A charging station power load adjustable potential prediction method and related device

Abstract

The application discloses a charging station power load adjustable potential prediction method and related devices, and the application obtains a load adjustable potential sequence B according to a historical sequence, a charging vehicle quantity prediction model, a power adjustable charging vehicle quantity prediction model, a charging load prediction model and a load adjustable potential prediction model, realizes more accurate charging station power load adjustable potential prediction, and can provide a decision basis for power load regulation and control.

Description

TECHNICAL FIELD

[0001] The application relates to a charging station power load adjustable potential prediction method and a related device, and belongs to the field of time series data prediction. BACKGROUND

[0002] In a power grid system, if the power demand of a power user (for example, an industrial enterprise, a commercial building, a residential area, a new energy vehicle charging station, etc.) exceeds the power transmission capacity of the power grid, the power grid system collapses. In order to ensure the smooth operation of the power system, the power load needs to be controlled to some extent to eliminate the power gap.

[0003] Demand response refers to the process in which the power dispatching system sends a demand to reduce the power consumption of a power user according to the dispatching demand, and the power user automatically or manually reduces the power consumption to respond to the dispatching demand. The power load adjustable potential refers to the size of the load adjustment capability of the power user during the demand response period. The current adjustable potential prediction method is based on the way of cutting off power supply, that is, the power consumption without cutting off power supply is the adjustable potential. However, for the charging piles of the charging station, the power limiting method is flexible and adjustable, and some charging vehicles can be adjusted, while some cannot. The number of vehicles to be charged is also unknown. These factors bring complexity to the prediction of the adjustable potential, and there is no related prior art. SUMMARY

[0004] The application provides a charging station power load adjustable potential prediction method and a related device, which solves the problems disclosed in the background art.

[0005] To solve the above technical problems, the technical scheme adopted by the application is as follows:

[0006] A charging station power load adjustable potential prediction method, comprising:

[0007] For different power adjustment ratios of the load regulation period, according to the initial historical charging vehicle quantity sequence in the observation window, the initial historical power adjustable charging vehicle quantity sequence in the observation window, the initial historical charging load sequence in the observation window, the initial historical load adjustable potential sequence in the observation window, the initial historical power adjustment ratio sequence in the observation window, a charging vehicle quantity prediction model, a power adjustable charging vehicle quantity prediction model, a charging load prediction model, and a load adjustable potential prediction model, a load adjustable potential sequence B is obtained; wherein the prediction window contains the load regulation period, the load regulation period is the period in which the dispatching system regulates the load of the charging station, and the load adjustable potential sequence B is the load adjustable potential sequence in the prediction window.

[0008] extracting a load adjustable potential sequence A from the load adjustable potential sequence B, and obtaining a load adjustable potential sequence C closest to a load gap; wherein the load gap is a load amount that needs to be regulated by the charging station in the load regulation process; the load adjustable potential sequence A is a load adjustable potential sequence in the load regulation period;

[0009] If a difference between the load adjustable potential sum of the load adjustable potential sequence C and the load gap is less than a threshold value, the load adjustable potential sequence C is taken as a charging station power load adjustable potential prediction result.

[0010] For different power regulation ratios of the load regulation period, the load adjustable potential sequence B is obtained according to the initial historical charging vehicle quantity sequence in the observation window, the initial historical power adjustable charging vehicle quantity sequence in the observation window, the initial historical charging load sequence in the observation window, the initial historical load adjustable potential sequence in the observation window, the initial historical power regulation ratio sequence in the observation window, a charging vehicle quantity prediction model, a power adjustable charging vehicle quantity prediction model, a charging load prediction model and a load adjustable potential prediction model, and includes:

[0011] 1) predicting a charging vehicle quantity at a next time according to the initial historical charging vehicle quantity sequence in the observation window and the charging vehicle quantity prediction model;

[0012] predicting a power adjustable charging vehicle quantity at a next time according to the initial historical power adjustable charging vehicle quantity sequence in the observation window and the power adjustable charging vehicle quantity prediction model;

[0013] predicting a charging load at a next time according to the initial historical charging load sequence in the observation window and the charging load prediction model;

[0014] 2) removing the earliest charging vehicle quantity in the initial historical charging vehicle quantity sequence, and combining the charging vehicle quantity at the next time and the remaining charging vehicle quantities in the initial historical charging vehicle quantity sequence into a new historical charging vehicle quantity sequence;

[0015] removing the earliest power adjustable charging vehicle quantity in the initial historical power adjustable charging vehicle quantity sequence, and combining the power adjustable charging vehicle quantity at the next time and the remaining power adjustable charging vehicle quantities in the initial historical power adjustable charging vehicle quantity sequence into a new historical power adjustable charging vehicle quantity sequence;

[0016] removing the earliest charging load in the initial historical charging load sequence, and combining the charging load at the next time and the remaining charging loads in the initial historical charging load sequence into a new historical charging load sequence;

[0017] The earliest power adjustment ratio in the initial historical power adjustment ratio sequence in the observation window is removed, and the power adjustment ratio at the next moment and the remaining power adjustment ratios in the initial historical power adjustment ratio sequence are combined into a new historical power adjustment ratio sequence; the power adjustment ratios at each moment in the load regulation period are consistent;

[0018] 3) According to the new historical number of charging vehicles sequence, the new historical power-adjustable number of charging vehicles sequence, the new historical charging load sequence, the new historical power adjustment ratio sequence, the initial historical load-adjustable potential sequence in the observation window and the load-adjustable potential prediction model, the load-adjustable potential at the next moment is obtained;

[0019] 4) The earliest load-adjustable potential in the initial historical load-adjustable potential sequence is removed, and the load-adjustable potential at the next moment and the remaining load-adjustable potentials in the historical load-adjustable potential sequence are combined into a new historical load-adjustable potential sequence;

[0020] 5) If the new historical load-adjustable potential sequence contains the load regulation period, go to 6), otherwise, take the new historical number of charging vehicles sequence as the initial historical number of charging vehicles sequence, take the new historical power-adjustable number of charging vehicles sequence as the initial historical power-adjustable number of charging vehicles sequence, take the new initial historical charging load sequence as the initial historical charging load sequence, take the new historical load-adjustable potential sequence as the initial historical load-adjustable potential sequence, take the new historical power adjustment ratio sequence as the initial historical power adjustment ratio sequence, and go to 1);

[0021] 6) Adjust the power adjustment ratio in the load regulation period according to the preset rule, if the adjusted power adjustment ratio in the load regulation period does not exceed the preset range, go to 1), otherwise, obtain the load-adjustable potential sequence B in the prediction window under different power adjustment ratios.

[0022] The preset range of the power adjustment ratio in the load regulation period is 0-1; the preset rule for adjusting the power adjustment ratio in the load regulation period is to adjust the power adjustment ratio in the load regulation period in an arithmetic sequence.

[0023] If the difference between the load-adjustable potential sum of the load-adjustable potential sequence C and the load gap is less than a threshold value, the load-adjustable potential sequence C is taken as the charging station power load-adjustable potential prediction result, and it is determined that the load-adjustable potential meets the load regulation demand, and the power adjustment ratio corresponding to the load-adjustable potential sequence C is sent to the dispatching system as the actual regulation parameter;

[0024] If the difference between the load-adjustable potential sum of the load-adjustable potential sequence C and the load gap is not less than a threshold value, it is determined that the load-adjustable potential does not meet the load regulation demand, and the load-adjustable potential sequence A corresponding to the maximum power adjustment ratio is sent to the dispatching system.

[0025] A charging station power load adjustable potential prediction device, comprising:

[0026] A prediction module, for different power adjustment ratios of a load regulation period, according to an initial historical charging vehicle quantity sequence in an observation window, an initial historical power adjustable charging vehicle quantity sequence in the observation window, an initial historical charging load sequence in the observation window, an initial historical load adjustable potential sequence in the observation window, an initial historical power adjustment ratio sequence in the observation window, a charging vehicle quantity prediction model, a power adjustable charging vehicle quantity prediction model, a charging load prediction model, and a load adjustable potential prediction model, obtains a load adjustable potential sequence B; wherein the prediction window contains the load regulation period, the load regulation period is the period of load regulation of the charging station by the dispatching system; the load adjustable potential sequence B is the load adjustable potential sequence in the prediction window;

[0027] An acquisition module extracts a load adjustable potential sequence A from the load adjustable potential sequence B, and obtains a load adjustable potential sequence C closest to a load gap; wherein the load gap is the load amount that needs to be regulated by the charging station in the load regulation process; the load adjustable potential sequence A is the load adjustable potential sequence in the load regulation period;

[0028] A result module, if the difference between the total load adjustable potential of the load adjustable potential sequence C and the load gap is less than a threshold value, takes the load adjustable potential sequence C as the charging station power load adjustable potential prediction result.

[0029] The process of the prediction module obtaining the load adjustable potential sequence B, comprising:

[0030] 1) According to the initial historical charging vehicle quantity sequence in the observation window and the charging vehicle quantity prediction model, predicting the charging vehicle quantity at the next time;

[0031] According to the initial historical power adjustable charging vehicle quantity sequence in the observation window and the power adjustable charging vehicle quantity prediction model, predicting the power adjustable charging vehicle quantity at the next time;

[0032] According to the initial historical charging load sequence in the observation window and the charging load prediction model, predicting the charging load at the next time;

[0033] 2) Eliminating the earliest charging vehicle quantity in the initial historical charging vehicle quantity sequence, combining the charging vehicle quantity at the next time and the remaining charging vehicle quantity in the initial historical charging vehicle quantity sequence into a new historical charging vehicle quantity sequence;

[0034] The earliest power adjustable charging vehicle quantity in the initial historical power adjustable charging vehicle quantity sequence is removed, and the power adjustable charging vehicle quantity at the next moment and the remaining power adjustable charging vehicle quantities in the initial historical power adjustable charging vehicle quantity sequence are combined into a new historical power adjustable charging vehicle quantity sequence.

[0035] The earliest charging load in the initial historical charging load sequence is removed, and the charging load at the next moment and the remaining charging loads in the initial historical charging load sequence are combined into a new historical charging load sequence.

[0036] The earliest power adjustment ratio in the initial historical power adjustment ratio sequence within the observation window is removed, and the power adjustment ratio at the next moment and the remaining power adjustment ratios in the initial historical power adjustment ratio sequence are combined into a new historical power adjustment ratio sequence; wherein the power adjustment ratios at each moment within the load regulation period are consistent.

[0037] 3) According to the new historical charging vehicle quantity sequence, the new historical power adjustable charging vehicle quantity sequence, the new historical charging load sequence, the new historical power adjustment ratio sequence, the initial historical load adjustable potential sequence within the observation window and the load adjustable potential prediction model, the load adjustable potential at the next moment is obtained.

[0038] 4) The earliest load adjustable potential in the initial historical load adjustable potential sequence is removed, and the load adjustable potential at the next moment and the remaining load adjustable potentials in the historical load adjustable potential sequence are combined into a new historical load adjustable potential sequence.

[0039] 5) If the new historical load adjustable potential sequence contains the load regulation period, go to 6), otherwise, take the new historical charging vehicle quantity sequence as the initial historical charging vehicle quantity sequence, take the new historical power adjustable charging vehicle quantity sequence as the initial historical power adjustable charging vehicle quantity sequence, take the new initial historical charging load sequence as the initial historical charging load sequence, take the new historical load adjustable potential sequence as the initial historical load adjustable potential sequence, take the new historical power adjustment ratio sequence as the initial historical power adjustment ratio sequence, and go to 1).

[0040] 6) Adjust the power adjustment ratio within the load regulation period according to the preset rule, if the adjusted power adjustment ratio within the load regulation period does not exceed the preset range, go to 1), otherwise, obtain the load adjustable potential sequence B within the prediction window under different power adjustment ratios.

[0041] The preset range of the power adjustment ratio within the load regulation period is 0-1; the preset rule for adjusting the power adjustment ratio within the load regulation period is to adjust the power adjustment ratio within the load regulation period in an arithmetic sequence.

[0042] The result module determines that the load adjustable potential meets the load regulation requirement if the difference between the total sum of the load adjustable potential sequence C and the load gap is less than the threshold value, and sends the power regulation ratio corresponding to the load adjustable potential sequence C to the dispatching system as the actual regulation parameter.

[0043] The result module determines that the load adjustable potential does not meet the load regulation requirement if the difference between the total sum of the load adjustable potential sequence C and the load gap is not less than the threshold value, and sends the load adjustable potential sequence A corresponding to the maximum power regulation ratio to the dispatching system.

[0044] A computer-readable storage medium storing one or more programs, the one or more programs comprising instructions which, when executed by a computing device, cause the computing device to perform the charging station power load adjustable potential prediction method.

[0045] A computing device comprising one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs comprise instructions for performing the charging station power load adjustable potential prediction method.

[0046] The present application has the following beneficial effects: The present application obtains the load adjustable potential sequence B according to the historical sequence, the charging vehicle quantity prediction model, the power adjustable charging vehicle quantity prediction model, the charging load prediction model, and the load adjustable potential prediction model, realizes more accurate charging station power load adjustable potential prediction, and provides a decision basis for power load regulation. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A flowchart of the charging station power load adjustable potential prediction method is shown.

[0048] Figure 2 A flowchart for obtaining the load adjustable potential sequence B is shown.

[0049] Figure 3 A model structure of DeepAR is shown. DETAILED DESCRIPTION

[0050] The present application will be further described below with reference to the accompanying drawings. The following examples are only used to more clearly illustrate the technical solutions of the present application, and cannot be used to limit the protection scope of the present application.

[0051] As shown in Figure 1 A charging station power load adjustable potential prediction method comprises the following steps:

[0052] Step 1, according to the initial historical charging vehicle quantity sequence in the observation window, the initial historical power-adjustable charging vehicle quantity sequence in the observation window, the initial historical charging load sequence in the observation window, the initial historical load-adjustable potential sequence in the observation window, the initial historical power adjustment ratio sequence in the observation window, the charging vehicle quantity prediction model, the power-adjustable charging vehicle quantity prediction model, the charging load prediction model, and the load-adjustable potential prediction model, a load-adjustable potential sequence B is obtained for different power adjustment ratios of the load regulation period; wherein, the prediction window contains the load regulation period, the load regulation period is the period of load regulation of the scheduling system to the charging station, and is a future period; the load-adjustable potential sequence B is the load-adjustable potential sequence in the prediction window.

[0053] Step 2, the load-adjustable potential sequence A is extracted from the load-adjustable potential sequence B, and the load-adjustable potential sequence C closest to the load gap is obtained; wherein, the load gap is the load amount that needs to be regulated by the charging station in the load regulation process; the load-adjustable potential sequence A is the load-adjustable potential sequence in the load regulation period.

[0054] Step 3, if the difference between the load-adjustable potential sum of the load-adjustable potential sequence C and the load gap is less than a threshold value, the load-adjustable potential sequence C is taken as the charging station power load-adjustable potential prediction result.

[0055] The above method obtains the load-adjustable potential sequence B according to the historical sequence, the charging vehicle quantity prediction model, the power-adjustable charging vehicle quantity prediction model, the charging load prediction model, and the load-adjustable potential prediction model, realizes more accurate charging station power load-adjustable potential prediction, and can provide decision basis for power load regulation. The above method is implemented at the charging station side. In the demand response scenario, the scheduling system sends a load regulation request to the charging station, which is a three-tuple <Gap, Time, Duration>, wherein, Gap is the load gap, i.e. the load amount that needs to be regulated by the charging station in the load regulation process; Time is when the load regulation needs to be performed, and Duration is how long the load regulation needs to be performed. The charging station side feeds back the adjustable potential prediction result to the scheduling system based on the above method.

[0056] Before implementing the above method, the charging vehicle quantity prediction model, the power-adjustable charging vehicle quantity prediction model, the charging load prediction model, and the load-adjustable potential prediction model need to be constructed and trained in advance.

[0057] An observation window and a prediction window are defined, wherein the sequence length of the observation window is n, the sequence length of the prediction window is m, and the size of the prediction window should be greater than or equal to Duration, i.e. the period in the prediction window should contain the load regulation period.

[0058] As Figure 2As shown, for different power adjustment ratios of the load regulation period, according to the initial historical charging car quantity sequence in the observation window, the initial historical power-adjustable charging car quantity sequence in the observation window, the initial historical charging load sequence in the observation window, the initial historical load-adjustable potential sequence in the observation window, the initial historical power adjustment ratio sequence in the observation window, the charging car quantity prediction model, the power-adjustable charging car quantity prediction model, the charging load prediction model, and the load-adjustable potential prediction model, the load-adjustable potential sequence B in the prediction window is obtained, which can specifically include:

[0059] 1) According to the initial historical charging car quantity sequence in the observation window and the charging car quantity prediction model, the charging car quantity at the next time is predicted; according to the initial historical power-adjustable charging car quantity sequence in the observation window and the power-adjustable charging car quantity prediction model, the power-adjustable charging car quantity at the next time is predicted; and according to the initial historical charging load sequence in the observation window and the charging load prediction model, the charging load at the next time is predicted.

[0060] That is, by inputting the initial historical series into the corresponding prediction model, the predicted value at the next time can be obtained.

[0061] 2) The earliest charging car quantity in the initial historical charging car quantity sequence is removed, and the charging car quantity at the next time and the remaining charging car quantities in the initial historical charging car quantity sequence are combined into a new historical charging car quantity sequence;

[0062] The earliest power-adjustable charging car quantity in the initial historical power-adjustable charging car quantity sequence is removed, and the power-adjustable charging car quantity at the next time and the remaining power-adjustable charging car quantities in the initial historical power-adjustable charging car quantity sequence are combined into a new historical power-adjustable charging car quantity sequence;

[0063] The earliest charging load in the initial historical charging load sequence is removed, and the charging load at the next time and the remaining charging loads in the initial historical charging load sequence are combined into a new historical charging load sequence;

[0064] The earliest power adjustment ratio in the initial historical power adjustment ratio sequence in the observation window is removed, and the power adjustment ratio at the next time and the remaining power adjustment ratios in the initial historical power adjustment ratio sequence are combined into a new historical power adjustment ratio sequence; wherein the power adjustment ratios at each time in the load regulation period are consistent.

[0065] Taking the initial historical charging vehicle quantity sequence as an example, the length of the initial historical charging vehicle quantity sequence is n, and the time corresponding to the last element in the initial historical charging vehicle quantity sequence is t. After 1), the charging vehicle quantity corresponding to t+1 can be obtained. The charging vehicle quantity corresponding to t+1 and the last n-1 elements in the initial historical charging vehicle quantity sequence are combined to form a new historical charging vehicle quantity sequence. The other sequences are similar.

[0066] 3) According to the new historical charging vehicle quantity sequence, the new historical power-adjustable charging vehicle quantity sequence, the new historical charging load sequence, the new historical power adjustment ratio sequence, the initial historical load-adjustable potential sequence in the observation window and the load-adjustable potential prediction model, the load-adjustable potential at the next time is obtained.

[0067] That is, the new historical charging vehicle quantity sequence, the new historical power-adjustable charging vehicle quantity sequence, the new historical charging load sequence, the new historical power adjustment ratio sequence, the initial historical load-adjustable potential sequence in the observation window are input into the load-adjustable potential prediction model, and the load-adjustable potential at the next time is obtained.

[0068] 4) The earliest load-adjustable potential in the initial historical load-adjustable potential sequence is removed, and the load-adjustable potential at the next time and the remaining load-adjustable potential in the historical load-adjustable potential sequence are combined to form a new historical load-adjustable potential sequence.

[0069] The length of the historical load-adjustable potential sequence is also n. If power adjustment is not performed in a certain time unit, 0 is supplemented.

[0070] 5) If the new historical load-adjustable potential sequence contains a load control period, go to 6), otherwise, the new historical charging vehicle quantity sequence is taken as the initial historical charging vehicle quantity sequence, the new historical power-adjustable charging vehicle quantity sequence is taken as the initial historical power-adjustable charging vehicle quantity sequence, the new initial historical charging load sequence is taken as the initial historical charging load sequence, the new historical load-adjustable potential sequence is taken as the initial historical load-adjustable potential sequence, and the new historical power adjustment ratio sequence is taken as the initial historical power adjustment ratio sequence, and go to 1).

[0071] 6) The power adjustment ratio in the load control period is adjusted according to the preset rule, and if the adjusted power adjustment ratio in the load control period does not exceed the preset range, go to 1), otherwise, obtain the load-adjustable potential sequence B in the prediction window under different power adjustment ratios; wherein the preset range of the power adjustment ratio in the load control period is 0-1; the preset rule for adjusting the power adjustment ratio in the load control period is to adjust the power adjustment ratio in the load control period in the form of an arithmetic sequence.

[0072] Since the sequence length of the prediction window is m, that is, the number of time points in the load regulation period is m, steps 1) to 5) are repeated m times without adjusting the power adjustment ratio, and a load adjustable potential sequence under the current power adjustment ratio is obtained.

[0073] The load adjustable potential under different power adjustment ratios is different, so the power adjustment ratio is adjusted, and 1) to 6) are repeated, that is, a plurality of load adjustable potential sequences under different power adjustment ratios are obtained.

[0074] Potential prediction essentially belongs to time series prediction, and all models can use DeepAR neural network time series model, which can support multivariate prediction of single variable, and of course can also use Informer and other time series models based on Transformer.

[0075] DeepAR learns the correlation characteristics inside different time series through deep recurrent neural network, and uses multiple or multiple target numbers to improve the overall prediction accuracy. DeepAR finally produces a multi-step prediction result with a selectable time span, and the prediction of a single time node is a probability prediction, which outputs P10, P50 and P90 by default. P10 here refers to the probability distribution, that is, 10% of the probability will be less than the value of P10. By giving a probability prediction, both a value prediction and a decision using the interval of P10 to P90 can be made.

[0076] The model structure of DeepAR is as shown in Figure 3 , wherein z i,t represents the value of the i-th sequence at time step t, x i,t represents a feature, and t0 represents the starting time of prediction. DeepAR predicts the probability distribution of z i,t based on an autoregressive recurrent neural network, and uses a likelihood function l(z i,t |Θ i,t ) to represent it.

[0077] During training, at each time step t, the input of the network includes the feature x i,t , the value z i,t-1 of the previous step, and the state h i,t-1 of the previous step.

[0078] First, calculate the current state:

[0079] h i,t =h(h i,t-1 ,z i,t-1 ,x i,t )

[0080] Where h is a function composed of the connection relationship of the neurons in the hidden layer of the neural network.

[0081] Further, l(z i,t |Θ i,t ), so that the gradient descent method can be used to calculate Θ i,t (Θ i,t indicates the weight on the connection edge of each layer of neurons, and the model parameter, and calculates Θ i,t is the trained model, which can be seen in the paper of DeapAR: Probabilistic Forecasting with Autoregressive Recurrent Networks, https: / / arxiv.org / abs / 1704.04110.

[0082] In the model inference process, the ancestral sampling method is used, and for t0, t0+1,..., T, z i,t is randomly sampled at each prediction time step. By repeating this process, the sampling values of t0, t0+1,..., T can be obtained, and the mathematical expectation and variance can be easily obtained by using these sampling values.

[0083] The load adjustable potential sequence A is extracted from each load adjustable potential sequence B, and the load adjustable potential sequence C closest to the load gap is obtained; wherein the load adjustable potential sequence A is the load adjustable potential sequence in the load regulation period.

[0084] If the difference between the total load adjustable potential of the load adjustable potential sequence C and the load gap is less than the threshold value, the load adjustable potential sequence C is taken as the charging station power load adjustable potential prediction result, and it is determined that the load adjustable potential meets the load regulation demand, and the power regulation ratio corresponding to the load adjustable potential sequence C is sent to the dispatching system as the actual regulation parameter; if the difference between the total load adjustable potential of the load adjustable potential sequence C and the load gap is not less than the threshold value, it is determined that the load adjustable potential does not meet the load regulation demand, and the load adjustable potential sequence A corresponding to the maximum power regulation ratio is sent to the dispatching system.

[0085] In order to further illustrate the above method, experiments are carried out by using certain charging station data, and Table 1 is the experimental data. Because the time sequence is relatively long, only a part of the data is given for observation.

[0086] Table 1 Experimental data

[0087]

[0088]

[0089] Among them, the power regulation ratio in a period of time is consistent.

[0090] This data is an equal-interval time series with a time resolution of 15 minutes, which contains detailed information on charging load, charging vehicle number, power-adjustable charging vehicle number, power adjustment ratio, load-adjustable potential, and other dimensions within each time period.

[0091] The data contains a time series with a resolution of 15 minutes for charging load, charging vehicle number, power-adjustable charging vehicle number, power adjustment ratio, load-adjustable potential, and other information between October 10, 2020 and August 25, 2021. Based on this data, four basic models are trained using DeepAR: charging vehicle number prediction model, power-adjustable charging vehicle number prediction model, charging load prediction model, and load-adjustable potential prediction model.

[0092] Taking the charging vehicle number prediction model as an example, assume the observation window is 8 and the prediction window is 1, i.e., using the historical 2-hour charging vehicle number to predict the future 15-minute charging vehicle number. Each training sample contains x and y, x is the historical data, and y is the data to be predicted. In this example, the first training sample is the charging vehicle number in 2020.10.10 00:00-2020.10.10 01:45, which is x, i.e., [2, 4, 7, 8, 5, 9, 6, 3], and the charging vehicle number at 2020.10.10 02:00 is y, i.e., [4]. The next training sample is constructed by shifting the previous sample by one time unit, and so on. All training samples are constructed, and the data from October 2020 to June 2021 is used as the training set, and the remaining data is used as the validation set for model validation. By adjusting the learning rate, batch size, learning rate decay coefficient, model layer number, and training round number, the best model is obtained. The MAPE and RMSE are used as double-index validation and evaluation.

[0093] The remaining model training is constructed according to the above method, the model is trained, and the best model is selected according to the evaluation method. When training the load-adjustable potential prediction model, it is a single variable prediction based on multiple variables, so the first training sample is the load-adjustable potential sequence between 2020.10.10 00:00-2020.10.10 01:45 and the covariate data such as charging vehicle number sequence, power-adjustable charging vehicle number sequence, charging load sequence, and power adjustment ratio sequence between 2020.10.10 00:00-2020.10.10 02:00, which are collectively used as x in the sample, and the load-adjustable potential at 2020.10.10 02:00 is y to construct the training sample. The next sample is still obtained by shifting the previous sample by one time unit, and the following training and evaluation process is consistent with other model training.

[0094] After the model is trained, the load adjustable potential is predicted, since the latest data is only updated to 2021.08.2523:45, the load adjustable potential on 2021.08.26 00:00 is predicted, since the load adjustable potential prediction model is a multivariate prediction univariate model, at this time, the charging vehicle quantity, the power adjustable charging vehicle quantity, the charging load, the power regulation ratio and the like information on 2021.08.2600:00 should be known; wherein the power regulation ratio is accessed by subsequent prediction, and the charging vehicle quantity, the power adjustable charging vehicle quantity, the charging load and the like information need to be predicted by using the model trained above.

[0095] Taking the prediction of the charging vehicle quantity as an example, since the observation window is 8 and the prediction window is 1 during the training, the charging vehicle quantity sequence between 2021.08.2522:00 and 2021.08.2523:45 is selected as the input of the model, and the model output result is the charging vehicle quantity on 2021.08.2600:00, the power adjustable charging vehicle quantity and the charging load are predicted by the above method.

[0096] At this time, the charging vehicle quantity, the power adjustable charging vehicle quantity and the charging load on 2021.08.2600:00 are obtained, the power regulation ratio can be input by self-definition, therefore, based on the trained load adjustable potential prediction model, the input data is the load adjustable potential sequence between 2021.08.2522:00 and 2021.08.2523:45 and the charging vehicle quantity sequence, the power adjustable charging vehicle quantity sequence, the charging load sequence and the power regulation ratio sequence between 2021.08.2522:00 and 2021.08.2600:00, the load adjustable potential on 2021.08.2600:00 can be predicted.

[0097] The new power regulation ratio is replaced, the load adjustable potential under different power regulation ratios is obtained, the load adjustable potential closest to the Gap is selected, if the difference between the load adjustable potential and the load gap is less than a threshold value, the power regulation ratio at this time is fed back to the scheduling system, otherwise the maximum load adjustable potential is returned.

[0098] Based on the same technical scheme, the application further discloses a software device of the above method, a charging station power load adjustable potential prediction device, comprising:

[0099] The prediction module obtains a load adjustable potential sequence B according to the initial historical charging vehicle quantity sequence in the observation window, the initial historical power adjustable charging vehicle quantity sequence in the observation window, the initial historical charging load sequence in the observation window, the initial historical load adjustable potential sequence in the observation window, the initial historical power adjustment ratio sequence in the observation window, the charging vehicle quantity prediction model, the power adjustable charging vehicle quantity prediction model, the charging load prediction model and the load adjustable potential prediction model for different power adjustment ratios in the load regulation period.

[0100] The process in which the prediction module obtains the load adjustable potential sequence B in the prediction window includes:

[0101] 1) predicting the charging vehicle quantity at the next moment according to the initial historical charging vehicle quantity sequence in the observation window and the charging vehicle quantity prediction model;

[0102] predicting the power adjustable charging vehicle quantity at the next moment according to the initial historical power adjustable charging vehicle quantity sequence in the observation window and the power adjustable charging vehicle quantity prediction model;

[0103] predicting the charging load at the next moment according to the initial historical charging load sequence in the observation window and the charging load prediction model;

[0104] 2) removing the earliest charging vehicle quantity in the initial historical charging vehicle quantity sequence, combining the charging vehicle quantity at the next moment and the remaining charging vehicle quantities in the initial historical charging vehicle quantity sequence into a new historical charging vehicle quantity sequence;

[0105] removing the earliest power adjustable charging vehicle quantity in the initial historical power adjustable charging vehicle quantity sequence, combining the power adjustable charging vehicle quantity at the next moment and the remaining power adjustable charging vehicle quantities in the initial historical power adjustable charging vehicle quantity sequence into a new historical power adjustable charging vehicle quantity sequence;

[0106] removing the earliest charging load in the initial historical charging load sequence, combining the charging load at the next moment and the remaining charging loads in the initial historical charging load sequence into a new historical charging load sequence;

[0107] removing the earliest power adjustment ratio in the initial historical power adjustment ratio sequence in the observation window, combining the power adjustment ratio at the next moment and the remaining power adjustment ratios in the initial historical power adjustment ratio sequence into a new historical power adjustment ratio sequence; wherein the power adjustment ratios at each moment in the load regulation period are consistent;

[0108] 3) obtaining the load adjustable potential at the next time according to the new historical charging vehicle quantity sequence, the new historical power adjustable charging vehicle quantity sequence, the new historical charging load sequence, the new historical power regulation ratio sequence, the initial historical load adjustable potential sequence within the observation window and the load adjustable potential prediction model;

[0109] 4) removing the earliest load adjustable potential in the initial historical load adjustable potential sequence, and combining the load adjustable potential at the next time and the remaining load adjustable potentials in the historical load adjustable potential sequence into a new historical load adjustable potential sequence;

[0110] 5) if the new historical load adjustable potential sequence contains the load regulation period, going to 6), otherwise, taking the new historical charging vehicle quantity sequence as the initial historical charging vehicle quantity sequence, taking the new historical power adjustable charging vehicle quantity sequence as the initial historical power adjustable charging vehicle quantity sequence, taking the new initial historical charging load sequence as the initial historical charging load sequence, taking the new historical load adjustable potential sequence as the initial historical load adjustable potential sequence, taking the new historical power regulation ratio sequence as the initial historical power regulation ratio sequence, and going to 1);

[0111] 6) adjusting the power regulation ratio within the load regulation period according to a preset rule, if the adjusted power regulation ratio within the load regulation period does not exceed a preset range, going to 1), otherwise, obtaining the load adjustable potential sequence B within the prediction window under different power regulation ratios.

[0112] The preset range of the power regulation ratio within the load regulation period is 0-1, and the preset rule for adjusting the power regulation ratio within the load regulation period is to adjust the power regulation ratio within the load regulation period in an arithmetic sequence manner.

[0113] The obtaining module extracts a load adjustable potential sequence A and a load adjustable potential sequence C closest to a load gap from the load adjustable potential sequence B, wherein the load gap is the load amount that needs to be regulated by the charging station in the load regulation process, and the load adjustable potential sequence A is the load adjustable potential sequence within the load regulation period.

[0114] The result module, if the difference between the load adjustable potential sum of the load adjustable potential sequence C and the load gap is less than a threshold value, takes the load adjustable potential sequence C as the charging station power load adjustable potential prediction result, determines that the load adjustable potential meets the load regulation demand, and sends the power regulation ratio corresponding to the load adjustable potential sequence C to the dispatching system as the actual regulation parameter;

[0115] If the difference between the load adjustable potential sum of the load adjustable potential sequence C and the load gap is not less than the threshold value, it is determined that the load adjustable potential does not meet the load regulation demand, and the load adjustable potential sequence A corresponding to the maximum power regulation ratio is sent to the dispatching system.

[0116] Based on the same technical solution, the application further discloses a computer readable storage medium storing one or more programs, the one or more programs including instructions which, when executed by a computing device, cause the computing device to perform the charging station power load adjustable potential prediction method.

[0117] Based on the same technical solution, the application further discloses a computing device including one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing the charging station power load adjustable potential prediction method.

[0118] Those skilled in the art will understand that embodiments of the application can be provided as methods, systems, or computer program products. Therefore, the application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer-usable program code.

[0119] The application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in a flow or multiple flows and / or blocks Figure 1 The functions specified in a flow or multiple flows and / or blocks

[0120] These computer program instructions can also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in a flow or multiple flows and / or blocks Figure 1 The functions specified in a flow or multiple flows and / or blocks

[0121] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are generated to realize the computer-implemented processes in the computer or other programmable devices, and the instructions executed in the computer or other programmable devices provide operational steps for implementing the functions specified in the flowchart Figure 1 one flow or a plurality of flows and / or the functions specified in the block Figure 1 one block or a plurality of blocks.

[0122] The above merely illustrates the embodiments of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of the claims of the present application.

Claims

1. A method for predicting the adjustable potential of power load in charging stations, characterized in that, include: For different power regulation ratios during load control periods, the adjustable load potential sequence B is obtained based on the following: the initial historical sequence of the number of charging vehicles, the initial historical sequence of the number of charging vehicles with adjustable power, the initial historical sequence of the charging load, the initial historical sequence of the adjustable load potential, the initial historical sequence of the power regulation ratio, the charging vehicle number prediction model, the adjustable power charging vehicle number prediction model, the charging load prediction model, and the adjustable load potential prediction model within the observation window. The prediction window includes the load control period, which is the period during which the dispatch system regulates the load of the charging station. The adjustable load potential sequence B is the adjustable load potential sequence within the prediction window. Extract load adjustable potential sequence A from load adjustable potential sequence B, and obtain load adjustable potential sequence C that is closest to the load gap; where the load gap is the load amount that needs to be regulated by the charging station during the load regulation process; and load adjustable potential sequence A is the load adjustable potential sequence during the load regulation period. If the difference between the sum of the adjustable load potentials of the adjustable load potential sequence C and the load gap is less than the threshold, then the adjustable load potential sequence C will be used as the prediction result of the adjustable load potential of the charging station. The above-mentioned different power regulation ratios for load control periods, based on the initial historical charging vehicle quantity sequence, the initial historical power adjustable charging vehicle quantity sequence, the initial historical charging load sequence, the initial historical load adjustable potential sequence, the initial historical power regulation ratio sequence, the charging vehicle quantity prediction model, the power adjustable charging vehicle quantity prediction model, the charging load prediction model, and the load adjustable potential prediction model within the observation window, obtain the load adjustable potential sequence B, including: 1) Based on the initial historical charging vehicle quantity sequence within the observation window and the charging vehicle quantity prediction model, predict the charging vehicle quantity at the next moment; Based on the initial historical sequence of the number of adjustable power charging vehicles in the observation window and the prediction model of the number of adjustable power charging vehicles, predict the number of adjustable power charging vehicles at the next moment. Based on the initial historical charging load sequence and charging load prediction model within the observation window, predict the charging load at the next moment; 2) Remove the earliest number of charging vehicles from the initial historical charging vehicle number sequence, and merge the number of charging vehicles at the next moment with the remaining number of charging vehicles in the initial historical charging vehicle number sequence into a new historical charging vehicle number sequence. Remove the earliest power adjustable charging vehicle quantity from the initial historical power adjustable charging vehicle quantity sequence, and merge the power adjustable charging vehicle quantity at the next moment with the remaining power adjustable charging vehicle quantity in the initial historical power adjustable charging vehicle quantity sequence into a new historical power adjustable charging vehicle quantity sequence. Remove the earliest charging load from the initial historical charging load sequence, and merge the charging load at the next moment with the remaining charging load in the initial historical charging load sequence into a new historical charging load sequence. Remove the earliest power regulation ratio from the initial historical power regulation ratio sequence within the observation window, and merge the power regulation ratio at the next moment with the remaining power regulation ratios in the initial historical power regulation ratio sequence into a new historical power regulation ratio sequence; wherein, the power regulation ratio is consistent at each moment within the load control period. 3) Based on the new historical charging vehicle quantity sequence, the new historical power adjustable charging vehicle quantity sequence, the new historical charging load sequence, the new historical power adjustment ratio sequence, the initial historical load adjustable potential sequence within the observation window, and the load adjustable potential prediction model, obtain the load adjustable potential at the next moment. 4) Remove the earliest load adjustable potential from the initial historical load adjustable potential sequence, and merge the load adjustable potential at the next moment and the remaining load adjustable potential in the historical load adjustable potential sequence into a new historical load adjustable potential sequence. 5) If the new historical load adjustable potential sequence includes a load control period, proceed to 6); otherwise, use the new historical charging vehicle quantity sequence as the initial historical charging vehicle quantity sequence, the new historical power adjustable charging vehicle quantity sequence as the initial historical power adjustable charging vehicle quantity sequence, the new initial historical charging load sequence as the initial historical charging load sequence, the new historical load adjustable potential sequence as the initial historical load adjustable potential sequence, and the new historical power adjustment ratio sequence as the initial historical power adjustment ratio sequence, then proceed to 1). 6) Adjust the power regulation ratio during the load control period according to the preset rules. If the adjusted power regulation ratio during the load control period does not exceed the preset range, then go to 1). Otherwise, obtain the load adjustable potential sequence B under different power regulation ratios.

2. The method for predicting the adjustable potential of power load in a charging station according to claim 1, characterized in that, The preset range of the power regulation ratio during the load control period is 0~1; the preset rule for adjusting the power regulation ratio during the load control period is to adjust the power regulation ratio during the load control period in an arithmetic sequence manner.

3. The method for predicting the adjustable potential of charging station power load according to claim 1, characterized in that, If the difference between the sum of the adjustable load potentials of the adjustable load potential sequence C and the load gap is less than the threshold, then the adjustable load potential sequence C is used as the predicted result of the adjustable load potential of the charging station, and it is determined that the adjustable load potential meets the load regulation requirements. The power regulation ratio corresponding to the adjustable load potential sequence C is sent to the dispatching system as the actual regulation parameter. If the difference between the sum of the load adjustable potentials of the load adjustable potential sequence C and the load gap is not less than the threshold, it is determined that the load adjustable potential does not meet the load regulation requirements, and the load adjustable potential sequence A corresponding to the maximum power regulation ratio is sent to the scheduling system.

4. A device for predicting the adjustable power load potential of a charging station, characterized in that, include: The prediction module, for different power adjustment ratios during load control periods, obtains the load adjustable potential sequence B based on the following: the initial historical sequence of the number of charging vehicles, the initial historical sequence of the number of charging vehicles with adjustable power, the initial historical sequence of the charging load, the initial historical sequence of the load adjustable potential, the initial historical sequence of the power adjustment ratio, the charging vehicle number prediction model, the power adjustable charging vehicle number prediction model, the charging load prediction model, and the load adjustable potential prediction model within the observation window. The prediction window includes the load control period, which is the period during which the dispatch system controls the load of the charging station; the load adjustable potential sequence B is the load adjustable potential sequence within the prediction window. The acquisition module extracts the load adjustable potential sequence A from the load adjustable potential sequence B and obtains the load adjustable potential sequence C that is closest to the load gap; where the load gap is the amount of load that needs to be regulated by the charging station during the load regulation process; and the load adjustable potential sequence A is the load adjustable potential sequence during the load regulation period. In the results module, if the difference between the sum of the adjustable load potentials of the adjustable load potential sequence C and the load gap is less than the threshold, then the adjustable load potential sequence C is taken as the prediction result of the adjustable load potential of the charging station. The process by which the above-mentioned prediction module obtains the load adjustability potential sequence B includes: 1) Based on the initial historical charging vehicle quantity sequence within the observation window and the charging vehicle quantity prediction model, predict the charging vehicle quantity at the next moment; Based on the initial historical sequence of the number of adjustable power charging vehicles in the observation window and the prediction model of the number of adjustable power charging vehicles, predict the number of adjustable power charging vehicles at the next moment. Based on the initial historical charging load sequence and charging load prediction model within the observation window, predict the charging load at the next moment; 2) Remove the earliest number of charging vehicles from the initial historical charging vehicle number sequence, and merge the number of charging vehicles at the next moment with the remaining number of charging vehicles in the initial historical charging vehicle number sequence into a new historical charging vehicle number sequence. Remove the earliest power adjustable charging vehicle quantity from the initial historical power adjustable charging vehicle quantity sequence, and merge the power adjustable charging vehicle quantity at the next moment with the remaining power adjustable charging vehicle quantity in the initial historical power adjustable charging vehicle quantity sequence into a new historical power adjustable charging vehicle quantity sequence. Remove the earliest charging load from the initial historical charging load sequence, and merge the charging load at the next moment with the remaining charging load in the initial historical charging load sequence into a new historical charging load sequence. Remove the earliest power regulation ratio from the initial historical power regulation ratio sequence within the observation window, and merge the power regulation ratio at the next moment with the remaining power regulation ratios in the initial historical power regulation ratio sequence into a new historical power regulation ratio sequence; wherein, the power regulation ratio is consistent at each moment within the load control period. 3) Based on the new historical charging vehicle quantity sequence, the new historical power adjustable charging vehicle quantity sequence, the new historical charging load sequence, the new historical power adjustment ratio sequence, the initial historical load adjustable potential sequence within the observation window, and the load adjustable potential prediction model, obtain the load adjustable potential at the next moment. 4) Remove the earliest load adjustable potential from the initial historical load adjustable potential sequence, and merge the load adjustable potential at the next moment and the remaining load adjustable potential in the historical load adjustable potential sequence into a new historical load adjustable potential sequence. 5) If the new historical load adjustable potential sequence includes a load control period, proceed to 6); otherwise, use the new historical charging vehicle quantity sequence as the initial historical charging vehicle quantity sequence, the new historical power adjustable charging vehicle quantity sequence as the initial historical power adjustable charging vehicle quantity sequence, the new initial historical charging load sequence as the initial historical charging load sequence, the new historical load adjustable potential sequence as the initial historical load adjustable potential sequence, and the new historical power adjustment ratio sequence as the initial historical power adjustment ratio sequence, then proceed to 1). 6) Adjust the power regulation ratio during the load control period according to the preset rules. If the adjusted power regulation ratio during the load control period does not exceed the preset range, then go to 1). Otherwise, obtain the load adjustable potential sequence B under different power regulation ratios.

5. The charging station power load adjustable potential prediction device according to claim 4, characterized in that, The preset range of the power regulation ratio during the load control period is 0~1; the preset rule for adjusting the power regulation ratio during the load control period is to adjust the power regulation ratio during the load control period in an arithmetic sequence manner.

6. The charging station power load adjustable potential prediction device according to claim 4, characterized in that, In the results module, if the difference between the sum of the adjustable load potentials of the adjustable load potential sequence C and the load gap is less than the threshold, then the adjustable load potential sequence C is used as the predicted result of the adjustable load potential of the charging station, and it is determined that the adjustable load potential meets the load regulation requirements. The power regulation ratio corresponding to the adjustable load potential sequence C is sent to the dispatch system as the actual regulation parameter. If the difference between the sum of the load adjustable potentials of the load adjustable potential sequence C and the load gap is not less than the threshold, it is determined that the load adjustable potential does not meet the load regulation requirements, and the load adjustable potential sequence A corresponding to the maximum power regulation ratio is sent to the scheduling system.

7. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods according to claims 1 to 3.

8. A computing device, characterized in that, include: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods according to claims 1 to 3.

Images