 shows simulation verification results of total energy output, train brake resistance loss energy and energy saving rate of all substations in a certain section of a certain subway line in Beijing in the above calculation example analysis, when different energy management strategies are adopted respectively.
Table 2 simulation verification results for different energy management strategies (departure interval: 420 s)
| Substation output energy (kWh) | Braking resistance loss (kWh) | Energy saving rate | |
| Uninstalled ground energy storage system | 123.1 | 30.85 | 0.00％ |
| EMS based on no-load voltage | 118.7 | 13.79 | 3.57％ |
| The method | 100.6 | 6.9 | 18.3％ |
According to the energy management strategy, on-line sensing of the train state in the power supply interval is achieved, the train regenerative braking energy is effectively absorbed and fully released in traction working conditions, so that the output energy of a transformer substation in the power supply interval is reduced, and meanwhile, the train braking resistance loss is reduced. Compared with the traditional energy management strategy based on no-load voltage identification, the energy saving rate of the transformer substation is greatly improved after the method is adopted, and the ground energy storage system achieves a good energy saving effect.
Therefore, the energy management method based on train state sensing is adopted, and on-line identification of the whole train state in the traction power supply section can be realized only by collecting part of ground information based on the deep learning method, so that the self-adaptive charge and discharge control of the energy storage system is realized according to the sensing result.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention and not for limiting it, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that: the technical scheme of the invention can be modified or replaced by the same, and the modified technical scheme cannot deviate from the spirit and scope of the technical scheme of the invention.
Claims (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311643801.5A CN117473408A (en) | 2023-12-04 | 2023-12-04 | An energy management method based on train status awareness |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311643801.5A CN117473408A (en) | 2023-12-04 | 2023-12-04 | An energy management method based on train status awareness |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117473408A true CN117473408A (en) | 2024-01-30 |
Family
ID=89634844
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311643801.5A Pending CN117473408A (en) | 2023-12-04 | 2023-12-04 | An energy management method based on train status awareness |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117473408A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5835901A (en) * | 1994-01-25 | 1998-11-10 | Martin Marietta Corporation | Perceptive system including a neural network |
| US20070240242A1 (en) * | 2006-04-06 | 2007-10-11 | Monsanto Technology Llc | Method for multivariate analysis in predicting a trait of interest |
| US20190041811A1 (en) * | 2017-08-03 | 2019-02-07 | Johnson Controls Technology Company | Building management system with augmented deep learning using combined regression and artificial neural network modeling |
| CN114492671A (en) * | 2022-02-17 | 2022-05-13 | 河海大学 | Non-invasive household load identification method based on improved FCM clustering algorithm and MLP neural network |
| CN117013647A (en) * | 2023-07-17 | 2023-11-07 | 北京交通大学 | An energy management method based on data pattern mining |
-
2023
- 2023-12-04 CN CN202311643801.5A patent/CN117473408A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5835901A (en) * | 1994-01-25 | 1998-11-10 | Martin Marietta Corporation | Perceptive system including a neural network |
| US20070240242A1 (en) * | 2006-04-06 | 2007-10-11 | Monsanto Technology Llc | Method for multivariate analysis in predicting a trait of interest |
| US20190041811A1 (en) * | 2017-08-03 | 2019-02-07 | Johnson Controls Technology Company | Building management system with augmented deep learning using combined regression and artificial neural network modeling |
| CN114492671A (en) * | 2022-02-17 | 2022-05-13 | 河海大学 | Non-invasive household load identification method based on improved FCM clustering algorithm and MLP neural network |
| CN117013647A (en) * | 2023-07-17 | 2023-11-07 | 北京交通大学 | An energy management method based on data pattern mining |
Non-Patent Citations (1)
| Title |
|---|
| 王晓东;苗宜之;卢奭瑄;刘颖明;: "基于SCM-ANFIS负荷预测的储能电站调峰控制策略", 太阳能学报, no. 06, 28 June 2018 (2018-06-28) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111091143B (en) | Distribution transformer weight overload early warning method based on deep belief network and K-means clustering | |
| CN108173273A (en) | A kind of intelligent phase-change switch system and method for adjusting three-phase imbalance | |
| CN102368610A (en) | Evaluation method based on distribution system security region | |
| CN102879680A (en) | Rail transit vehicle equipment universal detection and fault diagnosis method and system | |
| CN108805321A (en) | A kind of electric automobile charging station planing method | |
| CN120150194A (en) | Optimal scheduling method for hybrid energy storage system based on AI intelligent control | |
| US12151580B1 (en) | Energy management framework and method for rail transit energy storage | |
| Zhang et al. | Urban rail substation parameter optimization by energy audit and modified salp swarm algorithm | |
| CN114204675A (en) | Power distribution station electric energy data acquisition terminal based on cloud edge cooperation | |
| CN111738518B (en) | A charging and discharging scheduling method for electric vehicles based on taking into account the average discharge rate | |
| CN114640137A (en) | Cluster AGC load distribution method and system for efficient and flexible regulation of regional coal power | |
| CN105720676A (en) | Multi-mobile energy storage car emergency power-supply coordinating control algorithm | |
| CN118381118B (en) | A Data-Driven Wasserstein Fuzzy Set-Based Active Distribution Network Distributed Blue Day-Ahead Scheduling Method | |
| CN117473408A (en) | An energy management method based on train status awareness | |
| CN120280967B (en) | Flywheel energy storage optimal configuration method for urban rail transit regenerative braking energy recovery | |
| CN117013647A (en) | An energy management method based on data pattern mining | |
| CN120422725A (en) | Layered management method for fuel cell hybrid tramcar with self-adaptive working condition sensing function | |
| Chang et al. | Research of power consumers behavior using fuzzy c-means algorithm | |
| Ding et al. | Reliability evaluation of direct current distribution system for intelligent buildings based on big data analysis | |
| Zhong et al. | Coordinated control of the onboard and wayside energy storage system of an urban rail train based on rule mining | |
| CN117239762A (en) | A substation voltage adaptive control method based on urban rail flexible traction power supply system | |
| CN116644924A (en) | A Method for Scheduling Resources in Virtual Power Plant | |
| Yang et al. | Artificial Intelligence Based Fault Diagnosis and Analysis in the Distribution Network | |
| CN115483752A (en) | A comprehensive uninterrupted power supply load control and matching method based on emergency power supply vehicle | |
| Ren et al. | Optimal Coordination of Distance Protection Considering Multiple Setting Groups Strategy Based on Sideline Current Addition Coefficient |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20240130 |
|
| RJ01 | Rejection of invention patent application after publication |