New energy output prediction and demand response optimization method and device based on LSTM-BN network

Abstract

The invention relates to a new energy output prediction and demand response optimization method and device based on an LSTM-BN network. The method comprises the steps of obtaining target meteorological information and new energy output information of a region where a new energy power station is located in the same historical period; based on the target meteorological information and the new energy output information, utilizing an LSTM-BN prediction model to predict energy output of the new energy power station, and obtaining target output; and constructing a demand response optimization model with the purpose of minimizing the energy overhead cost based on the target output, and solving the demand response optimization model to obtain a demand response optimization strategy. The method combines the advantages of the LSTM and the BN network, constructs an energy output prediction model, introduces a price-based demand response mechanism, and achieves the flexible adjustment of a load and the efficient utilization of new energy.

Description

Technical Field

[0001] The present application relates to the field of smart grid technology, and in particular to a method and device for new energy output prediction and demand response optimization based on an LSTM-BN network. Background Art

[0002] Due to the volatility and uncertainty of renewable energy generation, the existing power consumption strategy cannot timely adjust the balance between the supply of generators and the demand of users, resulting in huge energy waste. Therefore, accurate and effective prediction of renewable energy output is crucial to optimize the demand response mechanism, which will help maintain the stability of the power grid and improve energy efficiency. Summary of the invention

[0003] Based on this, it is necessary to provide a new energy output prediction and demand response optimization method and device based on LSTM-BN network that can improve energy utilization efficiency in response to the above technical problems.

[0004] In the first aspect, the present application provides a new energy output prediction and demand response optimization method based on LSTM-BN network. The method includes:

[0005] Obtain target meteorological information and new energy output information for the same historical period in the area where the new energy power station is located;

[0006] Based on the target meteorological information and new energy output information, the energy output of the new energy power station is predicted using the LSTM-BN prediction model to obtain the target output;

[0007] Based on the target output, a demand response optimization model is constructed to minimize the energy expenditure cost, and the demand response optimization model is solved to obtain the demand response optimization strategy.

[0008] In one embodiment, obtaining target meteorological information of the area where the new energy power station is located includes:

[0009] Obtain meteorological data corresponding to several meteorological factors in the area where the new energy power station is located;

[0010] Obtain the correlation between meteorological factors corresponding to meteorological data and energy output data through path analysis;

[0011] Extract target meteorological information from meteorological data based on correlation.

[0012] In one embodiment, based on the target meteorological information and the new energy output information, the energy output of the new energy power station is predicted using the LSTM-BN prediction model, and obtaining the target output includes:

[0013] Input the target meteorological information and new energy output information into the LSTM network to extract the new energy meteorological time series features;

[0014] The new energy meteorological time series characteristics and the meteorological characteristics of the period to be predicted are input into the BN network to predict the target output based on the meteorological characteristics.

[0015] In one embodiment, constructing a demand response optimization model based on a target output for the purpose of minimizing energy expenditure costs includes:

[0016] Obtain energy overhead costs based on the initial investment cost of cloud storage technology and the operating expenses incurred by additional energy procurement required for electricity, heating and gas networks;

[0017] A demand response optimization model is constructed with the goal of minimizing energy expenditure costs.

[0018] In one embodiment, it includes:

[0019] Obtain load data on electricity, heating and gas networks, as well as electricity prices and number of users at different times of use;

[0020] Based on load data, electricity prices at different times of use and the number of users, obtain the operating expenses for additional energy purchases required for electricity, heating and gas networks.

[0021] In one embodiment, solving the demand response optimization model to obtain the demand response optimization strategy includes:

[0022] The demand response optimization model is linearized and solved using the MILP solver to obtain the demand response optimization strategy.

[0023] In the second aspect, the present application also provides a new energy output prediction and demand response optimization device based on LSTM-BN network. The device includes:

[0024] The acquisition module is used to obtain the target meteorological information and new energy output information of the area where the new energy power station is located in the same historical period;

[0025] The prediction module is used to predict the energy output of the renewable energy power station based on the target meteorological information and renewable energy output information, and to obtain the target output by using the LSTM-BN prediction model;

[0026] The optimization module is used to build a demand response optimization model based on the target output for the purpose of minimizing energy expenditure costs, solve the demand response optimization model, and obtain the demand response optimization strategy.

[0027] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps in the above-mentioned method for optimizing the output prediction and demand response of new energy based on the LSTM-BN network when executing the computer program.

[0028] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-mentioned method for optimizing the output forecasting and demand response of new energy based on the LSTM-BN network when the computer program is executed by the processor.

[0029] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps in the above-mentioned method for optimizing the output forecast and demand response of new energy based on LSTM-BN network when executed by a processor.

[0030] The above-mentioned new energy output prediction and demand response optimization method and device based on LSTM-BN network include: obtaining target meteorological information and new energy output information in the same historical period of the area where the new energy power station is located; based on the target meteorological information and new energy output information, and using the LSTM-BN prediction model to predict the energy output of the new energy power station, and obtain the target output; based on the target output, construct a demand response optimization model for the purpose of minimizing energy expenditure costs, solve the demand response optimization model, and obtain the demand response optimization strategy. Using the LSTM-BN prediction model to predict energy output can obtain higher prediction accuracy and faster convergence speed, providing reliable data support for subsequent energy system planning. Based on the prediction results, a user-side demand response optimization model is established, allowing users to better adapt to electricity price fluctuations, achieve more efficient energy utilization and higher economic returns. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flowchart of a new energy output prediction and demand response optimization method based on LSTM-BN network in one embodiment;

[0032] Figure 2 This is the LSTM minimum network unit structure in one embodiment;

[0033] Figure 3 A schematic diagram of a new energy meteorological time series feature extraction process in an embodiment;

[0034] Figure 4 It is a flow chart of a new energy output prediction and demand response optimization device based on LSTM-BN network in another embodiment;

[0035] Figure 5This is a schematic diagram comparing the matching degree between the predicted values ​​and the actual values ​​of the four models;

[0036] Figure 6 This is a schematic diagram comparing the calculation speeds of the four models;

[0037] Figure 7 This is a schematic diagram comparing the prediction results of new energy output power of four models;

[0038] Figure 8 This is a comparison chart before and after the adoption of the demand response program;

[0039] Fig. 9 This is a comparison chart of power load in different scenarios;

[0040] Fig.10 This is a comparison chart of the user's fixed investment cost and lifetime net income results. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] The embodiment of the present application provides a new energy output prediction and demand response optimization method based on LSTM-BN network, such as Figure 1 As shown, the following steps are included:

[0043] Step 102, obtaining target meteorological information and new energy output information of the region where the new energy power station is located in the same historical period.

[0044] Among them, new energy power stations such as photovoltaic power stations and wind power stations collect target meteorological information in the area where the new energy power stations are located within a certain period of time, such as temperature, wind speed, light intensity, wind direction, etc. The collected new energy output information includes the power generation of each new energy power station, such as photovoltaic power generation of photovoltaic power stations and wind power generation of wind power stations. Both the target meteorological information and the new energy output information contain time information, that is, the target meteorological information and the new energy output information include the changes in meteorology over time and the changes in new energy output over time, respectively.

[0045] Step 104, based on the target meteorological information and the new energy output information, and using the LSTM-BN prediction model, predict the energy output of the new energy power station to obtain the target output.

[0046] Among them, LSTM (Long Short-Term Memory) is a neural network architecture for processing and predicting time series data, which can better capture dependencies in long sequences. BN (Bayesian Neural) network is a graphical model for representing probabilistic relationships between variables. It represents a set of random variables and their conditional dependencies through a directed acyclic graph (DAG). Combining LSTM network and BN, the LSTM-BN prediction model combines the advantages of BN in uncertainty quantification and LSTM network in processing power, solving the problem of volatility and uncertainty of new energy output.

[0047] By inputting the target meteorological information and new energy output information into the LSTM-BN prediction model, the temporal correlation between the meteorology and power generation of the new energy power station can be mined, thereby achieving accurate prediction of the new energy output.

[0048] Step 106: construct a demand response optimization model based on the target output for the purpose of minimizing energy expenditure cost, solve the demand response optimization model, and obtain a demand response optimization strategy.

[0049] For the new energy output predicted in step 104, a price-based demand response optimization model is established to minimize the energy overhead cost. This embodiment adopts a price-based demand response optimization strategy to guide users to adjust their energy use behavior through changes in electricity prices, encourage users to increase electricity consumption when there is excess new energy generation, and reduce electricity consumption during peak hours, thereby achieving flexible load adjustment and efficient use of new energy.

[0050] This embodiment uses the LSTM-BN prediction model to predict energy output, which can achieve higher prediction accuracy and faster convergence speed, providing reliable data support for subsequent energy system planning. Based on the prediction results, a user-side demand response optimization model is established, allowing users to better adapt to electricity price fluctuations, achieve more efficient energy utilization and higher economic returns.

[0051] After accurately predicting the output of renewable energy, a demand response strategy is formulated to balance the supply and demand of electricity, and by optimizing the demand response strategy, the uncertainty of the power system to the output of renewable energy is reduced.

[0052] In one embodiment, obtaining target meteorological information of the area where the new energy power station is located in step 102 includes: obtaining meteorological data corresponding to several meteorological factors in the area where the new energy power station is located; obtaining the correlation between the meteorological factors corresponding to the meteorological data and the energy output data through path analysis; and extracting target meteorological information from the meteorological data based on the correlation.

[0053] Assume that a number of meteorological data are collected for Q meteorological factors, and define the meteorological data corresponding to meteorological factor q as x q , the meteorological data corresponding to the meteorological factor q collected at time t is x qt , the new energy output is y t .x q With y t The correlation between The calculation formula is as follows:

[0054]

[0055] Meteorological factor q corresponds to meteorological data x q The meteorological factor p corresponds to the meteorological data x p After the impact, with y t The correlation between The calculation formula is as follows:

[0056]

[0057] and x q 、x p and t The mean of .

[0058] Path analysis is used to obtain the correlation between meteorological factors and new energy output. The meteorological factors with greater correlation with new energy output are retained, and the meteorological factors with less correlation are deleted. The meteorological data corresponding to the retained meteorological factors are extracted to form the target meteorological information.

[0059] In this embodiment, the target meteorological information selected for photovoltaic power stations includes sunlight intensity, light duration, temperature, humidity, etc., and the target meteorological information selected for wind power stations includes wind force, wind direction, air temperature, humidity, etc. Finally, z groups of meteorological data were selected to form the target meteorological information required by the LSTM-BN prediction model.

[0060] In one embodiment, step 104 includes: inputting the target meteorological information and the new energy output information into the LSTM network to extract the new energy meteorological time series features; inputting the new energy meteorological time series features and the meteorological features of the time period to be predicted into the BN network to predict the target output for the meteorological features.

[0061] Set the LSTM network to include L network units, and predict the output of network unit a at time t as It is expressed as:

[0062]

[0063] In the above formula, Xt is the target meteorological information, y t Output information for new energy, It is the result of the splicing operation of the two.

[0064] The minimum network unit structure of the LSTM network is as follows Figure 2 As shown in Figure C t , C t-1 , C t-2 、h t 、h t-1 、h t-2 are the states of each learning unit, i t is the input gate, i t Defined as:

[0065]

[0066] In the above formula, δ is the activation function, b i is the bias constant, W i is the weight matrix of the input gate.

[0067] f t is the Forgotten gate, f t Defined as:

[0068]

[0069] In the above formula, W f is the weight matrix of the forget gate.

[0070] o t is the output gate, o t Defined as:

[0071]

[0072] In the above formula, W o is the weight matrix of the output gate.

[0073] Status of learning unit C t and h t It is expressed as:

[0074]

[0075] In the above formula, W c is the weight matrix of the learning unit.

[0076] The LSTM network unit captures the long-term dependencies in the time series through the gating mechanism and realizes the extraction of new energy weather information features. Figure 3 As shown, the cell state of LSTM network unit a is set to hat The LSTM network finally outputs the new energy meteorological time series characteristics

[0077] Then build a BN network to extract the meteorological data for the forecast period and obtain the meteorological characteristics Then the new energy meteorological time series features output by the LSTM network are and meteorological characteristics After splicing, It is input into the BN network as the basis for predicting the output value of new energy. The prior distribution of the BN network constructed in this embodiment is a Gaussian distribution, and the formula is expressed as:

[0078]

[0079] In the above formula, W is the probability layer parameter of the BN network, π is the mixing ratio of the Gaussian distribution, and w j is the j parameter of the probability layer of the BN network, and is the variance under normal distribution, N w is the total number of parameters in the probability layer of the network. In the BN network, assuming that the input feature concatenation vector is X and the output new energy output prediction is Y, the posterior probability of the BN network is p(W / X,Y). The approximate distribution of the new energy output prediction is q(W;θ), which is expressed as:

[0080]

[0081] In the above formula, θ is the learning parameter of the model, μ j is the mean value of the j parameter, is the variance of the j parameter. The a sampling value of the j parameter in the model is calculated according to the approximate distribution equation (9): The calculation formula is:

[0082]

[0083] In the above formula, is the Hadamard product, z is is the multidimensional distribution of the meteorological characteristics of the forecast period in the normal distribution. Based on equation (10), the sampling parameter of A in the BN network model is obtained: Taking the feature concatenation vector X as the input of the BN network model and propagating it forward in the model, we can get the output new energy outflow prediction Y, which is expressed as:

[0084]

[0085] In one embodiment, in step 106, constructing a demand response optimization model based on the target output for the purpose of minimizing energy expenditure costs includes: obtaining energy expenditure costs based on the initial investment cost of cloud energy storage technology and the operating expenses of additional energy procurement required for electricity, heating, and natural gas networks; and constructing a demand response optimization model with the goal of minimizing energy expenditure costs.

[0086] Among them, the load data of electricity, heating and natural gas networks, as well as the electricity price and number of users during electricity usage time are obtained; based on the load data, the electricity price and number of users during electricity usage time, the operating expenses incurred by the additional energy procurement required for the electricity, heating and natural gas networks are obtained.

[0087] This embodiment adopts a price-based demand response optimization strategy to guide users to adjust their energy usage behavior through electricity price change signals to achieve power optimization.

[0088] The power load transfer in the customer-side price demand response optimization model is:

[0089] p l =p 0 +Δp(12)

[0090] In the above formula, p l is the user load matrix before the demand response optimization model is adopted, p 0 is the user load matrix after adopting the demand response optimization model, Δp is the user load transfer matrix, and the Δp formula is expressed as:

[0091]

[0092] In the above formula, E is the elasticity coefficient matrix of the user in multiple time periods, Δ[p] is the normalized matrix of the tariff change rate at a specified time, and E and Δ[p] are expressed as:

[0093]

[0094] In the above formula, Δp T The difference between the electricity price used by the customer at time T and the benchmark electricity price, is the base electricity price used by the customer at time T, e TT is the self-elasticity coefficient of row T and column T in the customer's multi-time elasticity coefficient matrix.

[0095] In the target study area, users of different load types (electricity, heat, and gas) each share different forms of renewable energy output. Electricity load users integrate their own photovoltaic power generation and share the power of regional wind power. Thermal load users receive the thermal energy converted from wind energy, and gas load users distribute the gas energy converted from wind energy. For each moment t (t = 1, 2, ..., T) in each season s (s = 1, 2, ..., S), the output of renewable energy to these three load type users i is expressed as:

[0096]

[0097] In the above formula, s represents season, t represents time, and i represents user. represents the power load of renewable energy output at time t, represents the photovoltaic power generation at time t, represents the power distribution of wind power at time t, w represents its distribution coefficient, represents the heat load of the new energy output at time t, v represents its distribution coefficient, Represents the gas load providing new energy output at time t.

[0098] At the same time, there is uncertainty in the output of new energy. When there is a surplus of energy, energy storage and load are needed, and when there is a shortage of energy, external energy supplementation is needed. Therefore, this embodiment introduces a multi-energy cloud energy storage model for users, which uses big data and real-time information in the cloud platform to formulate and optimize charging and discharging strategies. Cloud energy storage users meet their load needs by leasing energy storage usage rights from cloud energy storage suppliers, avoiding the fixed costs of directly purchasing, operating and maintaining physical energy storage equipment. The total cost of the user consists of two parts: one is the rental service fee for equipment services as the initial investment cost of using cloud energy storage technology, and the other is the operating expenses borne by the additional energy procurement required for electricity, heating, and natural gas networks. The user's goal is to minimize its equivalent annual value cost, that is, to minimize the energy overhead cost in order to maximize economic efficiency. The demand response optimization model is expressed as:

[0099]

[0100] In the above formula, k pa is the annual value coefficient, d is the length of time the energy storage device is used, r is the annual interest rate, C2 is the operating cost of the user using cloud energy storage, C inv,i is the investment cost, kf1 and kf2 are the service fees per unit power and per unit capacity of the lithium battery respectively. and are the maximum power provided by cloud energy storage and the maximum limit of cloud gas storage capacity, kg1 and kg2 are the energy conversion rate and capacity cost of the heat storage tank, kv1 and kv2 are the energy conversion rate and capacity cost of the gas storage tank. Among them, C2 is determined by the product of the real-time electricity price and electricity consumption. The real-time electricity price is determined by formulas (12)-(14), and the electricity consumption is determined by formula (15).

[0101] In one embodiment, solving the demand response optimization model to obtain the demand response optimization strategy includes: linearizing the demand response optimization model, solving it using a MILP solver, and obtaining the demand response optimization strategy.

[0102] Since the demand response optimization model contains piecewise functions, for example, the electricity price will change over time and produce piecewise changes, which makes the demand response optimization model difficult to solve. Therefore, the "big M method" is used to linearize the nonlinear constraints containing piecewise functions, which is mainly achieved by introducing auxiliary variables and 0-1 variables. Assume that a piecewise function in the solution function is:

[0103]

[0104] In order to represent f(x), an auxiliary variable y is introduced to represent y=max(x,0). Then, a 0-1 variable is used to represent the symbol x. Assuming that the 0-1 variable is u, the piecewise function is represented as:

[0105]

[0106] In the above formula, M is a large constant, and u1 and u2 are both 0-1 variables.

[0107] Finally, all these constraints are integrated into a new energy user cost minimization problem. The original problem is transformed into a mixed integer linear programming (MILP) problem, and the standard MILP solver is used to solve the user's minimized annual cost, realizing a new energy output demand response optimization.

[0108] In order to verify the effectiveness of the new energy output prediction model constructed using the LSTM-BN network in this study, we selected the Google Earth Engineering dataset and the GEFCom2014 dataset as the training data for the model. The Google Earth Engineering dataset is an open and unified spatial dataset of wind and solar installations. The GEFCom2014 dataset contains data on four aspects: new energy load, electricity prices, and wind and solar power generation forecasts for specific regions. The operating environment and parameter settings of the model are shown in Table 1

[0109] Table 1 Operating environment and model parameterization

[0110]

[0111]

[0112] The degree of fit between the prediction and actual values ​​of the LSTM-BN model is first determined by two data sets. To illustrate the superiority of this research method, the LSTM-BN model, BN model, SEAM-LSTM model, and CL-LSTM model are used to form a control experiment under the same experimental conditions. The comparison results are shown in Figure 2. Figure 5 shown.

[0113] like Figure 5 As shown, the horizontal axis is the true value and the vertical axis is the predicted value. The predicted value of the LSTM-BN model is consistent with the true value to the greatest extent. The SEAM-LSTM model also has a high degree of fit, but there is still a gap with the LSTM-BN model, which reflects the effectiveness of the LSTM-BN model in processing long-term time-dependent data. By comparing the results of Figure 5 (a) and (b), (e) and (f), it can be found that the LSTM-BN model captures the characteristics of meteorological information through LSTM, and then uses BN technology to optimize the prediction of new energy power generation, which has the potential to greatly improve the prediction accuracy. When LSTM is missing in the LSTM-BN model, the greater the deviation of the prediction result, the lower the prediction accuracy. Comparison Figure 5 From the results of (a) and (e), we can see that the research on the LSTM-BN model is consistent with the real value GEFCom2014 dataset and the Google Earth Engin dataset, which shows that the model has a certain generalization ability and can produce better prediction results in various usage scenarios. In order to further prove the prediction accuracy of the LSTM-BN model constructed in this study, the normalized root mean square error (NRMSE), the correlation coefficient between the predicted value and the real value (R 2 ) and average percentage error (APE) are used to quantitatively evaluate the prediction ability of the model. NRMSE is an important indicator to measure the difference between the predicted value and the actual value. The lower the value, the more accurate the prediction. APE directly reflects the percentage of the predicted error to the actual value. The lower the APE value, the more accurate the prediction. 2 The value is used to quantify the degree of correlation between the predicted values ​​and the actual observed values ​​of the model. The predictive ability of the model increases with R 2 The value increases as it approaches 1. Table 2 shows the research results.

[0114] Table 2 Comparison of the prediction capabilities of the four models

[0115]

[0116] Table 2 shows the detailed results of the prediction of the four models for two data sets. It can be found that the LSTM-BN model has the highest R no matter it is tested on the GEFCom2014 dataset or the Google Earth Engin dataset. 2 This indicates that the model established in this study can more accurately predict the new energy power output data.

[0117] The prediction speed of the four models was tested, and the results are as follows: Figure 6 As shown, Figure 6 The horizontal axis is the number of iterations, and the vertical axis is the fitness value. Figure 6 (a) It can be seen that the LSTM-BN model constructed in this study has been iterated about 64 times on the Google Earth Engin dataset, and the fitness value has reached a stable state. The SEAM-LSTM model and the CL-LSTM model have been iterated 80 and 82 times respectively, and the fitness values ​​have reached a stable state. The fitness values ​​of the three models are not much different, while the BN model has been iterated 112 times before the fitness value is stable, and the fitness value is higher than the first three models. Figure 6 As shown in (b), in the test of GEFCom2014 data set, the LSTM-BN model constructed in this study reached a stable fitness value after about 82 iterations, while the other three models did not reach stability until the end of training. The results show that the LSTM-BN model constructed in this study has efficient computing power when processing different data sets and has good generalization ability, that is, it can maintain stable optimization effects and fast convergence speed in different data environments.

[0118] The model training data comes from the renewable energy power generation system built on the roof of the factory from January to December 2023. The data is sampled once every minute to eliminate abnormal and invalid information. Finally, 12,568 high-quality data points are retained for subsequent research. At the same time, numerical weather forecast data from nearby meteorological stations are introduced for training. The trained model is used to predict the renewable energy output power on a random day. The results are as follows: Figure 7 As shown, Figure 7 The horizontal axis is time, and the vertical axis is output power.

[0119] from Figure 7It can be seen that the maximum output power of the new energy power system on a sunny day is 18865W, and it is concentrated between 11:00-14:00. The maximum output power on a rainy day is only 1356W, concentrated between 13:00-14:00. In terms of overall prediction, the curve values ​​between the predicted curve and the true curve of the LSTM-BN model constructed in this study are almost overlapping, and the maximum prediction error does not exceed 1%. In contrast, the error between the predicted curve and the true curve of the SEAM-LSTM model, CL-LSTM model and BN model is relatively large, and the output power of the new energy power system cannot be accurately predicted.

[0120] Once the LSTM-BN model is used to accurately predict the output power of renewable energy in the target area, a corresponding demand response optimization strategy can be formulated based on the prediction results. This study sets each season as a 91-day cycle and determines a typical 24-hour dispatch cycle. Due to the characteristics of the cloud energy storage system, frequent instant response is not required, so the dispatch interval of the system is set to 1 hour to optimize energy distribution. In addition, the community's electric power price mechanism is adopted, which is specifically divided into peak period, regular period and valley period, and the impact of tariff changes on dispatch strategies is simulated according to the specific tariffs and time periods. After citing the demand response design method designed in this study, the comparison of community power load is as follows: Figure 8 As shown, Figure 8 The horizontal axis is time and the vertical axis is electrical load power.

[0121] exist Figure 8 In the study, the peak electricity consumption period of the community in both seasons is around 12:00-21:00, and the peak electricity consumption period in the low season is between 0:00-6:00 every day. Before the demand response scheme designed in this study was applied, the distribution of consumer electricity load fluctuated greatly. After adding the optimization scheme designed in this study, the load volume curve began to flatten regardless of the season, and the charging increased during the low peak electricity consumption period and decreased during the peak period. Next, three comparison schemes were established to compare the impact of demand response and time-of-use tariff strategies. Scenario 1: Electricity, heat, and gas users are independent, build their own energy storage systems, and use new energy output prediction models. Scenario 2: Users give up their own energy storage and purchase cloud energy storage services. Scheme 3: Based on Scheme 2, a demand response optimization model is added, that is, users adjust their energy usage according to electricity prices and their own needs. The comparison of electricity load in one day under the three scenarios is shown in the figure. Fig. 9 As shown, Fig. 9 The horizontal axis is time and the vertical axis is electrical load power.

[0122] like Fig. 9As shown in the figure, due to the configuration of the energy storage system, Scheme 2 shows a more significant peak load suppression effect. In contrast, the model of Scheme 3 not only reduces the operating cost of the energy storage system, but also better improves the fluctuation of the load curve and makes the load curve smoother by optimizing the energy storage capacity configuration and time-of-use electricity price strategy. The results of the user's fixed investment cost and the net benefit of the whole life cycle under the three scenarios are shown in Fig.10 shown.

[0123] like Fig.10 As shown in (a), the fixed investment costs of users in Scenario 1, Scenario 2, and Scenario 3 are $2.78 million, $3.52 million, and $3.52 million, respectively. This is because users of Scheme II and Scheme III need to purchase cloud storage power systems and bear their maintenance costs. Fig.10 As shown in (b), the net benefits of the users in Scenario I, Scenario II, and Scenario III over their entire life cycle are $2.15 million, $3.54 million, and $4.45 million, respectively. This shows that the use of cloud energy storage systems by users reduces overall operating costs and improves energy efficiency, thereby increasing net benefits. The use of demand response optimization models greatly increases net benefits. After optimization, users are able to better adapt to fluctuations in electricity prices, achieve more efficient energy use, and achieve higher economic returns.

[0124] Faced with the strong uncertainty of new energy output, this application proposes a new energy output prediction method based on LSTM-BN. This method uses the LSTM network to extract new energy weather quantity vector features, and then uses BN to predict power. It provides a solid database for subsequent energy system planning. Secondly, considering the important impact of price-based demand response on user behavior in the market mechanism, a user-side demand response model was established. Guide users to adjust their electricity consumption behavior through price signals. It is obvious that the LSTM-BN network constructed in this study is superior to other algorithms in terms of prediction accuracy and budget speed. In terms of prediction accuracy, the NRMSE value of the LSTM-BN model is less than 0.0090, and R 2 is greater than 0.99, and APE is less than 0.5. In terms of budget speed, the LSTM-BN model tends to be stable after 64 and 80 iterations in the Google Earth EnginEn and GEFCom2014 datasets, respectively, and the fitness value is the lowest. After accurately predicting the output of new energy for the LSTM-BN model, a price-based demand response strategy is proposed. The results show that the demand response method used in this study makes the load curve smoother and plays an important role in peak shaving and valley filling. At the same time, after applying the demand response optimization model, the net benefit of users in the region throughout their life cycle is maximized, which is 4.45 million yuan. The results show that the LSTM-BN model has high prediction accuracy and fast convergence speed in the prediction of new energy output, providing good support for the stability and efficiency of the power system.

[0125] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0126] Based on the same inventive concept, the embodiment of the present application also provides a device for new energy output prediction and demand response optimization based on LSTM-BN network for implementing the above-mentioned method for new energy output prediction and demand response optimization based on LSTM-BN network. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the new energy output prediction and demand response optimization device based on LSTM-BN network provided below can be referred to the limitations of the new energy output prediction and demand response optimization method based on LSTM-BN network above, and will not be repeated here.

[0127] In one embodiment, a new energy output prediction and demand response optimization device based on an LSTM-BN network is provided, comprising:

[0128] The acquisition module is used to obtain the target meteorological information and new energy output information of the area where the new energy power station is located in the same historical period;

[0129] The prediction module is used to predict the energy output of the renewable energy power station based on the target meteorological information and renewable energy output information, and to obtain the target output by using the LSTM-BN prediction model;

[0130] The optimization module is used to build a demand response optimization model based on the target output for the purpose of minimizing energy expenditure costs, solve the demand response optimization model, and obtain the demand response optimization strategy.

[0131] Each module in the above-mentioned new energy output prediction and demand response optimization device based on LSTM-BN network can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0132] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in all the above method embodiments when executing the computer program.

[0133] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in all the above method embodiments are implemented.

[0134] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in all the above method embodiments when executed by a processor.

[0135] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0136] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processors involved in each embodiment provided in this application may be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc., but are not limited thereto.

[0137] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A new energy output prediction and demand response optimization method based on LSTM-BN network, characterized in that: The method comprises: Obtain target meteorological information and new energy output information for the same historical period in the area where the new energy power station is located; Based on the target meteorological information and the new energy output information, the energy output of the new energy power station is predicted by using the LSTM-BN prediction model to obtain a target output; Based on the target output, a demand response optimization model is constructed for the purpose of minimizing energy expenditure cost, and the demand response optimization model is solved to obtain a demand response optimization strategy.

2. The method according to claim 1, characterized in that Obtaining target meteorological information for the area where the new energy power station is located includes: Obtain meteorological data corresponding to several meteorological factors in the area where the new energy power station is located; Obtaining the correlation between the meteorological factors corresponding to the meteorological data and the energy output data through path analysis; The target meteorological information is extracted from the meteorological data according to the correlation.

3. The method according to claim 1, characterized in that The step of predicting the energy output of the new energy power station based on the target meteorological information and the new energy output information and using the LSTM-BN prediction model to obtain the target output includes: Input the target meteorological information and the new energy output information into the LSTM network to extract new energy meteorological time series features; The new energy meteorological time series characteristics and the meteorological characteristics of the time period to be predicted are input into the BN network to predict the target output for the meteorological characteristics.

4. The method according to claim 1, characterized in that The demand response optimization model constructed based on the target output for the purpose of minimizing energy expenditure cost includes: Obtain the energy cost based on the initial investment cost of the cloud energy storage technology and the operating expenses incurred by the additional energy procurement required for the electricity, heating and gas networks; The demand response optimization model is constructed with the goal of minimizing the energy expenditure cost.

5. The method according to claim 4, characterized in that include: Obtain load data on electricity, heating and gas networks, as well as electricity prices and number of users at different times of use; The operating expenses of the additional energy purchase required for the electricity, heating and natural gas networks are obtained according to the load data, the electricity price during the electricity usage period and the number of users.

6. The method according to claim 1, characterized in that Solving the demand response optimization model to obtain the demand response optimization strategy includes: The demand response optimization model is linearized and solved using a MILP solver to obtain the demand response optimization strategy.

7. A new energy output prediction and demand response optimization device based on LSTM-BN network, characterized in that: The device comprises: The acquisition module is used to obtain the target meteorological information and new energy output information of the area where the new energy power station is located in the same historical period; A prediction module, used to predict the energy output of the new energy power station based on the target meteorological information and the new energy output information and using the LSTM-BN prediction model to obtain a target output; The optimization module is used to construct a demand response optimization model based on the target output for the purpose of minimizing energy expenditure cost, solve the demand response optimization model, and obtain a demand response optimization strategy.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images