CN115545354A - Artificial intelligence load prediction method and system for intelligent Internet of things platform - Google Patents
Artificial intelligence load prediction method and system for intelligent Internet of things platform Download PDFInfo
- Publication number
- CN115545354A CN115545354A CN202211512758.4A CN202211512758A CN115545354A CN 115545354 A CN115545354 A CN 115545354A CN 202211512758 A CN202211512758 A CN 202211512758A CN 115545354 A CN115545354 A CN 115545354A
- Authority
- CN
- China
- Prior art keywords
- data
- sequence data
- model
- artificial intelligence
- load prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 29
- 238000012216 screening Methods 0.000 claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 238000010276 construction Methods 0.000 claims abstract description 10
- 230000000875 corresponding effect Effects 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 5
- 230000002596 correlated effect Effects 0.000 claims description 4
- 230000005611 electricity Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Biophysics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of artificial intelligence load prediction, and particularly relates to an artificial intelligence load prediction method and system for an intelligent Internet of things platform, wherein the method comprises the following steps: processing the acquired data to obtain sequence data; constructing a model and carrying out model training on the sequence data; carrying out feature screening on the trained data; and predicting the power load according to the data after the characteristic screening. The system comprises a data acquisition unit, a model construction unit, a feature screening unit and a prediction unit. The method and the device realize accurate prediction of the future power load by acquiring the load data, using the CNN to perform feature extraction, screening and data enhancement, and performing internal association on the current data and the past and future load data through the BilSTM, have stronger processing capability on a complex nonlinear system, and are suitable for short-term load prediction with stronger randomness and volatility.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence load prediction, and particularly relates to an artificial intelligence load prediction method and system for an intelligent internet of things platform.
Background
At present, a power grid company needs to realize digital transformation of production, operation and management, and the digitization of each professional line needs to be supported, wherein the four aspects of object-object interconnection, object-object mutual control, data sharing and full-service online are basic requirements of the digital transformation of the company, and the intelligent Internet of things platform is pregnant with the basic requirements. The intelligent internet of things platform has the characteristics of flexible and various communication, flexible and flexible interface, intelligent and efficient addressing, real-time acquisition, fusion and sharing, massive data storage throughput, real-time acquisition, fusion and sharing and flexible interface.
At present, the problem of insufficient power supply often occurs during the peak period of power utilization, and power utilization data are acquired fussy and are processed to consume a large amount of manpower and material resources, so that the traditional power load prediction algorithm is difficult to adapt to the power load under the new potential. The power load prediction is an important component of power system planning and is the basis of the economic operation of a power grid. The method is divided into different prediction objects, and the power load prediction mainly comprises prediction of future power demand, prediction of future power consumption and prediction of load curves. By predicting the time distribution and the space distribution of the future power load, an effective decision basis is provided for power planning operation.
The existing power grid load prediction methods mainly include methods based on mathematical statistics, such as multivariate linear regression, kalman filtering, exponential smoothing and the like, and although the methods are simple in calculation and high in calculation efficiency, the methods have poor processing capability on a complex nonlinear system and are not suitable for short-term load prediction with strong randomness and volatility. The other method is a power grid load prediction method based on machine learning, wherein the short-term load prediction method based on the TCA-CNN-LSTM builds a model by combining a CA-CNN module and a TA-LSTM module, and realizes the reinforcement of feature expression and the fine depiction of load time sequence data.
In order to solve the above problems, it is necessary to design an artificial intelligence load prediction method and system for an intelligent internet of things platform.
Disclosure of Invention
In order to solve the above problems, the present invention provides an artificial intelligence load prediction method for an intelligent internet of things platform, the method comprising:
processing the acquired data to obtain sequence data;
constructing a model and carrying out model training on the sequence data;
carrying out feature screening on the trained data;
and predicting the power load according to the data after the characteristic screening.
Preferably, the collected data includes current time, load, electricity price, temperature and humidity;
wherein the temperatures include dew point temperature, dry bulb temperature, and wet bulb temperature.
Preferably, the processing the acquired data comprises the following steps:
sequencing the acquired data according to the sequence of time points to form a data queue;
extracting a plurality of data with the length of m from the data queue as feature data;
inputting the characteristic data into a BilSTM module to obtain a load predicted value at the next moment;
moving the characteristic data forward to form new characteristic data with the same length, and predicting the load predicted value at the next moment again until all data are traversed;
and sorting all the load predicted values to form sequence data in continuous sub-time.
Preferably, the characteristic data forwarding specifically includes: and taking a plurality of data with the length of m from the data of the second time point in the current characteristic data as a starting point to form new characteristic data.
Preferably, the model training of the sequence data comprises the steps of:
inputting the sequence data into a model, and performing feature extraction on the sequence data through two layers of one-dimensional convolution layers;
carrying out dimensionality reduction and filtration on the sequence data subjected to feature extraction through a pooling layer and a Flatten layer;
inputting the sequence data into a BilSTM layer, and performing correlation processing on the sequence data through the BilSTM layer;
the sequence data enters a Dense layer to carry out feature enhancement;
and outputting a model training result.
Preferably, the step of feature extraction comprises: the sequence data is mapped in a one-dimensional convolutional layer in a high-dimensional manner.
Preferably, the associating the sequence data through the BilSTM layer comprises the following steps:
carrying out recursive feedback on the hidden layer state of the past data and the hidden layer state of the future data in the sequence data;
the current data, past data, and future data in the sequence data are inherently correlated.
Preferably, the hidden layer state is formed by combining hidden layer output at a previous moment of forward propagation, hidden state at a previous moment of backward propagation, and input at a current moment in the BiLSTM layer, and an expression formula of the hidden layer state is as follows:
wherein,
indicating the state of the forward hidden layer,
a backward hidden layer state is represented,
representing the weight of the hidden layer of the forward propagating unit,
representing the weight of the hidden layer of the back propagation unit,
indicating the bias of the hidden layer at the current moment,
an input representing the current time of day is presented,
representing the hidden layer state at the previous instant of the back propagation,
the hidden layer output at the previous time point representing forward propagation,
the overall hidden state is obtained through front and back bidirectional calculation.
Preferably, the associating the sequence data through the BilSTM layer further comprises:
applying an attention mechanism to highlight key information in the sequence data by weight configuration;
data characteristics dependent on more than a predetermined time are mined from the sequence data.
Preferably, the model is constructed and the performance of the model is evaluated effectively.
Preferably, the formula of the effectiveness evaluation is:
wherein,
the root mean square error is represented as a function of,
the mean absolute percentage error is expressed as a percentage error,
the true value of the ith sample is represented,
represents the predicted value of the ith sample,
indicating the number of sample points.
Preferably, the feature screening comprises: removing irrelevant feature data in the sequence data according to an L1 regularization formula;
the L1 regularization formula is
；
Wherein,
，
respectively a training sample and its corresponding label,
in order to be a weight parameter, the weight parameter,
in order to be the objective function, the target function,
in order to be a penalty term,
in order to obtain the penalty term coefficient,
the target function after the irrelevant features are removed.
The invention also provides an artificial intelligence load prediction system facing the intelligent Internet of things platform, which comprises a data acquisition unit, a model construction unit, a feature screening unit and a prediction unit;
the data acquisition unit is used for processing the acquired data to obtain sequence data;
the model building unit is used for building a model and carrying out model training on the sequence data;
the characteristic screening unit is used for screening the characteristics of the trained data;
and the prediction unit is used for predicting the power load according to the data after the characteristic screening.
Preferably, the data acquisition unit is configured to process the acquired data to obtain sequence data, and the sequence data includes:
the data acquisition unit is used for sequencing the acquired data according to the sequence of time points to form a data queue;
extracting a plurality of data with the length of m from the data queue as feature data;
inputting the characteristic data into a BilSTM module to obtain a load predicted value at the next moment;
moving the characteristic data forward to form new characteristic data with the same length, and predicting the load predicted value at the next moment again until all data are traversed;
and sorting all the load predicted values to form sequence data in continuous sub-time.
Preferably, the model building unit is configured to build a model and train the model of the sequence data, and includes:
the model construction unit is used for inputting the sequence data into a model and extracting the characteristics of the sequence data through two layers of one-dimensional convolution layers;
carrying out dimensionality reduction and filtration on the sequence data subjected to feature extraction through a pooling layer and a Flatten layer;
inputting the sequence data into a BilSTM layer, and performing correlation processing on the sequence data through the BilSTM layer;
the sequence data enters a Dense layer to carry out feature enhancement;
and outputting a model training result.
Preferably, the model construction unit is configured to perform feature extraction on the sequence data, and includes:
the model building unit is used for performing high-dimensional mapping on the sequence data in the one-dimensional convolutional layer.
Preferably, the model building unit is configured to perform association processing on the sequence data through a BilSTM layer, and includes:
the model building unit is used for carrying out recursive feedback on the states of a hidden layer of past data and a hidden layer of future data in the sequence data;
the current data, past data, and future data in the sequence data are inherently correlated.
Preferably, the feature screening unit is configured to perform feature screening on the trained data, and includes:
and the characteristic screening unit is used for removing irrelevant characteristic data in the sequence data according to the L1 regularization formula.
Preferably, the L1 regularization formula is
；
Wherein,
，
respectively a training sample and its corresponding label,
in order to be a weight parameter, the weight parameter,
in order to be the objective function, the target function,
in order to be a penalty term,
in order to make the penalty term coefficient,
the target function after the irrelevant features are removed.
The invention has the following beneficial effects:
(1) The method comprises the steps of firstly, acquiring load data, using CNN to carry out feature extraction, screening and data enhancement, and realizing accurate prediction of future power load through the internal association of BilSTM on current data and past and future load data;
(2) The CNN-BilSTM is applied to the load prediction of the smart power grid, the problems that the original CNN-LSTM model is poor in performance in time sequence capture and the model prediction effect is unstable are solved, the load prediction efficiency is improved, and the efficient operation of the power grid is guaranteed.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a flowchart illustrating an artificial intelligence load prediction method for an intelligent IOT platform according to an embodiment of the present invention;
FIG. 2 shows a schematic diagram of data acquisition in an embodiment of the invention;
FIG. 3 illustrates a flow diagram for model building in an embodiment of the invention;
FIG. 4 shows a BiLSTM layer network structure diagram in an embodiment of the invention;
FIG. 5 is a graph showing the fitting effect of training and testing results in an embodiment of the present invention;
fig. 6 is a diagram illustrating an artificial intelligence load prediction system oriented to an intelligent internet of things platform according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the present invention provides an artificial intelligence load prediction method for an intelligent internet of things platform, the method includes: processing the acquired data to obtain sequence data; constructing a model and performing model training on the sequence data, wherein the model is constructed by adopting a CNN-BilSTM structure; carrying out feature screening on the trained data; and predicting the power load according to the data after the characteristic screening.
The acquired data comprises current time, load, electricity price, temperature and humidity; wherein the temperatures include dew point temperature, dry bulb temperature, and wet bulb temperature.
The processing of the collected data comprises the following steps: sequencing the acquired data according to the sequence of time points to form a data queue; extracting a plurality of data with the length of m from the data queue as feature data; inputting the characteristic data into a BilSTM module to obtain a load predicted value at the next moment; moving the characteristic data forward to form new characteristic data with the same length, and predicting the load predicted value at the next moment again until all data are traversed; the characteristic data forward specifically comprises the following steps: taking data at a second time point in the current characteristic data as a starting point, and taking a plurality of data with the length of m to form new characteristic data; and sorting all the load predicted values to form sequence data in continuous sub-time. As shown in fig. 2, at time t, the small box in the dashed line frame represents input data of one time node, the small box on the right side of the dashed line frame represents output data, the data of the time node represented by the small box in the dashed line frame is used as data in the input prediction small box, and the whole is moved forward by one time node at time t +1 until the whole data is predicted.
As shown in fig. 3, the model training of the sequence data comprises the following steps:
inputting the sequence data into a model, and performing feature extraction on the sequence data through two layers of one-dimensional convolution layers; the step of feature extraction comprises: performing high-dimensional mapping on the sequence data in the one-dimensional convolutional layer; in this embodiment, the number of convolution kernels to be convolutional layers is 64, and the step length is 2;
carrying out dimensionality reduction and filtration on the sequence data subjected to feature extraction through a pooling layer and a Flatten layer; the pooling layer can analyze and reduce the dimension of the original data, so that the over-fitting problem of the model can be effectively relieved, and the extraction efficiency of the data characteristics can be effectively improved;
inputting the sequence data into a BilSTM layer, and performing correlation processing on the sequence data through the BilSTM layer, wherein in the embodiment, the BilSTM layer contains 20 neurons; as shown in fig. 4, the associating process of the sequence data includes the following steps: carrying out recursive feedback on the hidden layer state of the past data and the hidden layer state of the future data in the sequence data; for current data in sequence dataIntrinsic correlation of past and future data; the hidden layer state is formed by combining hidden layer output at the previous moment of forward propagation, hidden state at the previous moment of backward propagation and input at the current moment in the BilSTM layer, and the expression formula of the hidden layer state is as follows:
；
wherein,
indicating the state of the forward hidden layer,
a backward hidden layer state is represented,
representing the weight of the hidden layer of the forward propagation unit,
representing the weight of the hidden layer of the back propagation unit,
indicating the bias of the hidden layer at the current time,
an input indicative of a current time of day,
representing the hidden layer state at the previous instant of the back propagation,
the hidden layer output at the previous time point representing forward propagation,
the overall hidden state is obtained through front and back bidirectional calculation. In particular, the input at the previous moment
Past the last moment forward hidden state
Or a backward hidden state at the next moment
Output is as
Input of the current time
Past the forward hidden state at the current time
Or a backward hidden state at the current time
Output is as
Input of the next time
Past the next moment forward hidden state
Or a last-time backward hidden state
Output is as
. The associating the sequence data through the BilSTM layer further comprises the following steps: applying an attention mechanism to highlight critical information in sequence data by weight configuration(ii) a Mining data characteristics which depend on more than a preset time in the sequence data; in the embodiment, the preset time can be adjusted according to the actual situation;
the sequence data enters a Dense layer to carry out feature enhancement;
and outputting a model training result.
After the model is built, the effectiveness evaluation of the performance of the model is also carried out; the formula of the effectiveness evaluation is as follows:
The feature screening comprises the following steps: removing irrelevant feature data in the sequence data according to an L1 regularization formula;
the L1 regularization formula is
；
Wherein,
，
respectively a training sample and its corresponding label,
in order to be a weight parameter, the weight parameter,
in order to be the objective function, the target function,
in order to be a penalty term,
in order to make the penalty term coefficient,
the objective function after the irrelevant features are removed.
As shown in fig. 5, the present embodiment selects the power data from 2006-2010 in australia as experimental data, wherein the data characteristics include date, time, temperature, electricity price and power load, wherein the temperature includes dry bulb, dew point and humidity. The acquisition frequency was half an hour. The experimental parameters were set as: the time step is 18, the number of nodes of the two-way LSTM memory layer is 128, and it is found through experiments that the loss function is stable when the number of parameters batchsize which are passed to the program for training is 108. Algorithm optimization was performed by adaptive moment estimation Adam and an attention mechanism was applied, where training and test results of the data were fitted. It can be seen from the figure that after more than 9 iterations, the root mean square error of the training set and the test set measures the deviation between the observed value and the true value to be consistent gradually.
As shown in fig. 6, the present invention further provides an artificial intelligence load prediction system for an intelligent internet of things platform, wherein the system includes a data acquisition unit, a model construction unit, a feature screening unit and a prediction unit;
the data acquisition unit is used for acquiring data and processing the acquired data to obtain sequence data; the data acquisition unit is used for processing the acquired data to obtain sequence data and comprises: the data acquisition unit is used for sequencing the acquired data according to the sequence of time points to form a data queue; extracting a plurality of data with the length of m from the data queue as feature data; inputting the characteristic data into a BilSTM module to obtain a load predicted value at the next moment; moving the characteristic data forward to form new characteristic data with the same length, and predicting the load predicted value at the next moment again until all data are traversed; sorting all the load predicted values to form sequence data in continuous sub-time;
the model building unit is used for building a model and carrying out model training on the sequence data; the model construction unit is used for constructing a model and performing model training on the sequence data and comprises the following steps: the model construction unit is used for inputting the sequence data into a model and extracting the characteristics of the sequence data through two layers of one-dimensional convolution layers; carrying out dimensionality reduction and filtration on the sequence data subjected to feature extraction through a pooling layer and a Flatten layer; the model building unit is used for carrying out association processing on the sequence data through the BilSTM layer, and comprises the following steps: the model building unit is used for carrying out recursive feedback on the states of a hidden layer of past data and a hidden layer of future data in the sequence data; correlating internally current data, past data, and future data in the sequence data; the sequence data enters a Dense layer to carry out feature enhancement; outputting a model training result; the model construction unit is used for extracting the characteristics of the sequence data and comprises the following steps: the model building unit is used for carrying out high-dimensional mapping on the sequence data in the one-dimensional convolutional layer;
the characteristic screening unit is used for screening the characteristics of the trained data; wherein, the feature screening unit is used for carrying out feature screening on the trained data and comprises: the characteristic screening unit is used for removing irrelevant characteristic data in the sequence data according to the L1 regularization formula; the L1 regularization formula is
(ii) a Wherein,
，
respectively a training sample and its corresponding label,
in order to be a weight parameter, the weight parameter,
in order to be the objective function, the target function,
in order to be a penalty term,
in order to make the penalty term coefficient,
the target function after the irrelevant features are removed.
And the prediction unit is used for predicting the power load according to the data after the characteristic screening.
Those of ordinary skill in the art will understand that: although the present invention has been described in detail with reference to the foregoing embodiments, modifications may be made to the technical solutions described in the foregoing embodiments, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (19)
1. An artificial intelligence load prediction method for an intelligent internet of things platform is characterized by comprising the following steps:
processing the acquired data to obtain sequence data;
constructing a model and carrying out model training on the sequence data;
carrying out feature screening on the trained data;
and predicting the power load according to the data after the characteristic screening.
2. The intelligent load prediction method for intelligent IOT platform as recited in claim 1,
the acquired data comprises current time, load, electricity price, temperature and humidity;
wherein the temperatures include dew point temperature, dry bulb temperature, and wet bulb temperature.
3. The method of claim 1, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the processing of the acquired data comprises the following steps:
sequencing the acquired data according to the sequence of time points to form a data queue;
extracting a plurality of data with the length of m from the data queue as characteristic data;
inputting the characteristic data into a BilSTM module to obtain a load predicted value at the next moment;
moving the characteristic data forward to form new characteristic data with the same length, and predicting the load predicted value at the next moment again until all data are traversed;
and sorting all the load predicted values to form sequence data in continuous sub-time.
4. The method of claim 3, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the characteristic data forward specifically comprises the following steps: and taking a plurality of data with the length of m from the data of the second time point in the current characteristic data as a starting point to form new characteristic data.
5. The method of claim 1, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the model training of the sequence data comprises the following steps:
inputting the sequence data into a model, and performing feature extraction on the sequence data through two layers of one-dimensional convolution layers;
carrying out dimensionality reduction and filtration on the sequence data subjected to feature extraction through a pooling layer and a Flatten layer;
inputting the sequence data into a BilSTM layer, and performing correlation processing on the sequence data through the BilSTM layer;
the sequence data enters a Dense layer to carry out feature enhancement;
and outputting a model training result.
6. The method of claim 5, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the step of feature extraction comprises: the sequence data is mapped in a one-dimensional convolutional layer in a high-dimensional manner.
7. The method of claim 5, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the correlation processing of the sequence data through the BilSTM layer comprises the following steps:
carrying out recursive feedback on the hidden layer state of the past data and the hidden layer state of the future data in the sequence data;
the current data, past data, and future data in the sequence data are inherently correlated.
8. The method of claim 7, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the hidden layer state is formed by combining hidden layer output at the previous moment of forward propagation, hidden state at the previous moment of backward propagation and input at the current moment in the BilSTM layer, and the expression formula of the hidden layer state is as follows:
；
wherein,
indicating the state of the forward hidden layer,
a backward hidden layer state is represented,
representing the weight of the hidden layer of the forward propagation unit,
representing the weight of the hidden layer of the back propagation unit,
indicating the bias of the hidden layer at the current moment,
an input representing the current time of day is presented,
representing the hidden layer state at the previous instant of the back propagation,
the hidden layer output at the previous time point representing forward propagation,
the overall hidden state is obtained through front and back bidirectional calculation.
9. The method of claim 7, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the associating the sequence data through the BilSTM layer further comprises the following steps:
applying an attention mechanism to highlight key information in the sequence data by weight configuration;
data characteristics dependent for more than a predetermined time are mined from the sequence data.
10. The method of claim 1, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
and after the model is built, the effectiveness evaluation of the performance of the model is also carried out.
11. The method of claim 10, wherein the load prediction model of the intelligent IOT platform,
the formula of the effectiveness evaluation is as follows:
wherein,
the root mean square error is represented as a function of,
the mean absolute percentage error is expressed as a percentage error,
the true value of the i-th sample is represented,
represents the predicted value of the ith sample,
indicating the number of sample points.
12. The method of claim 1, wherein the artificial intelligence load prediction method for the intelligent IOT platform,
the feature screening comprises the following steps: removing irrelevant feature data in the sequence data according to an L1 regularization formula;
the L1 regularization formula is
；
Wherein,
，
respectively a training sample and its corresponding label,
in order to be a weight parameter, the weight parameter,
in order to be the objective function, the target function,
in order to be a penalty term,
in order to make the penalty term coefficient,
the target function after the irrelevant features are removed.
13. An artificial intelligence load prediction system for an intelligent Internet of things platform is characterized by comprising a data acquisition unit, a model construction unit, a feature screening unit and a prediction unit;
the data acquisition unit is used for processing the acquired data to obtain sequence data;
the model building unit is used for building a model and carrying out model training on the sequence data;
the characteristic screening unit is used for screening the characteristics of the trained data;
and the prediction unit is used for predicting the power load according to the data after the characteristic screening.
14. The system of claim 13, wherein the artificial intelligence load prediction system for the intelligent IOT platform,
the data acquisition unit is used for processing the acquired data to obtain sequence data and comprises:
the data acquisition unit is used for sequencing the acquired data according to the sequence of time points to form a data queue;
extracting a plurality of data with the length of m from the data queue as feature data;
inputting the characteristic data into a BilSTM module to obtain a load predicted value at the next moment;
moving the characteristic data forward to form new characteristic data with the same length, and predicting the load predicted value at the next moment again until all data are traversed;
and sorting all the load predicted values to form sequence data in continuous sub-time.
15. The system of claim 13, wherein the artificial intelligence load prediction system for the intelligent IOT platform,
the model building unit is used for building a model and carrying out model training on the sequence data and comprises the following steps:
the model building unit is used for inputting the sequence data into a model and extracting the characteristics of the sequence data through two layers of one-dimensional convolutional layers;
carrying out dimensionality reduction and filtration on the sequence data subjected to feature extraction through a pooling layer and a Flatten layer;
inputting the sequence data into a BilSTM layer, and performing correlation processing on the sequence data through the BilSTM layer;
entering sequence data into a Dense layer for feature enhancement;
and outputting a model training result.
16. The system of claim 15, wherein the artificial intelligence load prediction system for the intelligent IOT platform,
the model construction unit is used for extracting the characteristics of the sequence data and comprises the following steps:
the model building unit is used for performing high-dimensional mapping on the sequence data in the one-dimensional convolutional layer.
17. The system of claim 15, wherein the artificial intelligence load prediction system for the intelligent IOT platform,
the model building unit is used for performing correlation processing on the sequence data through a BilSTM layer and comprises the following steps:
the model building unit is used for carrying out recursive feedback on the states of a hidden layer of past data and a hidden layer of future data in the sequence data;
the current data, past data, and future data in the sequence data are inherently correlated.
18. The system of claim 13, wherein the load prediction system comprises a plurality of load prediction units,
the feature screening unit is used for performing feature screening on the trained data and comprises the following steps:
and the characteristic screening unit is used for removing irrelevant characteristic data in the sequence data according to the L1 regularization formula.
19. The system of claim 18, wherein the artificial intelligence load prediction system for the intelligent IOT platform,
the L1 regularization formula is
；
Wherein,
，
respectively a training sample and its corresponding label,
in order to be a weight parameter, the weight parameter,
in order to be the objective function, the target function,
in order to be a penalty term,
in order to make the penalty term coefficient,
the target function after the irrelevant features are removed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211512758.4A CN115545354A (en) | 2022-11-30 | 2022-11-30 | Artificial intelligence load prediction method and system for intelligent Internet of things platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211512758.4A CN115545354A (en) | 2022-11-30 | 2022-11-30 | Artificial intelligence load prediction method and system for intelligent Internet of things platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115545354A true CN115545354A (en) | 2022-12-30 |
Family
ID=84722031
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211512758.4A Pending CN115545354A (en) | 2022-11-30 | 2022-11-30 | Artificial intelligence load prediction method and system for intelligent Internet of things platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115545354A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112529283A (en) * | 2020-12-04 | 2021-03-19 | 天津天大求实电力新技术股份有限公司 | Comprehensive energy system short-term load prediction method based on attention mechanism |
| CN113222279A (en) * | 2021-05-28 | 2021-08-06 | 国网浙江桐乡市供电有限公司 | Short-term load prediction method considering demand response |
-
2022
- 2022-11-30 CN CN202211512758.4A patent/CN115545354A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112529283A (en) * | 2020-12-04 | 2021-03-19 | 天津天大求实电力新技术股份有限公司 | Comprehensive energy system short-term load prediction method based on attention mechanism |
| CN113222279A (en) * | 2021-05-28 | 2021-08-06 | 国网浙江桐乡市供电有限公司 | Short-term load prediction method considering demand response |
Non-Patent Citations (1)
| Title |
|---|
| 朱凌建等: ""基于CNN-BiLSTM 的短期电力负荷预测"", 《电网技术》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| CN107609667B (en) | Heat supply load prediction method and system based on Box _ cox transformation and UFCNN | |
| CN111461463B (en) | Short-term load prediction method, system and equipment based on TCN-BP | |
| CN109242169B (en) | Short-term load prediction method for large power customer | |
| CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
| CN114119273B (en) | Non-invasive load decomposition method and system for park comprehensive energy system | |
| CN110717610A (en) | Wind power prediction method based on data mining | |
| CN111079989A (en) | Water supply company water supply amount prediction device based on DWT-PCA-LSTM | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN113128666A (en) | Mo-S-LSTMs model-based time series multi-step prediction method | |
| CN111008726A (en) | Class image conversion method in power load prediction | |
| CN116757465A (en) | Line risk assessment method and device based on double training weight distribution model | |
| CN117114160A (en) | Short-term photovoltaic power prediction method | |
| CN113393034A (en) | Electric quantity prediction method of online self-adaptive OSELM-GARCH model | |
| CN117748495A (en) | Photovoltaic power station short-term generation power prediction method based on residual neural network | |
| CN116739118A (en) | Power load prediction method for realizing error correction mechanism based on LSTM-XGBoost | |
| CN114118401A (en) | Neural network-based power distribution network flow prediction method, system, device and storage medium | |
| CN116937559A (en) | Power system load prediction system and method based on cyclic neural network and tensor decomposition | |
| CN109146191B (en) | Short-term power load prediction method and system based on generic implication model | |
| CN115545354A (en) | Artificial intelligence load prediction method and system for intelligent Internet of things platform | |
| CN111126645A (en) | Wind power prediction algorithm based on data mining technology and improved support vector machine | |
| Wu et al. | Optimizing CNN-LSTM model for short-term PV power prediction using northern goshawk optimization | |
| Yeh et al. | Short-term load power prediction based deep learning gated recurrent unit in solar power plant | |
| Dai et al. | Short-Term Prediction of Energy Consumption in Smart Buildings Based on AC-BiLSTM | |
| CN114462476A (en) | Time series feature generation method and device based on evolutionary algorithm and LSTM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221230 |