CN111564848A - Intelligent power dispatching method and power utilization load prediction device for micro power grid - Google Patents

Abstract

An intelligent power dispatching method and a power load forecasting device of a micro power grid are used for acquiring power utilization data of the micro power grid for a period of time; inputting the electricity utilization data of the period of time into a pre-established electricity utilization load prediction model to predict the electricity utilization condition at a future moment relative to the period of time, wherein the electricity utilization condition comprises a predicted value and a probability distribution range of the predicted value; and carrying out matching scheduling on the electric energy of the micro-grid according to the predicted electricity utilization condition at a future moment relative to the period of time.

Description

A kind of intelligent power dispatching method and power load forecasting device for micro grid

technical field

The invention relates to an intelligent power dispatching method and a power load forecasting device of a micro-grid.

Background technique

Microgrid refers to the connection of scattered power generation resources (such as self-powered power generation equipment or backup generator sets, solar power generation devices, wind power generation equipment and other renewable energy power generation devices) in a certain area or in some enterprises and institutions. It supplies power to users, and runs in parallel with the main large-scale power grid through the distribution network, forming a system in which a large-scale grid and small power generation equipment operate jointly. In a sense, when the distributed power source reaches a certain proportion, it can be called a micro grid.

In a microgrid, electricity consumption will not remain constant. For example, there are peaks and troughs in electricity consumption at different times of the day, peaks and troughs at different times of the week, and also in a month. During the peak period of electricity consumption of a microgrid, if the supply situation does not meet the electricity demand, it will bring many problems, but if the microgrid is kept in a high power supply state, it will generate a lot of unnecessary waste of resources. . Therefore, this is a problem that needs to be solved.

SUMMARY OF THE INVENTION

The present application provides an intelligent power dispatching method and a power load forecasting device for a microgrid.

According to a first aspect, an embodiment provides an electric load prediction device, including:

Sensors, used to obtain electricity consumption data for a period of time;

The processor is configured to input the electricity consumption data of the period of time into a pre-established electricity consumption load prediction model to predict the electricity consumption situation at a moment in the future relative to the period of time, and the electricity consumption situation includes the normal value , minimum and maximum values;

The electricity load prediction model is established in the following ways:

Obtain a training set, the data in the training set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the most recent electricity consumption data at the same moment. value, the minimum value of electricity consumption data at the same time in several recent times;

Using the training set, the electricity load prediction model is obtained by training; specifically, it includes: constructing a prediction model based on integrated deep learning in advance. The prediction model first decomposes the input electricity consumption data through an empirical mode decomposition algorithm to obtain Sub-signals of different frequencies, and then apply the deep recurrent neural network to analyze and predict each sub-signal, and then integrate the output obtained after analyzing and predicting each sub-signal, as the predicted electricity consumption; will be trained through the training set The latter prediction model is used as the electricity load prediction model. .

The electricity load prediction model is also established in the following ways:

Obtain a test set, the data in the test set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data at the same moment in recent times is the largest value, the minimum value of electricity consumption data at the same time in several recent times;

Use the test set to verify the electricity load prediction model obtained by training the training set; take the error between the label of the data in the test set and the normal value of the electricity consumption situation predicted by the electricity load prediction model as the standard, Perform hyperparameter tuning on the power consumption load prediction model to obtain a power consumption load prediction model after hyperparameter optimization.

In one embodiment, the power usage includes at least power.

In an embodiment, the electricity consumption data of the microgrid for a period of time includes power, corresponding voltage, current and phase angle.

In one embodiment, the power consumption data at the same time in the last several days includes the power consumption data at the same time in the most recent days, or the power consumption data at the same time in the same week in the most recent weeks.

According to a second aspect, an embodiment provides an intelligent power dispatching method for a microgrid, including:

Obtain electricity consumption data for a period of time in the microgrid;

Input the power consumption data of the period of time into a pre-established power consumption load prediction model to predict the power consumption situation at a moment in the future relative to the period of time, and the power consumption situation includes the predicted value, or the predicted value and its probability distribution range;

According to the predicted electricity consumption at a moment in the future relative to the period of time, the electric energy of the microgrid is matched and dispatched;

The electricity load prediction model is established in the following ways:

Obtain a training set, the data in the training set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the most recent electricity consumption data at the same moment. value, the minimum value of electricity consumption data at the same time in several recent times;

Using the training set, the electricity load prediction model is obtained by training; specifically, it includes: constructing a prediction model based on integrated deep learning in advance. The prediction model first decomposes the input electricity consumption data through an empirical mode decomposition algorithm to obtain Sub-signals of different frequencies, and then apply the deep recurrent neural network to analyze and predict each sub-signal, and then integrate the output obtained after analyzing and predicting each sub-signal, as the predicted electricity consumption; will be trained through the training set The latter prediction model is used as the electricity load prediction model.

The electricity load prediction model is also established in the following ways:

Obtain a test set, the data in the test set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data at the same moment in recent times is the largest value, the minimum value of electricity consumption data at the same time in several recent times;

Use the test set to verify the electricity load prediction model obtained through the training set;

Taking the error between the label of the data in the test set and the normal value of the electricity consumption predicted by the electricity load prediction model as the standard, the hyperparameter tuning of the electricity load prediction model is carried out to obtain the hyperparameter tuning. Electricity load forecasting model.

In an embodiment, the electricity usage situation includes at least a predicted value of power, or a predicted value of power and a probability distribution range thereof.

In one embodiment, the predicted value of the power includes a maximum value, a minimum value and a normal value of the power.

In an embodiment, the electricity consumption data of the microgrid for a period of time includes power, corresponding voltage, current and phase angle.

In one embodiment, the power consumption data at the same time in the last several days includes the power consumption data at the same time in the most recent days, or the power consumption data at the same time in the same week in the most recent weeks.

According to a third aspect, an embodiment provides a computer-readable storage medium comprising a program executable by a processor to implement the method described in any of the embodiments herein.

According to the smart power scheduling method of the microgrid, the electricity load forecasting device and the computer-readable storage medium according to the above embodiments, the power consumption of the microgrid is matched and dispatched by predicting the electricity consumption of the microgrid.

Description of drawings

FIG. 1 is a schematic structural diagram of an electric load prediction device according to an embodiment;

FIG. 2 is an algorithm flow chart of an electricity load prediction model according to an embodiment;

Fig. 3 is the algorithm flow chart of the electricity load prediction model of another embodiment;

FIG. 4 is a flowchart of a smart power dispatching method for a microgrid according to an embodiment;

5 is a flow chart of establishing an electricity load prediction model according to an embodiment;

FIG. 6 is a flow chart of establishing an electricity load prediction model according to another embodiment.

Detailed ways

The present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments have used associated similar element numbers. In the following embodiments, many details are described so that the present application can be better understood. However, those skilled in the art will readily recognize that some of the features may be omitted under different circumstances, or may be replaced by other elements, materials, and methods. In some cases, some operations related to the present application are not shown or described in the specification, in order to avoid the core part of the present application from being overwhelmed by excessive description, and for those skilled in the art, these are described in detail. The relevant operations are not necessary, and they can fully understand the relevant operations according to the descriptions in the specification and general technical knowledge in the field.

Additionally, the features, acts, or characteristics described in the specification may be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can also be exchanged or adjusted in order in a manner obvious to those skilled in the art. Therefore, the various sequences in the specification and drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a necessary order unless otherwise stated, a certain order must be followed.

The serial numbers themselves, such as "first", "second", etc., for the components herein are only used to distinguish the described objects, and do not have any order or technical meaning. The "connection" and "connection" mentioned in this application, unless otherwise specified, include both direct and indirect connections (connections).

The purpose of this application is to perform a prediction and match on the power consumption load of the microgrid, so as to achieve the effect of intelligent power dispatching, thereby saving more energy and realizing low carbon and environmental protection. Specifically, the application believes that although a microgrid has high and low electricity loads in different time periods, after research, the application believes that the electricity consumption in the microgrid is regular. Of course, different microgrids consume electricity. The laws are different. For example, a hotel may have a high electricity load at night and a low electricity load during the day, while a factory may have a high electricity load during the day and a low electricity load at night. The time is extended to a week, and the electricity load is also regular. , this application has found that the electricity consumption at each moment is related to the previous period of time, and is related to the electricity consumption at the same moment in the previous few days. Then, the future electricity consumption situation is predicted through the previous electricity consumption situation, so as to further match and dispatch the electric energy of the microgrid.

The present application will be specifically described below.

In some embodiments of the present application, an electric load prediction device is disclosed. Referring to FIG. 1 , in some embodiments, the electricity load prediction apparatus includes a sensor 10 and a processor 30 , which will be described in detail below.

The sensor 10 is used to obtain electricity consumption data for a period of time. The number of sensors 10 may be one or more. In some embodiments, the electrical data may include one, two, or three of power, corresponding voltage, and current. In some embodiments, the power consumption data may further include corresponding voltage phase angles or current phase angles.

In some embodiments, the sensor 10 may include a Phasor Measurement Unit (PMU). The PMU is a phasor measurement unit composed of a GPS (Global Positioning System, Global Positioning System) second pulse as a synchronous clock. The invention uses the PMU to measure the voltage vector of each node of the power grid in the transient process, so that the PMU is applied to the fields of dynamic monitoring, state estimation, system protection, regional stability control, system analysis and prediction of the power grid, so as to ensure the safe operation of the power grid. .

In the present invention, one or more PMUs are installed in the microgrid to construct a dynamic monitoring system for the microgrid, and the power consumption data of the microgrid, such as generator power angle, internal potential, and three-phase fundamental wave voltage at the machine terminal, are uploaded through the PMU. Phasor, machine terminal fundamental wave positive sequence voltage phasor, machine terminal three-phase fundamental wave current phasor, machine terminal fundamental wave positive sequence current phasor, active power, reactive power, excitation current, excitation voltage and rotor speed, etc.

In some examples, the PMU may report data every 10ms, and collect data at a relatively high frequency, thereby obtaining a large amount of data for training and learning of the following models. For example, GSA (Grid State Analyser) developed and formulated by Jianke Yunzhi (Shenzhen) Technology Co., Ltd. can be used. Its phase acquisition principle is similar to that of traditional PMU, and its acquisition accuracy is better than that of traditional PMU. The GSA of Jianke Yunzhi (Shenzhen) Technology Co., Ltd. can obtain various parameters of grid nodes with high precision, such as AC and DC voltage and current, as well as AC frequency and global phase angle, etc., and make preliminary online grid nodes in real time. State estimation. In some examples, the above-mentioned GSA also integrates the following processor 30, so that it can conduct in-depth analysis of the current state of the power grid. The exact state of the node and estimated future state.

These are some descriptions of the sensor 10 .

The processor 30 is configured to predict the electricity consumption situation at a moment in the future according to the electricity consumption data acquired by the sensor 10 for a period of time, and the electricity consumption situation includes the predicted value, or the predicted value and its probability distribution range. For example, taking power as an example, the processor 30 is configured to predict the electricity consumption at a moment in the future according to the electricity consumption data acquired by the sensor 10 for a period of time, where the electricity consumption includes the predicted value of the power at a moment in the future, or The predicted value of the power at a future moment and its probability distribution range. In some embodiments, the processor 30 is configured to predict the electricity consumption at a future moment, such as power, normal value, minimum value and maximum value at a future moment, according to the electricity consumption data acquired by the sensor 10 for a period of time. In some embodiments, the processor 30 is configured to predict the electricity consumption at a future moment according to the electricity consumption data acquired by the sensor 10 for a period of time - such as power, the normal value and its probability, minimum value and its probability and its maximum value and its probability. The specific description is given below.

In some embodiments, the processor 30 is configured to input the electricity consumption data of the period of time into a pre-established electricity consumption load prediction model, so as to predict the electricity consumption situation at a moment in the future relative to the period of time, using Electrical conditions include normal, minimum and maximum values. For example, the processor 30 obtains the electricity consumption data from 9:50 am to 55 am on the current day, and predicts the electricity consumption situation at 10 am. Let us take the power as an example, that is, predict the normal value, the minimum value and the maximum value of the power at 10 am. . In some embodiments, the processor 30 further performs matching and scheduling on the power of the microgrid according to the predicted power consumption at a moment in the future relative to the period of time.

The establishment of the electricity load prediction model will be described below. In some embodiments, the electricity load prediction model may be established in the following ways:

(1) Obtain a training set, the data in the training set is the electricity consumption data of the microgrid for a period of time, and the labels of the data are the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data of several recent times at the same time. The maximum value of the electricity data, and the minimum value of the electricity consumption data of several recent times at the same time.

In some embodiments, the power consumption data at the same moment in the most recent several days includes the power consumption data at the same moment in the most recent days, or the power consumption data at the same moment in the same week in the most recent weeks. For example, several recent identical moments here are the same moment relative to the above-mentioned future moment, which can be the same moment in the last N days, or the same moment in the same week in the last N weeks, and N can be set by the user.

For example, let's take data A as the data in the training set as an example, then data A is the electricity consumption data of a certain Wednesday from T11 to T12 in the microgrid, and the label of the data is the future one of the relative time period T1 to T2. Power consumption data at time T13, maximum power consumption data and minimum power consumption data at time T13 of the last N days (for example, Saturday, Sunday, this week, Monday and Tuesday this week).

For another example, let's take data B as the data in the training set as an example, then data B is the electricity consumption data of a certain Thursday in the microgrid from T21 to T22, and the data label is the relative time period T21 to T22. The electricity consumption data at T23 in the future, the maximum electricity consumption data and the minimum electricity consumption data at the time T3 of the same week in the last N weeks (for example, Thursday last week, Thursday last week).

(2) Using the above training set, the electricity load prediction model is obtained by training. Specifically, it can be trained in this way: pre-build a prediction model based on integrated deep learning. The prediction model first decomposes the input electricity data through the empirical mode decomposition algorithm to obtain sub-signals of different frequencies, and then applies the deep recurrent neural network to Each sub-signal is analyzed and predicted, and the output obtained after each sub-signal is analyzed and predicted is integrated as the predicted electricity consumption; the prediction model trained by the training set is used as the electricity load prediction model .

以及R1；对W1进行经验模式分解得到

以及Rm；再应用长短期网络(Long short-term memory，LSTM)对各个子信号分析得到各自的预测结果，例如应用长短期网络

对

分析得到

应用长短期网络

对

分析得到

应用长短期网络

对R1分析得到

应用长短期网络

对

分析得到

应用长短期网络

对

分析得到

应用长短期网络

对Rm分析得到

再将各个子信号的预测结果当作另一个长短期神经网络LSTM的输入，训练得到最后的预测结果(Prediction Results，PR)。Specifically, please refer to FIG. 2 , which is an algorithm flow chart of the electricity load prediction model. Take electricity consumption data as power as an example. In the figure, the time series signal (TSS) refers to the power signal. Wavelet transform (DiscreteWaveletTransformation, DWT) obtains the items W1, W2, ...., Wm; and then perform empirical mode decomposition on the items W1, W2, ...., Wm, respectively, to obtain several sub-signals (for example, Fig. IMF, R), specifically, the empirical mode decomposition of W1 gets

and R1; empirical mode decomposition of W1 is obtained

and Rm; then apply the long short-term network (Long short-term memory, LSTM) to analyze each sub-signal to obtain the respective prediction results, such as applying the long and short-term network

right

Analysis got

Applying long and short term networks

right

Analysis got

Applying long and short term networks

Analysis of R1 gets

Applying long and short term networks

right

Analysis got

Applying long and short term networks

right

Analysis got

Applying long and short term networks

By analyzing Rm, we get

Then, the prediction results of each sub-signal are used as the input of another long-term and short-term neural network LSTM, and the final prediction results (Prediction Results, PR) are obtained by training.

In other examples, please refer to FIG. 3 , the prediction results of each sub-signal and the corresponding additional features (Additional Features, AF) can also be input into the long-term and short-term neural network LSTM for training to obtain the final prediction results. Additional features here can be current, voltage and phase angle or phase angle difference etc. Take the phase angle as an example: power is divided into active power and reactive power, and the respective proportions are determined by the phase angle difference between voltage and current, so current, voltage and their respective phase angles can provide more detailed information for power calculation , so it is also helpful for power prediction in the above model.

The above (1) and (2) are trained through the training set to obtain the electricity load prediction model. In some embodiments, the model may also be corrected. For example, the electricity load prediction model is also established in the following ways:

(3) Acquire a test set, the data in the test set is the electricity consumption data of the microgrid for a period of time, and the labels of the data are the electricity consumption data at a moment in the future relative to the period of time, and the consumption data of several recent times at the same moment. The maximum value of the electricity data, and the minimum value of the electricity consumption data of several recent times at the same time. The meanings of the last several identical moments have been described in detail above, and will not be repeated here.

(4) Use the test set to verify the electricity load prediction model obtained by training the training set.

(5) Taking the error between the label of the data in the test set and the normal value of the electricity consumption predicted by the electricity load prediction model as the standard, perform hyperparameter tuning on the electricity consumption load prediction model to obtain the hyperparameter adjustment The optimized electricity load forecasting model.

Through (3), (4) and (5), the hyperparameter tuning of the electricity load prediction model trained by the above (1) and (2) is completed.

It should be noted that the data of the training set and the test set can also be collected by the sensor 10, for example, collected by the PMU mentioned above. After the electricity consumption data is collected, the electricity consumption data can also be cleaned, because the electricity consumption data obtained by the sensors may have problems such as missing, recording errors, or abnormal values (such as abnormal values caused by short circuits). . Then, a training set can be constructed based on the cleaned electricity consumption data, and a test set can also be constructed.

The present invention adopts the above-mentioned model algorithm and cooperates with the GSA developed and formulated by Jianke Yunzhi (Shenzhen) Technology Co., Ltd., which makes the prediction result have higher accuracy. In addition, it can be seen from the model algorithm and the actual effect verified by the inventor that the established electricity load prediction model can have good performance without complicated feature engineering. Taking the predicted power as an example, the present invention also predicts its upper and lower ranges when predicting the power. This upper and lower limit range can provide a reference for the electricity consumption unit, so that the electricity consumption unit can better match and dispatch the electric energy of the microgrid.

In some embodiments of the present invention, an intelligent power dispatching method for a micro-grid is also disclosed, which will be described in detail below.

Referring to FIG. 4 , the smart power dispatching method for a microgrid in an embodiment includes the following steps:

Step 100: Acquire electricity consumption data in the microgrid for a period of time.

In some embodiments, the electrical data may include one, two, or three of power, corresponding voltage, and current. In some embodiments, the power consumption data may further include corresponding voltage phase angles or current phase angles.

In some embodiments, step 100 may acquire power consumption data through a sensor 10 such as a phasor measurement unit. Specifically, one or more PMUs are installed in the microgrid to construct a dynamic monitoring system for the microgrid, and the power consumption data of the microgrid, such as generator power angle, internal potential, and three-phase fundamental wave at the machine end, are uploaded through the PMU. Voltage phasor, machine terminal fundamental wave positive sequence voltage phasor, machine terminal three-phase fundamental wave current phasor, machine terminal fundamental wave positive sequence current phasor, active power, reactive power, excitation current, excitation voltage and rotor speed, etc. . In some examples, the PMU may report data every 10ms, and collect data at a relatively high frequency, thereby obtaining a large amount of data for training and learning of the following models. For example, GSA (Grid State Analyser) developed and formulated by Jianke Yunzhi (Shenzhen) Technology Co., Ltd. can be used. Its phase acquisition principle is similar to that of traditional PMU, and its acquisition accuracy is better than that of traditional PMU. The GSA of Keyunzhi (Shenzhen) Technology Co., Ltd. can obtain various parameters of grid nodes with high precision, such as AC and DC voltage and current, as well as AC frequency and global phase angle, etc., and make the initial status of online grid nodes in real time. estimate. In some examples, the above-mentioned GSA also integrates the following processor 30, so that it can conduct in-depth analysis of the current state of the power grid. The exact state of the node and estimated future state.

Step 200: Input the power consumption data of the period of time in step 100 into a pre-established power consumption load prediction model to predict the power consumption situation at a moment in the future relative to the period of time, and the power consumption conditions include: Measured value, or predicted value and its probability distribution range. For example, taking power as an example, step 200 inputs the electricity consumption data of the period of time in step 100 into a pre-established electricity consumption load prediction model to predict the electricity consumption at a moment in the future relative to the period of time The electricity consumption situation includes the predicted value of the power at a future moment, or the predicted value of the power at a future moment and its probability distribution range. In some embodiments, step 200 inputs the electricity consumption data of the period of time in step 100 into a pre-established electricity consumption load prediction model, so as to predict the electricity consumption situation at a moment in the future relative to the period of time— - For example power, normal, minimum and maximum values for a future moment. In some embodiments, step 200 inputs the electricity consumption data of the period of time in step 100 into a pre-established electricity consumption load prediction model, so as to predict the electricity consumption situation at a moment in the future relative to the period of time— - For example power, normal value and its probability at a future moment, minimum value and its probability and maximum value and its probability.

Specifically, for example, in step 200, the electricity consumption data from 9:50 am to 55 am on the current day is obtained, and the electricity consumption at 10 am is predicted. Let us take the power as an example, that is, predict the normal value and minimum value of the power at 10 am. and the maximum value.

The establishment of the electricity load forecasting model will be described below.

In some embodiments, referring to FIG. 5 , the electricity load prediction model in step 200 may be established in the following manner:

Step 210: Acquire a training set, the data in the training set is the electricity consumption data of the microgrid for a period of time, and the labels of the data are the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data of several recent times at the same moment. The maximum value of the electricity data, and the minimum value of the electricity consumption data of several recent times at the same time.

In some embodiments, the power consumption data at the same moment in the most recent several days includes the power consumption data at the same moment in the most recent days, or the power consumption data at the same moment in the same week in the most recent weeks. For example, several recent identical moments here are the same moment relative to the above-mentioned future moment, which can be the same moment in the last N days, or the same moment in the same week in the last N weeks, and N can be set by the user.

For example, let's take data A as the data in the training set as an example, then data A is the electricity consumption data of a certain Wednesday from T11 to T12 in the microgrid, and the label of the data is the future one of the relative time period T1 to T2. Power consumption data at time T13, maximum power consumption data and minimum power consumption data at time T13 of the last N days (for example, Saturday, Sunday, this week, Monday and Tuesday this week).

For another example, let's take data B as the data in the training set as an example, then data B is the electricity consumption data of a certain Thursday in the microgrid from T21 to T22, and the data label is the relative time period T21 to T22. The electricity consumption data at T23 in the future, the maximum electricity consumption data and the minimum electricity consumption data at the time T3 of the same week in the last N weeks (for example, Thursday last week, Thursday last week).

Step 220: Use the above training set to obtain an electricity load prediction model through training. Specifically, it can be trained in this way: pre-build a prediction model based on integrated deep learning. The prediction model first decomposes the input electricity data through the empirical mode decomposition algorithm to obtain sub-signals of different frequencies, and then applies the deep recurrent neural network to Each sub-signal is analyzed and predicted, and the output obtained after each sub-signal is analyzed and predicted is integrated as the predicted electricity consumption; the prediction model trained by the training set is used as the electricity load prediction model .

For the algorithm of the electricity load prediction model in step 220, reference may be made to the description of the algorithm flow in FIG. 2 and FIG. 3 above, which will not be repeated here.

In the above steps 210 and 220, the electricity load prediction model is obtained through training on the training set. In some embodiments, the model may also be corrected. In some embodiments, referring to FIG. 6 , the electricity load prediction model trained in steps 210 and 220 is also corrected in the following manner:

Step 230: Acquire a test set, the data in the test set is the electricity consumption data of the microgrid for a period of time, and the labels of the data are the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data of several recent times at the same moment. The maximum value of the electricity data, and the minimum value of the electricity consumption data of several recent times at the same time. The meanings of the last several identical moments have been described in detail above, and will not be repeated here.

Step 240: Use the test set to verify the electricity load prediction model obtained by training on the training set.

Step 250: Using the error between the label of the data in the test set and the normal value of the electricity consumption predicted by the electricity load prediction model as a standard, perform hyperparameter tuning on the electricity consumption load prediction model to obtain hyperparameter tuning. The optimized electricity load forecasting model.

Through steps 230 to 250, hyperparameter tuning is completed for the electricity load prediction model trained in the above steps 210 and 220.

It should be noted that the data of the training set and the test set can also be collected by the sensor 10, for example, collected by the PMU mentioned above. After the electricity consumption data is collected, the electricity consumption data can also be cleaned, because the electricity consumption data obtained by the sensors may have problems such as missing, recording errors, or abnormal values (such as abnormal values caused by short circuits). . Then, a training set can be constructed based on the cleaned electricity consumption data, and a test set can also be constructed.

Step 300: According to the electricity consumption situation predicted in step 200 at a moment in the future relative to the period of time, matching and scheduling the electric energy of the microgrid. For example, if it is predicted that the peak period of electricity consumption is about to arrive, the microgrid will be controlled to increase the power supply; on the contrary, if it is predicted that the valley of electricity consumption will soon be reached, the microgrid will be controlled to reduce the power supply.

The present invention adopts the above-mentioned model algorithm and cooperates with the GSA developed and formulated by Jianke Yunzhi (Shenzhen) Technology Co., Ltd., which makes the prediction result have higher accuracy. In addition, it can be seen from the model algorithm and the actual effect verified by the inventor that the established electricity load prediction model can have good performance without complicated feature engineering. Taking the predicted power as an example, the present invention also predicts its upper and lower ranges when predicting the power. This upper and lower limit range can provide a reference for the electricity consumption unit, so that the electricity consumption unit can better match and dispatch the electric energy of the microgrid.

Descriptions are made herein with reference to various exemplary embodiments. However, those skilled in the art will recognize that changes and modifications may be made to the exemplary embodiments without departing from the scope of this document. For example, various operational steps, and components for performing operational steps, may be implemented in different ways depending on the particular application or considering any number of cost functions associated with the operation of the system (eg, one or more steps may be deleted, modified or incorporated into other steps).

In the above-mentioned embodiments, it may be implemented in whole or in part by software, hardware, firmware or any combination thereof. Additionally, as understood by those skilled in the art, the principles herein may be reflected in a computer program product on a computer-readable storage medium preloaded with computer-readable program code. Any tangible, non-transitory computer-readable storage medium may be used, including magnetic storage devices (hard disks, floppy disks, etc.), optical storage devices (CD to ROM, DVD, Blu Ray disks, etc.), flash memory, and/or the like . These computer program instructions may be loaded on a general purpose computer, special purpose computer or other programmable data processing apparatus to form a machine such that execution of the instructions on the computer or other programmable data processing apparatus may generate means for implementing the specified functions. These computer program instructions may also be stored in a computer-readable memory that instructs a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable memory form a piece of Articles of manufacture, including implementing means for implementing specified functions. Computer program instructions may also be loaded on a computer or other programmable data processing device to perform a series of operational steps on the computer or other programmable device to produce a computer-implemented process such that a process executed on the computer or other programmable device Instructions may provide steps for implementing specified functions.

Although the principles herein have been shown in various embodiments, many modifications may be made in structure, arrangement, proportions, elements, materials and components as are particularly suited to particular environmental and operating requirements without departing from the principles and scope of the present disclosure use. The above modifications and other changes or corrections are intended to be included within the scope of this document.

The foregoing Detailed Description has been described with reference to various embodiments. However, those skilled in the art will recognize that various modifications and changes can be made without departing from the scope of the present disclosure. Accordingly, this disclosure is to be considered in an illustrative rather than a restrictive sense, and all such modifications are intended to be included within its scope. Likewise, the advantages, other advantages, and solutions to problems of the various embodiments have been described above. However, the benefits, advantages, solutions to the problems, and any elements that give rise to these, or make them more explicit, should not be construed as critical, necessary, or essential. As used herein, the term "comprising" and any other variations thereof are non-exclusive inclusion, such that a process, method, article or device that includes a list of elements includes not only those elements, but also not expressly listed or part of the process , method, system, article or other elements of a device. Furthermore, as used herein, the term "coupled" and any other variations thereof refer to physical connections, electrical connections, magnetic connections, optical connections, communication connections, functional connections, and/or any other connection.

Those skilled in the art will recognize that many changes may be made to the details of the above-described embodiments without departing from the underlying principles of the invention. Accordingly, the scope of the invention should be determined only by the claims.

Claims (10)

1. An electrical load forecasting device, comprising: Sensors, used to obtain electricity consumption data for a period of time; The processor is configured to input the electricity consumption data of the period of time into a pre-established electricity consumption load prediction model to predict the electricity consumption situation at a moment in the future relative to the period of time, and the electricity consumption situation includes the normal value , minimum and maximum values; The electricity load prediction model is established in the following ways: Obtain a training set, the data in the training set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the most recent electricity consumption data at the same moment. value, the minimum value of electricity consumption data at the same time in several recent times; Using the training set, the electricity load prediction model is obtained by training; specifically, it includes: constructing a prediction model based on integrated deep learning in advance. The prediction model first decomposes the input electricity consumption data through an empirical mode decomposition algorithm to obtain Sub-signals of different frequencies, and then apply the deep recurrent neural network to analyze and predict each sub-signal, and then integrate the output obtained after analyzing and predicting each sub-signal, as the predicted electricity consumption; will be trained through the training set The latter prediction model is used as the electricity load prediction model. The electricity load prediction model is also established in the following ways: Obtain a test set, the data in the test set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data at the same moment in recent times is the largest value, the minimum value of electricity consumption data at the same time in several recent times; Use the test set to verify the electricity load prediction model obtained by training the training set; take the error between the label of the data in the test set and the normal value of the electricity consumption situation predicted by the electricity load prediction model as the standard, Perform hyperparameter tuning on the power consumption load prediction model to obtain a power consumption load prediction model after hyperparameter optimization. 2 . The power consumption load forecasting device according to claim 1 , wherein the power consumption situation at least includes power. 3 . 3 . The power consumption load prediction device according to claim 1 , wherein the power consumption data of the microgrid for a period of time includes power, corresponding voltage, current and phase angle. 4 . 4. The power consumption load forecasting device according to claim 1 or 2, wherein the power consumption data at the same time in the last several days includes the power consumption data at the same time in the most recent days, or the same week in the most recent weeks. electricity consumption data at the same time. 5. An intelligent power dispatching method for a microgrid, comprising: Obtain electricity consumption data for a period of time in the microgrid; Input the power consumption data of the period of time into a pre-established power consumption load prediction model to predict the power consumption situation at a moment in the future relative to the period of time, and the power consumption situation includes the predicted value, or the predicted value and its probability distribution range; According to the predicted electricity consumption at a moment in the future relative to the period of time, the electric energy of the microgrid is matched and dispatched; The electricity load prediction model is established in the following ways: Obtain a training set, the data in the training set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the most recent electricity consumption data at the same moment. value, the minimum value of electricity consumption data at the same time in several recent times; Using the training set, the electricity load prediction model is obtained by training; specifically, it includes: constructing a prediction model based on integrated deep learning in advance. The prediction model first decomposes the input electricity consumption data through an empirical mode decomposition algorithm to obtain Sub-signals of different frequencies, and then apply the deep recurrent neural network to analyze and predict each sub-signal, and then integrate the output obtained after analyzing and predicting each sub-signal, as the predicted electricity consumption; will be trained through the training set The latter prediction model is used as the electricity load prediction model. The electricity load prediction model is also established in the following ways: Obtain a test set, the data in the test set is the electricity consumption data of the microgrid for a period of time, and the label of the data is the electricity consumption data at a moment in the future relative to the period of time, and the electricity consumption data at the same moment in recent times is the largest value, the minimum value of electricity consumption data at the same time in several recent times; Use the test set to verify the electricity load prediction model obtained through the training set; Taking the error between the label of the data in the test set and the normal value of the electricity consumption predicted by the electricity load prediction model as the standard, the hyperparameter tuning of the electricity load prediction model is carried out to obtain the hyperparameter tuning. Electricity load forecasting model. 6 . The smart power dispatching method according to claim 5 , wherein the power consumption situation includes at least a predicted value of power, or a predicted value of power and its probability distribution range. 7 . 7 . The intelligent power dispatching method according to claim 6 , wherein the predicted value of the power includes a maximum value, a minimum value and a normal value of the power. 8 . 8 . The smart power scheduling method according to claim 5 , wherein the power consumption data of the microgrid for a period of time includes power, corresponding voltage, current and phase angle. 9 . 9. The intelligent power dispatching method according to any one of claims 5 to 8, wherein the power consumption data at the same time in the last several days includes the power consumption data at the same time in the most recent days, or the most recent power consumption data at the same time. Electricity consumption data at the same time in the same week of the week. 10. A computer-readable storage medium comprising a program executable by a processor to implement the method according to any one of claims 5 to 9.

Images