CN113570414A - Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm - Google Patents
Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm Download PDFInfo
- Publication number
- CN113570414A CN113570414A CN202110859266.1A CN202110859266A CN113570414A CN 113570414 A CN113570414 A CN 113570414A CN 202110859266 A CN202110859266 A CN 202110859266A CN 113570414 A CN113570414 A CN 113570414A
- Authority
- CN
- China
- Prior art keywords
- neural network
- deep neural
- electricity price
- data
- adam algorithm
- 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
- 230000005611 electricity Effects 0.000 title claims abstract description 98
- 238000000034 method Methods 0.000 title claims abstract description 55
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 title claims abstract description 54
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 40
- 238000003062 neural network model Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 9
- 210000002569 neuron Anatomy 0.000 claims abstract description 7
- 238000012546 transfer Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 26
- 230000004913 activation Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000002759 z-score normalization Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000012216 screening 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0206—Price or cost determination based on market factors
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Entrepreneurship & Innovation (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Primary Health Care (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The real-time electricity price prediction method for optimizing the deep neural network based on the improved Adam algorithm comprises the following steps: (1) acquiring relevant data of electricity prices and influence factors thereof in the power system as sample data; (2) carrying out normalization pretreatment on relevant data of electricity price and influence factors thereof in the power system; (3) determining an input mode, an output mode, the number of hidden layers, the number of hidden layer neurons, a hidden layer transfer function, an output layer transfer function and a loss function of a neural network; improving the traditional Adam algorithm, and optimizing a deep neural network model by using the improved Adam algorithm; (4) the method comprises the steps of taking an influence factor with large correlation with actual electricity prices as an input quantity, predicting the electricity prices as an output quantity, training a deep neural network model based on an improved Adam algorithm, and optimizing parameters of the deep neural network model; (5) and processing the electricity price influence factor data of different nodes in the power system by using the finally optimized deep neural network model, and predicting the real-time electricity prices of the different nodes. The invention can improve the sufficiency of the utilization data, accelerate the convergence speed of the training and improve the accuracy of the electricity price prediction.
Description
Technical Field
The invention relates to the technical field of power markets, in particular to a real-time electricity price prediction method for optimizing a deep neural network based on an improved Adam algorithm.
Background
The real-time electricity price is the marginal cost of providing electricity to users within a limited extremely short period of time under the condition of considering the operation and the basic investment of the power system, directly reflects the relation between the market price and the electricity purchasing cost of the day ahead or the real-time market, and is one of the most ideal electricity price mechanisms. The accurate prediction of the real-time electricity price can provide reliable value basis for electricity purchasing users on one hand, so that a scientific electricity utilization strategy is formulated; on the other hand, the method can provide important reference for the electric power market supervision department, further formulate reasonable market rules and promote healthy, stable and orderly development of the electric power market. However, the real-time electricity price is susceptible to various factors, so that the real-time electricity price presents strong volatility and sparseness, and effective prediction of the real-time electricity price is difficult. Therefore, the problem of predicting the real-time electricity price has become one of the important issues in the field of marketable operation of the current power system.
At present, the real-time electricity price prediction methods mainly comprise two types, one type is a prediction method based on a time sequence, and the prediction method comprises an autoregressive conditional variance model, an autoregressive moving average model and the like, and is mainly used for representing the linear relation between the real-time electricity price and time and determining the similarity association between sample data and the electricity price to be predicted. The other type is an intelligent prediction method based on machine learning, and comprises prediction methods such as a Support Vector Machine (SVM), an artificial neural network and the like, wherein the SVM has better nonlinear mapping capability, and the generalization performance of the system is improved; the artificial neural network has good parallel distribution processing capacity and high fault tolerance performance on the electricity price noise.
The above real-time electricity price-based prediction method has the limitations that: with the new energy and new equipment being incorporated into each level of power grid, the electricity price time sequence presents more complex nonlinear characteristics, so that the time sequence prediction method is difficult to select the appropriate number of input variables; the real-time electricity price prediction method of the artificial neural network is adopted, so that the phenomenon of over-fitting of the prediction model is easily caused, and the prediction performance of the model is influenced; although the prediction method based on the support vector machine overcomes the defects of poor generalization capability, slow convergence and the like in the artificial neural network prediction method, the calculation timeliness of the large-scale training sample data is greatly reduced. Therefore, the ideal effect is difficult to achieve by adopting the existing real-time electricity price prediction method.
Disclosure of Invention
The invention aims to overcome the defects of the existing real-time electricity price prediction method, and aims to improve the timeliness and the efficiency stability of a prediction model by aiming at the volatility, the sparsity and the nonlinear characteristics of the real-time electricity price and adopting improved Adam algorithm optimization parameters to use the improved Adam algorithm for deep neural network optimization, so as to construct a real-time electricity price prediction system and method based on the improved Adam algorithm optimized Deep Neural Network (DNN).
The invention aims to overcome the defects in the prior art, and provides a real-time electricity price prediction method based on an improved Adam optimized deep neural network, which can accurately predict the real-time electricity price of an electric power market.
In order to solve the technical problem, the invention provides a real-time electricity price prediction method based on an improved Adam optimized deep neural network, which is characterized by comprising the following steps of:
s1, determining the electricity price influence factors, acquiring historical data of each influence factor, and forming each monitoring quantity sequence;
s2, analyzing the correlation between the actual electricity price and other influence factors based on each index sequence, selecting the influence factor sequence with large correlation as a training data set, and carrying out z-score normalization processing on the used training sample data;
s3, selecting proper hidden layer transfer functions, output layer transfer functions and loss functions, and constructing a depth neural network prediction model based on an improved Adam algorithm;
s4, taking an influence factor with large correlation with the actual electricity price as an input quantity, predicting the electricity price as an output quantity, training a deep neural network model based on an improved Adam algorithm, and optimizing parameters of the deep neural network model;
and S5, processing the electricity price influence factor data of different nodes in the power system by using the finally optimized deep neural network model, and predicting the electricity prices of the different nodes.
Further, the electricity price prediction method based on the advanced Adam algorithm optimized deep neural network as claimed in claim 1, wherein in step (1), the data partitioning process divides the electricity utilization area according to the actual geographic location of the load and the related electricity utilization data in the local power grid management system, divides the electricity utilization area into four areas, namely an industrial area, a commercial area, a residential area and a public activity area, integrates and processes the data of the divided areas, and transmits the processed data to the database for storage.
Furthermore, the acquired data mainly refers to historical electricity price data, historical load data, relevant weather data and the like of the region to which the data belongs; input variables of the deep neural network model are influence factor data such as historical load data and relevant weather data, and output variables of the deep neural network model are predicted electricity prices; the correlation between the real-time electricity rate and other indicators is analyzed by using a gray correlation analysis method.
Further, the normalization preprocessing adopts a formula of
Wherein x is*Represents the normalized sample values, μ represents the sample mean, and σ represents the sample standard deviation.
Further, the deep learning technique can be divided into the following 3 steps: parameterizing by using the weight, measuring the output quality by using a loss function and adjusting the weight by using the loss value.
The interior of the deep neural network model can be divided into three categories: an input layer, a hidden layer, and an output layer. Suppose that the l-th layer has nlA neuron inputting a vector of zlThe vector of the output component is hlLet u be hlSo as to distinguish the final output from the output of the hidden layer, according to the calculation rule of the deep neural network, the method comprises the following steps:
zl＝Wlzl-1+bl,l＝1,2,…,L
hl＝fl(zl)
in the formula:
is a weight matrix from layer l-1 to layer l,
is a bias vector of the l-th layer, flIs the activation function of the l-th layer.
Furthermore, the improved Adam algorithm is characterized in that a momentum idea is integrated into the Adam algorithm, so that the Adam algorithm is more stable and has a faster convergence speed.
Compared with the prior art, the method has higher prediction precision, can reliably analyze the change of the electricity price when the load and the weather condition dynamically change, can better guide the load operation of the power system, and has important significance for maintaining the stability of the electricity price.
Drawings
Fig. 1 is a schematic flow chart of the electricity price prediction method for optimizing the deep neural network based on the improved Adam algorithm.
Fig. 2 is a schematic diagram of a training and prediction process of an improved Adam algorithm optimized deep neural network model provided by an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. 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.
It should be noted that the deep neural network driven by data completely can effectively solve the problem of power price prediction in a larger data scale as a deep learning model fitting a complex nonlinear relation. With the rapid innovation of the large-power spot market, the data of the power system rapidly increase, and in order to meet the large-scale electricity price data processing requirements, an electricity price prediction model which can process more data and has better accuracy and practicability is established to be of great importance. Meanwhile, parameters are optimized by adopting an improved Adam algorithm, and the timeliness and the robustness of the prediction model are further improved.
The invention discloses an electricity price prediction system and method for optimizing a deep neural network based on an improved Adam algorithm. Firstly, based on the characteristics of the clear price data of the spot market, screening out key influence factors to construct input quantity, and further discussing a data preprocessing mode; secondly, researching a specific framework flow of the electricity price prediction of the optimized deep neural network based on the improved Adam algorithm based on the basic principle of the artificial intelligence deep learning technology.
Referring to fig. 1 and fig. 2, an embodiment of the present invention provides an electricity price prediction method for optimizing a deep neural network based on an improved Adam algorithm, including the steps of:
s1, obtaining training sample data according to a preset feature selection rule;
it should be noted that the data is the key for predicting the electricity price by using the optimized deep neural network algorithm of the improved Adam, and the prediction accuracy is directly influenced. Therefore, the input quantity database of the embodiment of the invention selects the microscopic influence factors which have influence effect on the electricity price, and covers the time, the supply-demand ratio, the load, the weather and the electricity price of the first three days. And the features are respectively used as vector forms to form input quantities of the daily electricity price prediction model. In the optimized deep neural network algorithm prediction model based on the improved Adam, the output quantity totals 24-dimensional characteristics and respectively corresponds to the electricity price of each 24-hour day.
The input quantities total 123-dimensional features, as detailed below. The first dimension input quantity is the day type of the day of the forecast day, and Monday-Sunday are respectively marked as 1-7; the 2 nd dimension input quantity is whether the forecast day is a working day or not, if the forecast day is a working day, the working day is marked as 0, and the weekend is marked as 1; the 3 rd dimension input quantity is a system supply-demand ratio for predicting a day-ahead peak time; the 4 th-6 th dimension input quantity is the minimum value, the mean value and the maximum value of the system load statistic on the day of the prediction day; the 7 th-30 th dimension input quantity is a system demand side load value at each moment 24 hours on the forecast day; the 31 st-39 th dimension input quantity is the minimum value, the mean value and the maximum value of the electricity price statistic of the day before the forecast day, two days before the forecast day and three days before the forecast day; the input quantities of the 40 th dimension to the 48 th dimension are the minimum value, the mean value and the maximum value of the temperature statistics of the day before, two days before and three days before the prediction day, the input quantities of the 49 th dimension to the 51 th dimension are the minimum value, the mean value and the maximum value of the temperature forecast of the day before the prediction day, and the input quantities of the 52 th dimension to the 123 th dimension are the electricity prices (the default price is once per hour) at 72 hours of the day before, two days before and three days before the prediction day.
It should be noted that, in the data partitioning process, the power utilization area is divided according to the actual geographic location of the load and the relevant power utilization data in the local power grid management system, the power utilization area is divided into four areas, namely an industrial area, a commercial area, a residential area and a public activity area, then the data integration and processing are performed on each divided area, and the processed data are transmitted to the database for storage.
And S2, carrying out normalization preprocessing on the training sample data.
In the embodiment of the present invention, variable dimensions such as a day type, a supply-demand ratio, a load value, an air temperature, and an electricity price included in the input quantity of the electricity price prediction model are not consistent and have a large difference in numerical value, which is not favorable for training. Therefore, in the embodiment of the present invention, a z-score normalization method is adopted to perform normalization on the training sample, and further, a formula adopted by the preprocessing is as follows:
in the formula, x*Represents the normalized sample values, μ represents the sample mean, and σ represents the sample standard deviation.
Considering that the input quantities of the training sample and the data to be predicted are both subjected to the normalization processing of the sample data, it is essential to perform the inverse normalization processing on the output electricity price data. The denormalization is the inverse of the normalization, as shown in the following equation:
x＝x*×σ+μ
wherein the variables have the same meaning as above.
And S3, inputting the training sample data after normalization preprocessing into a pre-constructed electricity price prediction model of the advanced Adam algorithm optimized deep neural network for training to obtain the electricity price prediction model of the advanced Adam algorithm optimized deep neural network.
In the embodiment of the present invention, further, step S3 specifically includes:
the interior of the deep neural network model can be divided into three categories: an input layer, a hidden layer, and an output layer. Suppose that the l-th layer has nlA neuron inputting a vector of zlThe vector of the output component is hlLet u be hlSo as to distinguish the final output from the output of the hidden layer, according to the calculation rule of the deep neural network, the method comprises the following steps:
zl＝Wlzl-1+bl,l＝1,2,…,L
hl＝fl(zl)
in the formula：
Is a weight matrix from layer l-1 to layer l,
is a bias vector of the l-th layer, flIs the activation function of the l-th layer.
The deep learning technique can be divided into 3 steps: parameterizing by using the weight, measuring the output quality by using a loss function and adjusting the weight by using the loss value. Therefore, first, the hidden layer activation function, the output layer activation function, and the loss function are determined as follows, respectively.
The Activation function (Activation function) simulates the threshold Activation characteristic of a human brain neuron, introduces nonlinear characteristics into a deep neural network, and realizes the conversion from a simple linear space to a highly nonlinear space. Considering the advantages of fast convergence speed and strong generalization capability of model training using the ReLU function, the ReLU function is selected as the hidden layer activation function, specifically as follows:
f1(zl)＝max(0,zl)
the activation function of the output layer is determined according to the problem to be solved, the rock burst intensity level prediction belongs to a classification task, and a Softmax function is usually adopted, and the function form is as follows:
in the formula:
is the output of the kth neuron of the output layer.
Meanwhile, the Loss function (Loss function) generally selects a Cross entropy error (Cross entropy error) which is a function of:
in the formula (I), the compound is shown in the specification,
in the form of an actual value of the value,
as a predicted value, N is the number of learning samples, and T is the number of classifications.
And S4, optimizing parameters of the deep neural network model according to the improved Adam algorithm. The improved Adam algorithm is characterized in that the momentum idea is integrated into the Adam algorithm, so that the Adam algorithm is more stable and has faster convergence speed. The specific process can be seen in FIG. 2.
The method comprises the following specific steps:
(1) initializing parameters:
the initial learning rate eta, the exponential decay rates of the first moment and the second moment estimation are respectively beta1And beta2，β1,β2E [0,1) for a numerically stable small constant δ;
an initial parameter theta;
first moment vector m00, second order moment vector v0Time step t of 00Improved Adam algorithm update quantity at 0
Adam algorithm iteration direction
(2) While stop criteria not satisfied do
And when the criterion is met, the Adam optimization algorithm can be stopped, the deep network neural system is directly returned to output a predicted value, otherwise, the parameters and the Adam algorithm are continuously updated until the criterion is met.
(3) Randomly selecting m samples from the training set, y(i)Is a sample x(i)Corresponding true value
(4) Calculate the average gradient of m samples:
(5)t←t+1
(6) updating biased first moment estimates:
mt←β1·mt-1+(1-β1)·gt
(7) updating the biased second moment estimation:
vt←β2·vt-1+(1-β2)·gt⊙gt
(8) correcting the deviation of the first moment:
(9) correcting the deviation of the second moment:
(10) the updating amount of each step of iteration of the Adam algorithm is improved:
(11) updating parameters:
θt←θt-1-Pt AM
(12)end while
and S5, obtaining the influence factors at the moment to be predicted to form a prediction data set, and inputting the prediction data set into the deep neural network model to obtain the predicted value of the electricity price.
The input quantity database of the embodiment of the invention selects microscopic influence factors which have influence effect on the electricity price, and covers time, supply-demand ratio, load, weather and the electricity price of the first three days. And the features are respectively used as vector forms to form input quantities of the daily electricity price prediction model. In the optimized deep neural network algorithm prediction model based on the improved Adam, the output quantity totals 24-dimensional characteristics and respectively corresponds to the electricity price of each 24-hour day.
The input quantities total 123-dimensional features, as detailed below. The first dimension input quantity is the day type of the day of the forecast day, and Monday-Sunday are respectively marked as 1-7; the 2 nd dimension input quantity is whether the forecast day is a working day or not, if the forecast day is a working day, the working day is marked as 0, and the weekend is marked as 1; the 3 rd dimension input quantity is a system supply-demand ratio for predicting a day-ahead peak time; the 4 th-6 th dimension input quantity is the minimum value, the mean value and the maximum value of the system load statistic on the day of the prediction day; the 7 th-30 th dimension input quantity is a system demand side load value at each moment 24 hours on the forecast day; the 31 st-39 th dimension input quantity is the minimum value, the mean value and the maximum value of the electricity price statistic of the day before the forecast day, two days before the forecast day and three days before the forecast day; the input quantities of the 40 th dimension to the 48 th dimension are the minimum value, the mean value and the maximum value of the temperature statistics of the day before, two days before and three days before the prediction day, the input quantities of the 49 th dimension to the 51 th dimension are the minimum value, the mean value and the maximum value of the temperature forecast of the day before the prediction day, and the input quantities of the 52 th dimension to the 123 th dimension are the electricity prices (the default price is once per hour) at 72 hours of the day before, two days before and three days before the prediction day.
Compared with the prior art, the method has higher prediction precision, can reliably analyze the change of the electricity price when the load and the weather condition dynamically change, can better guide the load operation of the power system, and has important significance for maintaining the stability of the electricity price.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.
Claims (10)
1. A power price prediction method for optimizing a deep neural network based on an improved Adam algorithm is characterized by comprising the following steps:
(S1) acquiring relevant data of the electricity price and the influence factors thereof in the power system as sample data;
(S2) carrying out normalization preprocessing on the relevant data of the electricity price and the influence factors thereof in the power system;
(S3) determining an input pattern, an output pattern, a number of hidden layers, a number of hidden layer neurons, a hidden layer transfer function, an output layer transfer function, and a loss function of the neural network; improving the traditional Adam algorithm, and optimizing a deep neural network model by using the improved Adam algorithm;
(S4) taking an influence factor with large correlation with the actual electricity price as an input quantity, predicting the electricity price as an output quantity, training a deep neural network model based on an improved Adam algorithm, and optimizing parameters of the deep neural network model;
(S5) processing the electricity price influence factor data of different nodes in the power system by using the finally optimized deep neural network model, and predicting the real-time electricity prices of the different nodes.
2. The method for predicting the electricity price based on the advanced Adam algorithm optimized deep neural network as claimed in claim 1, wherein in the step (S1), the data partitioning process divides the electricity utilization area according to the actual geographical location of the load and the related electricity utilization data in the local grid management system, divides the electricity utilization area into four areas, namely an industrial area, a commercial area, a residential area and a public activity area, integrates and processes the data of the divided areas, and transmits the processed data to the database for storage.
3. The electricity price prediction method based on the advanced Adam algorithm optimized deep neural network of claim 1, characterized in that in the step (S1), the obtained data mainly refers to historical electricity price data, historical load data, relevant weather data and the like of the region.
4. The electricity price prediction method based on the advanced Adam algorithm optimized deep neural network of claim 1, wherein in the step (S1), the correlation between the real-time electricity price and other indexes is analyzed by using a grey correlation analysis method.
5. The method for predicting the electricity price based on the advanced Adam algorithm optimized deep neural network as claimed in claim 1, wherein in the step (S2), the historical data of each index is normalized, and the formula adopted in the normalization preprocessing is
Wherein x is*Represents the normalized sample values, μ represents the sample mean, and σ represents the sample standard deviation.
6. The method for predicting the electricity prices based on the advanced Adam algorithm optimized deep neural network of claim 1, wherein in the step (S3), the input variables of the deep neural network model are influence factor data such as historical load data and related weather data, and the output variables of the deep neural network model are predicted electricity prices.
7. The electricity price prediction method based on the advanced Adam algorithm optimized deep neural network of claim 1, wherein in the step (S3), the deep learning technique can be divided into the following 3 steps: parameterizing by using the weight, measuring the output quality by using a loss function and adjusting the weight by using the loss value.
8. The electricity price prediction method based on the advanced Adam algorithm optimized deep neural network of claim 1, wherein in the step (S3), the interior of the deep neural network model can be divided into three categories: an input layer, a hidden layer, and an output layer. Suppose that the l-th layer has nlA neuron inputting a vector of zlThe vector of the output component is hlLet u be hlSo as to distinguish the final output from the output of the hidden layer, according to the calculation rule of the deep neural network, the method comprises the following steps:
zl＝Wlzl-1+bl,l＝1,2,…,L…，L
hl＝fl(zl)
in the formula:
is a weight matrix from layer l-1 to layer l,
is a bias vector of the l-th layer, flIs the activation function of the l-th layer.
9. The electricity price prediction method based on the advanced Adam algorithm optimized deep neural network as claimed in claim 1, wherein in the step (S4), the advanced Adam algorithm is to blend momentum idea into the Adam algorithm, so that the advanced Adam algorithm is more stable and has faster convergence speed.
10. The improved Adam algorithm-based power price prediction method for optimizing the deep neural network according to claim 1, wherein in the step (S5), the neural network model is trained based on the improved Adam algorithm based on the normalized training data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110859266.1A CN113570414A (en) | 2021-07-29 | 2021-07-29 | Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110859266.1A CN113570414A (en) | 2021-07-29 | 2021-07-29 | Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113570414A true CN113570414A (en) | 2021-10-29 |
Family
ID=78168646
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110859266.1A Pending CN113570414A (en) | 2021-07-29 | 2021-07-29 | Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113570414A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114720956A (en) * | 2022-06-07 | 2022-07-08 | 成都信息工程大学 | Water condensate particle classification and identification method and device for dual-polarization weather radar |
| CN115619437A (en) * | 2022-12-12 | 2023-01-17 | 中国华能集团清洁能源技术研究院有限公司 | Real-time electricity price determining method and system |
-
2021
- 2021-07-29 CN CN202110859266.1A patent/CN113570414A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114720956A (en) * | 2022-06-07 | 2022-07-08 | 成都信息工程大学 | Water condensate particle classification and identification method and device for dual-polarization weather radar |
| CN115619437A (en) * | 2022-12-12 | 2023-01-17 | 中国华能集团清洁能源技术研究院有限公司 | Real-time electricity price determining method and system |
| CN115619437B (en) * | 2022-12-12 | 2023-08-29 | 中国华能集团清洁能源技术研究院有限公司 | Real-time electricity price determining method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| Cai et al. | Predicting the energy consumption of residential buildings for regional electricity supply-side and demand-side management | |
| CN108846517B (en) | Integration method for predicating quantile probabilistic short-term power load | |
| CN111860979B (en) | Short-term load prediction method based on TCN and IPSO-LSSVM combined model | |
| CN111428926B (en) | Regional power load prediction method considering meteorological factors | |
| CN111260136A (en) | Building short-term load prediction method based on ARIMA-LSTM combined model | |
| CN108388962B (en) | Wind power prediction system and method | |
| CN110163429B (en) | Short-term load prediction method based on similarity day optimization screening | |
| CN105069525A (en) | All-weather 96-point daily load curve prediction and optimization correction system | |
| CN110380444B (en) | Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula | |
| CN107292534A (en) | The yardstick competition evaluation method and device of urban power distribution network long term dynamics investment | |
| CN106503851A (en) | A kind of improved Short-Term Load Forecasting Method based on wavelet analysises | |
| CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
| CN111079989A (en) | Water supply company water supply amount prediction device based on DWT-PCA-LSTM | |
| CN113570414A (en) | Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm | |
| CN112329990A (en) | User power load prediction method based on LSTM-BP neural network | |
| CN110163444A (en) | A kind of water demand prediction method based on GASA-SVR | |
| CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
| CN115222106A (en) | User day-ahead load prediction method of self-adaptive model | |
| CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
| Hu et al. | Multi-objective optimization of time-of-use price for tertiary industry based on generalized seasonal multi-model structure | |
| CN115358437A (en) | Power supply load prediction method based on convolutional neural network | |
| CN116151464A (en) | Photovoltaic power generation power prediction method, system and storable medium | |
| CN116826745B (en) | Layered and partitioned short-term load prediction method and system in power system background | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |