Power load prediction and supply and demand balance method and system considering climate influence

Abstract

The invention discloses a power load prediction and supply and demand balance method and system considering climate influence. The method comprises the following steps: acquiring historical energy demand data and corresponding temperature data; dividing the historical energy demand data into basic load data and meteorological load data; predicting to obtain a basic load prediction result and a meteorological load prediction result; combining the basic load prediction result and the meteorological load prediction result to obtain an electrical load prediction result; and predicting the supply and demand situation by using the power consumption load prediction result, generating a supply and demand balance strategy based on the recommendation strategy, performing power consumption customer demand response invitation, calculating the response amount of the user when the power consumption customer demand response invitation is performed, and dynamically adjusting the supply and demand balance strategy. According to the method, the demand balance strategy can be formed in a targeted manner, scientific decision support is provided for planning and management of the power system, the whole demand response scheme can be automatically and intelligently generated, and the scheme execution effect is pre-estimated.

Description

Technical Field

[0001] The present invention belongs to the technical field of power supply and demand forecasting, and specifically relates to a method and system for power load forecasting and supply and demand balancing taking climate influence into consideration. Background Art

[0002] With the progress of society, the demand for energy is increasing, and the problem of energy supply and demand is becoming increasingly prominent. Specifically, due to the influence of extreme weather, residential and service electricity users often experience a surge in winter and summer electricity consumption, including winter heating electricity and summer air conditioning and refrigeration electricity. It is also manifested in the influence of holidays, where there will be an increase in residential electricity consumption on non-working days and a decrease in electricity consumption by industrial and service electricity users. At present, the power grid cannot store a large amount of electricity. The imbalance between electricity supply and demand will affect the quality of electricity consumption, and in severe cases, it will endanger the safety and stability of the power grid. In order to cope with the problem of power supply and demand balance, the power grid needs to invite electricity users in the load resource pool in advance based on demand response load forecasting. Within a specific time period, the electricity users participating in the invitation respond to continuous quantitative loads according to the response contract to ensure the balance of the power grid.

[0003] At present, the methods of demand response load forecasting often use mathematical fitting methods or deep learning methods. It is difficult to accurately predict extreme weather using mathematical fitting methods. It can only predict relatively short-term results based on the current load, and there is no way to take into account factors such as periodicity. Load forecasting using deep learning methods can achieve relatively accurate results in medium- and long-term load forecasts, but there is a problem of reduced accuracy in short-term forecasts under sudden meteorological changes. In addition, the formulation of demand response plans is still simplistic, unitary, and does not consider principles such as fairness; the response effect of demand response is lagging, and there is no way to accurately judge whether the demand response invitation balances the supply and demand of the power grid. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a method and system for power load forecasting and supply-demand balancing taking into account climate impact.

[0005] The present invention adopts the following technical solution.

[0006] A first aspect of the present invention provides a method for power load forecasting and supply-demand balancing considering climate impact, comprising:

[0007] Obtain historical energy demand data and corresponding temperature data;

[0008] dividing the historical energy demand data into basic load data and meteorological load data according to the temperature data;

[0009] The basic load data is used to train the deep learning network to obtain the basic load prediction model, and the basic load prediction result is predicted; the meteorological load prediction model is obtained by fitting the meteorological load data and the corresponding temperature data, and the meteorological load prediction result is predicted; the basic load prediction result and the meteorological load prediction result are combined to obtain the power load prediction result;

[0010] The power load forecast results are used to predict the supply and demand situation, and a supply and demand balance strategy is generated based on the recommended strategy. The power customers are invited to respond to the demand and the user's response amount is calculated when the power customers are invited to respond to the demand, and the supply and demand balance strategy is adjusted dynamically.

[0011] Preferably, the historical energy demand data is divided into basic load data and meteorological load data according to the temperature data, as follows:

[0012] BasePow i = {Pow i (t)|10℃<t<20℃}

[0013]

[0014] Pow weather =Pow-Pow base

[0015] Among them, BasePow i is the basic load of user i within the statistical time range;

[0016] Pow i (t) is the set of power loads of user i at temperature t;

[0017] Pow base is the basic load within the statistical time range;

[0018] Pow is the total power load within the statistical time range;

[0019] Pow weather is the meteorological load within the statistical time range;

[0020] n represents the number of all users in the user resource pool.

[0021] Preferably, the method of using the power load forecasting result to forecast the supply and demand situation, generating a supply and demand balancing strategy based on the recommended strategy, and inviting power customers to respond to demand includes:

[0022] Compare the load forecast results with the guaranteed load data to obtain the gap period and maximum gap amount;

[0023] The time period of the gap and the maximum gap amount are combined with the recommendation strategy to generate a supply and demand balance strategy, recommend demand response users from the demand response resource pool, and send demand response invitations to electricity users in batches according to the recommendation list.

[0024] Preferably, the recommendation strategy considers 7 indicators of the user: fairness Historical performance reliability Adjustable potential Two high attributes P1, fast up and down capability P2 and monthly electricity consumption The demand response capability of each user is obtained by weighted summing the 7 indicators. In this way, the user recommendation list is dynamically adjusted to respond to demand, forming a supply-demand balance strategy;

[0025] Specific, demand responsiveness The calculation formula is:

[0026]

[0027] Where k1-k7 are dynamically set weights.

[0028] Preferably, the fairness calculation formula is:

[0029]

[0030] Where P all represents the number of demand response events in the previous ten days, P attend is the number of times user i participates in demand response.

[0031] Preferably, the calculation formula for historical performance is:

[0032]

[0033] Where L invited is the number of responses to the invitation from user i to participate in the demand response, L response is the actual response amount of user i participating in demand response.

[0034] Preferably, the calculation formula of reliability is:

[0035]

[0036] Where P attend is the number of times user i participates in demand response, P success is the probability that user i participates in demand response and responds successfully.

[0037] Preferably, the calculation formula for adjustable potential is:

[0038]

[0039] Where T max It is the maximum gap between load forecast and guaranteed load;

[0040] The adjustable potential for user i;

[0041] t1, t2, and t3 represent the morning, mid-day, and evening periods, respectively;

[0042] R(T) represents the overlap between the gap period T and the early, mid, and late periods;

[0043] is the baseline load of user i at time point j;

[0044] X i represents the initial value of the adjustable potential of user i;

[0045] n represents the number of all users in the user resource pool.

[0046] Preferably, the calculation formula of the two high attributes is P1=wh*H;

[0047] Among them, H indicates whether the user has the "two high attributes", 1 means yes, 0 means no, and the weight wh is 1;

[0048] The calculation formula for the ability to quickly go up and down is P2 = wf*F;

[0049] Among them, F indicates whether the user has the "fast up and down ability", 1 means yes, 0 means no, and the weight wf is 1;

[0050] Preferably, the calculation formula for monthly electricity consumption is:

[0051]

[0052] in Indicates the monthly maximum electricity consumption of user i.

[0053] Preferably, the calculation formula of the user's response amount is as follows:

[0054]

[0055] in The predicted user response volume;

[0056] Ability to respond to user needs;

[0057] m random is a randomly generated random number between [0,1];

[0058] m1 is the statistical probability of users responding to demand;

[0059] m2 is the statistical probability that the user does not respond to the demand.

[0060] A1 and A2 are random numbers between [0.8,1.2].

[0061] A second aspect of the present invention provides a power load forecasting and supply-demand balancing system considering climate impact, comprising:

[0062] An information acquisition module is used to obtain historical energy demand data and corresponding temperature data;

[0063] An influencing factor analysis module, used for dividing the historical energy demand data into basic load data and meteorological load data according to the temperature data;

[0064] The power load forecasting module is used to train the deep learning network with basic load data to obtain a basic load forecasting model, and predict the basic load forecasting results; fit the meteorological load forecasting model according to the meteorological load data and the corresponding temperature data, and predict the meteorological load forecasting results; combine the basic load forecasting results and the meteorological load forecasting results to obtain the power load forecasting results;

[0065] The supply and demand balance strategy generation module is used to use the power load forecast results to predict the supply and demand situation, and generate a supply and demand balance strategy based on the recommended strategy, invite power customers to respond to demand and calculate the user's response amount when inviting power customers to respond to demand, and dynamically adjust the supply and demand balance strategy.

[0066] A third aspect of the present invention provides a terminal, comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method.

[0067] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method when executed by a processor.

[0068] Compared with the prior art, the beneficial effects of the present invention include at least:

[0069] The present invention mainly uses deep learning methods for prediction, analyzes the impact of temperature factors on electricity consumption on the demand side to make predictions, superimposes the basic load predicted by the deep learning method and the meteorological load obtained by fitting the temperature factors to obtain the electricity load prediction result, realizes the correction of the basic load by meteorological information, and improves the accuracy of electricity load prediction.

[0070] The present invention forms a demand balancing strategy based on a recommendation strategy, wherein the recommendation strategy takes into account 7 indicators of users: fairness, historical performance, reliability, adjustable potential, two high attributes, fast up and down capabilities, and monthly electricity consumption; by dynamically setting weights for the 7 indicators, the weighted sum is used to obtain the demand response capability of each user, thereby dynamically adjusting the demand response user recommendation list, and the entire demand response plan can be automatically and intelligently generated; and when inviting electricity users to respond to demand and when inviting electricity users to respond to demand, the user's response amount is calculated, and it is accurately judged whether the demand response invitation balances the supply and demand of the power grid, and the supply and demand balancing strategy is further dynamically adjusted to provide scientific decision support for power system planning and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is the overall flow chart of the present invention.

[0072] Figure 2 It is a flow chart of neural network model training of the present invention.

[0073] Figure 3 It is a module composition collaboration diagram of the present invention.

[0074] Figure 4 It is the temperature and meteorological load fitting diagram of the present invention.

[0075] Figure 5 This is an example of comparing the prediction results of the prediction model. Figure 1 .

[0076] Figure 6 This is an example of comparing the prediction results of the prediction model. Figure 2 .

[0077] Figure 7 It is a hardware structure diagram of the present invention. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0079] Embodiment 1 of the present invention provides a method for power load forecasting and supply-demand balancing considering the impact of climate change, which assists the power system in reasonably formulating a demand response strategy for efficient and sustainable development, such as Figure 1 As shown, the method includes:

[0080] Step 1: Obtain historical energy demand data and corresponding temperature data;

[0081] Further preferably, a variety of meteorological data in the target area can be obtained, including but not limited to temperature, humidity, precipitation and wind speed, etc. Since the temperature influence weight is relatively high, the temperature data is mainly considered in the specific implementation;

[0082] Obtain energy demand information for the target area, including load information for each electricity user in the area within a specific period of time;

[0083] It also obtains the energy supply information (i.e., guaranteed load data / power generation) of the target area, including the power supply load information of the area within a specific time;

[0084] In this embodiment, power supply data, historical power load data and temperature data in the target area are obtained as follows:

[0085] The historical power load data of a certain area from January 1, 2022 to December 31, 2023 (one sampling point every 15 minutes) and the temperature data from January 1, 2022 to January 2, 2024 (temperature, humidity, rainfall and wind speed, one sampling point every 15 minutes) are obtained from the system.

[0086] Step 2: dividing the historical energy demand data into basic load data and meteorological load data according to the temperature data;

[0087] Further preferably, the acquired data can be preprocessed and the influencing factors analyzed to divide the energy demand data into basic load data and meteorological load data;

[0088] The data preprocessing includes: aggregating energy supply and demand data into daily, weekly and monthly load data; interpolating and aligning the load data and corresponding temperature data to unify the units; and managing, caching and updating the preprocessed data sets.

[0089] The influencing factor analysis includes: analyzing the main influencing factors affecting load changes, including user characteristics, holidays, weekends and meteorological factors; forming a user portrait based on the main influencing factors, and dividing the load data into a basic load not affected by meteorological factors and a meteorological load related to meteorological factors.

[0090] The characteristics of demand-side electricity load include: basic load without the influence of meteorological factors; meteorological load related to meteorological factors.

[0091] Basic load data refers to the part of the electricity load that is not related to meteorological factors, which is obtained by statistically analyzing the data of that year. The specific formula is as follows:

[0092] BasePowi = {Pow i (t)|10℃<t<20℃}

[0093]

[0094] Pow weather =Pow-Pow base

[0095] Among them, Pow refers to the total power load within the statistical time range;

[0096] Pow(t) refers to the set of electrical loads at temperature t;

[0097] Pow base Refers to the base load within the statistical time range;

[0098] Pow weather Refers to the meteorological load within the statistical time range.

[0099] In this embodiment, the demand-side power load is decomposed according to meteorological factors, including: basic load without meteorological factors; meteorological load related to meteorological factors. By statistically selecting power load data and temperature data from January 1, 2022 to January 1, 2024 for statistical analysis, the statistical results are as follows:

[0100] Table 1

[0101]

[0102]

[0103] Step 3: Use basic load data to train the deep learning network to obtain a basic load forecasting model, and predict the basic load forecasting results; fit the meteorological load forecasting model according to the meteorological load data and the corresponding temperature data, and predict the meteorological load forecasting results; combine the basic load forecasting results and the meteorological load forecasting results to obtain the power load forecasting results;

[0104] Further preferably, the load forecasting method includes:

[0105] (1) Predict the base load of power users through a neural network-based time series forecasting model;

[0106] The system provides a variety of neural network models, including MLP network, LSTM network and Transformer network. According to the performance on the validation set, the model with higher accuracy is selected as the final prediction model. The calculation formula for model evaluation is as follows:

[0107]

[0108] Among them, y i represents the actual load value, y pred Represents the predicted load value. MSE represents the gap between the true value and the predicted value. The smaller the value, the higher the accuracy of the model prediction.

[0109] Considering the possibility that the user's electricity consumption pattern may change, the model platform periodically updates the load data used for training. The load forecasting model is added to the automatic training plan, and its update cycle and start time are specified for model training. This ensures the accuracy of the neural network model prediction.

[0110] In this embodiment, the basic load of the power unit is trained and predicted by a time series prediction model based on a neural network.

[0111] In the daily prediction model, after training, the neural network model can predict the basic electricity load for the next two days based on the basic electricity load for the previous seven days.

[0112] Specifically, the training process of the neural network time series model uses the historical power data of the previous two years as the training data set. The training set is divided into n training data with 15 minutes as a sampling point. The input of the model prediction of each training data is the load data of the previous 7 days, so the input of the model is 672 points, and the output of the model is the load data of the next 2 days, so the output of the model is 192 points.

[0113] When predicting load data for multiple days, the model uses the model's prediction results as part of the input and recursively predicts the next load data.

[0114] The specific process of model training is as follows Figure 2 shown.

[0115] (2) Use a climate impact model based on statistical analysis to make meteorological load corrections to electricity load.

[0116] In this embodiment, a climate impact model based on statistical analysis is used to perform polynomial fitting on the meteorological load, and the formula is as follows:

[0117] X t =A×t 4 +B×t 3 +C×t 2 +D×t+E

[0118] Among them, t represents the temperature data at the current moment, X t represents the fitting value of meteorological load under current meteorological conditions. The fitting result is as follows Figure 4 As shown, the fitting polynomial is as follows:

[0119] X t =-0.01×t 4 +0.65×t 3 -11.36×t 2 -51.15×t+1664.45

[0120] (3) In this embodiment, the prediction results of (1) and (2) are combined to obtain the specific power load. Compared with the current technology, the neural network-based time series prediction model can predict the basic law of the basic load. On the basis of ensuring the power consumption law, it takes into account the impact of meteorology and solves the problem of low prediction accuracy caused by the surge in power load due to extreme weather. Figure 5 , Figure 6 shown.

[0121] Step 4: Use the power load forecast results to predict the supply and demand situation, and generate a supply and demand balance strategy based on the recommended strategy, invite power customers to respond to demand, calculate the user's response amount when inviting power customers to respond to demand, and dynamically adjust the supply and demand balance strategy.

[0122] The power load forecast results are used to predict the supply and demand situation, and a supply and demand balance strategy is generated based on the recommended strategy to invite power customers to respond to demand, including:

[0123] The results of load forecasting are used to predict the supply and demand situation and obtain the time period of the gap and the maximum gap amount; the time period of the gap and the maximum gap amount are combined with the recommendation strategy to generate a supply and demand balance strategy, and demand response users are recommended from the demand response resource pool. Demand response invitations are issued to electricity users in batches according to the recommendation list.

[0124] The characteristics of electricity users that are considered during the formulation of the supply and demand balancing strategy include at least: adjustable potential to measure the user's ability to respond during the execution period; the average execution status of the user's demand response executions in the previous ten execution days; the number of times the user participated in the demand response in the previous ten execution calendars; the probability that the user's response is effective in the demand response in the previous ten execution days; whether the user is a "two highs" user; whether the user's adjustable load has the characteristic of fast increase and decrease; and the user's monthly usage level.

[0125] The recommended strategy then considers seven indicators of the user: fairness, historical performance, reliability, adjustable potential, two-high attributes, fast up and down capabilities, and monthly electricity consumption;

[0126] The demand response capability of each user is obtained by weighted summing the 7 indicators. In this way, the user recommendation list is dynamically adjusted to respond to demand, forming a supply-demand balance strategy;

[0127] By dynamically setting weights for the seven indicators, the user recommendation list for demand response is dynamically adjusted;

[0128] Demand responsiveness The calculation formula is:

[0129]

[0130] Where k1-k7 are dynamically set weights.

[0131] The specific indicators are as follows:

[0132] (1) Fairness is:

[0133]

[0134] Where P all represents the number of demand response events in the previous ten days, P attend is the number of times user i participates in demand response;

[0135] (2) Historical performance:

[0136]

[0137] Where L invited is the number of responses to the invitation from user i to participate in the demand response, L response is the actual response amount of user i participating in demand response;

[0138] (3) Reliability:

[0139]

[0140] Where P attend is the number of times user i participates in demand response, P success is the probability that user i participates in demand response and responds successfully;

[0141] (4) Adjustable potential:

[0142]

[0143] in The adjustable potential for user i;

[0144] t1, t2, and t3 represent the morning, mid-day, and evening periods, respectively;

[0145] R(T) represents the coincidence of the adjustment time with the morning, waist, and evening periods;

[0146] L j i is the baseline load of user i at time point j; according to the above piecewise function, such as T max= t3, which is the evening period, j is equal to any time point between 17:00 and 22:00;

[0147] X i represents the adjustable potential value of user i;

[0148] (5) Two high attributes: Two high users should be prioritized in the demand response invitation user list; two high attributes P1 = wh*H;

[0149] Among them, H indicates whether the user has the "two high attributes", 1 means yes, 0 means no, and the weight wh is 1;

[0150] (6) Fast on / off capability: Industrial electricity users that support fast on / off are given priority in the demand response invitation user list;

[0151] Fast up and down ability P2 = wf*F;

[0152] Among them, F indicates whether the user has the "fast up and down ability", 1 means yes, 0 means no, and the weight wf is 1;

[0153] (7) Monthly electricity consumption:

[0154]

[0155] in Indicates the monthly maximum electricity consumption of user i.

[0156] When inviting electricity users to respond to demand, calculate the user's response amount and adjust the scope of the response invitation;

[0157] Further preferably, the formula for predicting the user's response amount is as follows:

[0158]

[0159] in The predicted user response volume;

[0160] Ability to respond to user needs;

[0161] m random It is a randomly generated random number. Generally, users will respond to the demand or not. The random number here is a probability from 0 to 1, indicating the probability of the user responding;

[0162] m1 is the statistical probability of users responding to demand;

[0163] m2 is the statistical probability that the user does not respond to the demand.

[0164] A1 and A2 are random numbers between [0.8,1.2];

[0165] Further preferably, the method may further include: Step 5: after the demand response is completed, re-evaluating the demand response capability of each user and analyzing the regional demand response adjustment capability, including:

[0166] After the demand response is completed, the execution effect of the demand response is counted, the indicators of each user in the recommended strategy are updated, and the demand response capability of each user is re-evaluated;

[0167] Evaluate the actual implementation effects of regions and industries, count the overall response effects of all users in the region (actual total load reduction / planned total load reduction), analyze the regional demand response and regulation capabilities, analyze the corresponding contributions of different time periods and different user types, and evaluate the balance within the region.

[0168] Embodiment 2 of the present invention provides a power load forecasting and supply and demand balancing system considering climate impact, the system comprising: an information acquisition module, a data preprocessing module, an influencing factor analysis module, a power load forecasting module, a supply and demand balancing strategy generation module, a demand response effect estimation module and a demand response execution module;

[0169] (1) Information acquisition module:

[0170] The information acquisition module includes the acquisition of energy supply information (load data for guaranteed supply), the acquisition of energy demand information, and the acquisition of various meteorological data. In general, this module is an entire collection system. In the actual implementation process, the data of this module are all obtained from external systems, such as meteorological data from the China Meteorological Administration. The basic electricity consumption data of power users is 15-minute load data, while the meteorological data is obtained as 1-hour meteorological collection data. In order to make the data granularity of meteorological data and electricity consumption data correspond to each other, the data preprocessing module is required to interpolate the meteorological data, and the interpolation method generally adopts the linear interpolation method.

[0171] (2) Data preprocessing module

[0172] The data preprocessing module mainly performs interpolation alignment on the initial data and unifies the units. The 15-minute load data is aggregated into daily, weekly, and monthly load data. The aggregation methods include taking the average, maximum, or median. In order to ensure that the prediction result of the load data after aggregation is greater than the prediction result before aggregation during the prediction process, the maximum value within the aggregation time period is taken.

[0173] The directly aggregated data is made into a training set, which has a certain overlap, which is called the sampling step. If the sampling step is 1, the input and output data will be shifted 15 minutes later than the previous training data, so the overlap between the two training data will be high, and the amount of training data will increase a lot, but this also means that the data will be very repetitive and the training efficiency will be very poor.

[0174] In addition, the data preprocessing module is also responsible for managing and caching the data set to facilitate model training. The data set will update the data of the information acquisition module in a timely manner after a certain period of time to ensure the accuracy of the model prediction.

[0175] (3) Influencing factor analysis module

[0176] The influencing factor analysis module analyzes the main factors affecting load changes by analyzing the characteristics of the load data collected by the information acquisition module. The main influencing factors of load include user characteristics, holiday impact, weekend impact and meteorological factors.

[0177] This module includes the analysis of each user and the extraction of user characteristics to form a user profile, including the daily maximum load, daily minimum load, daily peak-to-valley difference, workday type, and seasonal load changes.

[0178] The impact of holidays is analyzed by comparing the load curve changes on holidays and before and after holidays, and analyzing the impact of holidays on the region and industry. This includes analyzing the maximum load level before and after holidays, load characteristics before and after holidays, such as the maximum load during the morning peak, and the time of occurrence.

[0179] Similarly, working days and rest days have a great impact on certain industries, which has the following characteristics: Generally speaking, the electricity consumption on rest days of the whole society will be about 1 / 5 lower than that on working days; the electricity consumption of urban and rural residents at home will increase instead; the difference between working days and rest days has a greater impact on industry and services, but a weaker impact on agriculture and animal husbandry, etc.

[0180] The meteorological factor analysis module includes the fitting analysis of meteorological information such as temperature and meteorological load, which is used to fine-tune the prediction results. Meteorological data include temperature, humidity, precipitation and wind speed, which need to be fitted separately. The results of the four fittings can be weighted and summed, and the weight of each meteorological influence is different. Since the weight of temperature influence is the highest, this embodiment only considers the influence of temperature. An example of the fitting result of meteorological load and temperature is as follows: Figure 4 shown.

[0181] (4) Power load prediction module

[0182] This module is mainly responsible for the management of the load forecasting module based on deep learning, including model training, automatic training management, model training result comparison and multiple forecasts.

[0183] The power load forecasting module includes forecasting tasks of various time granularities, including 15-minute, daily, weekly, and monthly forecasts. For training tasks of different time granularities, the input and output sizes of the models are also different. For example, the recommended input number for the 15-minute model is 672 and the output number is 192, the recommended input number for the daily model is 28 and the output number is 7, the recommended input number for the weekly model is 12 and the output number is 4, and the recommended input number for the monthly model is 3 and the output number is 1. In comparison, the 15-minute forecasting task has the highest level of sophistication and the most forecast inputs and outputs. The monthly model forecast uses less data and has a lower forecast accuracy.

[0184] (5) Supply and demand balance strategy generation module

[0185] This module predicts the supply and demand situation by using the results of load forecasting, and obtains the load supply of the whole period from the information collection module. By predicting the supply and demand situation, the time period and the maximum gap amount of the gap are obtained. This guides the demand response to digest and decompose the gap. For example, a province will have a load gap of 1 million kW from 16:00 to 19:00. In order to digest the gap in the province, the gap will be decomposed to each city according to the capacity of each city. For example, a city is responsible for a load gap of 200,000 kW. In order to ensure that the invitation guarantee is completed, the invitation will be responded to at a 1.2 times invitation coefficient. Then the city needs to respond to demand from 16:00 to 19:00 to reduce electricity consumption by 240,000 kW.

[0186] The supply and demand balance strategy generation module determines which users are most suitable for demand response invitations based on the response demand. Users suitable for demand response in the demand response resource pool will be recommended first, and the system will issue demand response invitations in batches according to the recommendation list. The recommendation strategy should comply with the following principles:

[0187] Fairness: The number of times users participate in demand response should be reasonably distributed so that users of the same size have the same chance to be invited to participate in demand response. The higher the index, the easier it is to be selected. The calculation formula is as follows:

[0188]

[0189] Where P all represents the number of demand response events in the previous ten days, P attend The number of times the user participated in the demand response. The larger the index is, the more likely the user is to be selected.

[0190] Historical performance: There is another indicator for evaluating the user's response effect, which indicates the performance of the user in performing demand response. The formula is as follows:

[0191]

[0192] Where L invited is the number of responses to the user's invitation to participate in the demand response, L response The actual amount of responses for this user to participate in demand response. The larger the indicator is, the more likely the user is to be selected.

[0193] Reliability: This indicator describes the execution effect of the user's participation in demand response. The better the execution effect, the more likely it is to complete the demand response. Such users should be recommended. The calculation formula is as follows:

[0194]

[0195] Where P attend is the number of times the user participates in demand response, P success The probability of the user participating in the demand response and successfully responding. When the actual response volume is between 0.8-1.2 times the invited response volume, it is considered a successful response. The larger the index is, the more likely the user is to be selected.

[0196] Adjustable potential: Under the same demand response invitation, the fewer users affected, the better. The one with greater ability to execute demand response period is more likely to be selected. The formula is as follows:

[0197]

[0198] Among them, t1, t2, and t3 represent the three periods of early waist and late respectively, R(T) represents the overlap between the adjustment time and the three periods of early waist and late, and L i is the baseline load of the ith user, with a total of 24 points. i represents the adjustable potential of the i-th user, and then scales the adjustable potential to the range of [0,1].

[0199] Priority for users with high-energy consumption and high-dose attributes: users with high-energy consumption and high-dose attributes should be prioritized in the demand response invitation user list.

[0200] Fast up and down capability / fast up and down priority: Industrial electricity users that support fast up and down are given priority in the demand response invitation user list. Their fast up and down capability can quickly balance the supply and demand of the power grid.

[0201] Monthly electricity consumption: The ones with higher monthly electricity consumption have stronger adjustable capabilities.

[0202]

[0203] in Indicates the maximum monthly electricity consumption of user i. The larger the index is, the more likely the user is to be selected.

[0204] Based on the above principles, this module considers 7 indicators of users: adjustable potential, historical performance, fairness, reliability, two high attributes, fast up and down capabilities, and monthly electricity consumption. By dynamically setting weights for the seven indicators, the list of invited users can be dynamically adjusted.

[0205] (6) Response effect estimation module

[0206] When intelligently inviting electricity users to respond to demand, this module can make an estimated analysis of the results of the demand response and provide a regional electricity consumption forecast after the demand response is executed.

[0207] In actual demand response, users do not strictly follow the response invitation, and each type of user has different response performance. The user's response invitation is predicted based on the historical performance of each user. The specific formula is as follows:

[0208]

[0209] in, To provide users with the ability to respond to their needs, is the predicted user response volume, m random is a randomly generated random number, m1 is the statistical probability that the user responds to the demand, and m2 is the statistical probability that the user does not make a negative response.

[0210] When the user responds effectively, A1 is a random number between [0.8, 1.2], indicating that the user has responded to the demand response invitation and reduced electricity consumption.

[0211] When the user does not respond, it means that the user does not consider responding to the invitation and continues to use electricity normally.

[0212] In some cases, some users will have a negative response, and the user will increase electricity consumption.

[0213] After calculating the response results of each user, the power consumption curve of the area will be generated according to the predicted response results. By comparing with the power generation, we can have a general understanding of the actual response results and adjust the scope of the response invitation.

[0214] (7) Demand response execution module

[0215] For devices with demand response control capabilities, the system can perform rigid or flexible control on the target device, and the control status of the specific device can be viewed through loop monitoring.

[0216] After the demand response is completed, the demand response execution module will collect statistics on the execution effect of the demand response, update the historical performance, reliability, fairness and other demand response indicators of each user, and re-evaluate the demand response capabilities of each power user. It will also evaluate the actual execution effects of regions and industries and analyze the regional demand response regulation capabilities.

[0217] Embodiment 3 of the present invention provides a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method.

[0218] Embodiment 4 of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method when executed by a processor.

[0219] Figure 7 A schematic diagram of a storage medium provided in an optional embodiment of the present invention. Figure 7 As shown in the figure, the hardware is generally divided into three parts, namely, the server, the smart energy unit and the adjustable and controllable device. The server is equipped with a central processing unit, an image processor, a memory and a network adapter. Inside the server, each unit communicates with each other through a bus. The processor can process the instructions stored in the memory and communicate with the smart energy unit through the network adapter.

[0220] The server includes an image processor, which is used for heterogeneous parallel processing of large-scale data calculations, including the training and reasoning of deep learning models used in this embodiment.

[0221] The smart energy unit is responsible for issuing control instructions to adjustable and controllable devices. Its control process and results are displayed on the software terminal through the server. The demand response execution module performs statistics and analysis of the effects.

[0222] The present invention provides a computer readable medium, or a computer code originally stored in a remote storage medium and downloaded through a network. The method described herein can be stored in a server using a general-purpose computer, a dedicated processor and a memory for processing.

[0223] An embodiment of the present invention provides a computer program, which is stored in a machine-readable storage medium. The program is executed by at least one processor to implement the various processes of the above-mentioned method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0224] Compared with the prior art, the beneficial effects of the present invention include at least:

[0225] The present invention mainly uses deep learning methods for prediction, analyzes the impact of factors such as temperature on demand-side electricity consumption to make electricity load forecasts, and forms a demand balancing strategy based on the recommendation strategy; the recommendation strategy takes into account 7 indicators of users: fairness, historical performance, reliability, adjustable potential, two high attributes, fast up and down capabilities, and monthly electricity consumption; by dynamically setting weights for the 7 indicators, the weighted sum is used to obtain the demand response capability of each user, and the demand response user recommendation list is dynamically adjusted to automatically and intelligently generate the entire demand response plan; and when inviting electricity users to respond to demand and when inviting electricity users to respond to demand, the user's response amount is calculated, and it is accurately judged whether the demand response invitation balances the supply and demand of the power grid, and the supply and demand balancing strategy is further dynamically adjusted to provide scientific decision support for power system planning and management.

[0226] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0227] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0228] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0229] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0230] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for power load forecasting and supply-demand balance considering climate impact, characterized in that: include: Obtain historical energy demand data and corresponding temperature data; dividing the historical energy demand data into basic load data and meteorological load data according to the temperature data; The basic load data is used to train the deep learning network to obtain the basic load prediction model, and the basic load prediction result is obtained; A meteorological load prediction model is obtained by fitting the meteorological load data and the corresponding temperature data, and a meteorological load prediction result is obtained by prediction; The basic load forecast result and the meteorological load forecast result are combined to obtain the power load forecast result; The power load forecast results are used to predict the supply and demand situation, and a supply and demand balance strategy is generated based on the recommended strategy. The power customers are invited to respond to the demand and the user's response amount is calculated when the power customers are invited to respond to the demand, and the supply and demand balance strategy is adjusted dynamically.

2. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 1, characterized in that: The historical energy demand data is divided into basic load data and meteorological load data according to the temperature data, as follows: BasePow i ={Pow i (t)|10℃<t<20℃} Pow weather =Pow-Pow base Among them, BasePow i is the basic load of user i within the statistical time range; Pow i (t) is the set of power loads of user i at temperature t; Pow base is the basic load within the statistical time range; Pow is the total power load within the statistical time range; Pow weather is the meteorological load within the statistical time range; n represents the number of all users in the user resource pool.

3. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 1, characterized in that: The method of using the power load forecasting result to forecast the supply and demand situation, generating a supply and demand balance strategy based on the recommended strategy, and inviting power users to respond to demand includes: Compare the load forecast results with the guaranteed load data to obtain the gap period and maximum gap amount; The time period of the gap and the maximum gap amount are combined with the recommendation strategy to generate a supply and demand balance strategy, recommend demand response users from the demand response resource pool, and send demand response invitations to electricity users in batches according to the recommendation list.

4. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 3, characterized in that: The recommendation strategy considers 7 indicators of users: fairness Historical Performance reliability Adjustable potential Two high attributes P1, fast up and down capability P2 and monthly electricity consumption The demand response capability of each user is obtained by weighted summing the 7 indicators. In this way, the user recommendation list is dynamically adjusted to respond to demand, forming a supply-demand balance strategy; Demand responsiveness The calculation formula is: Where k1-k7 are dynamically set weights.

5. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 4, characterized in that: The calculation formula for fairness is: Where P all represents the number of demand response events in the previous ten days, P attend is the number of times user i participates in demand response.

6. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 4, characterized in that: The historical performance is calculated as: Where L invited is the number of responses to the invitation from user i to participate in the demand response, L response is the actual response amount of user i participating in demand response.

7. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 4, characterized in that: The reliability calculation formula is: Where P attend is the number of times user i participates in demand response, P success is the probability that user i participates in demand response and responds successfully.

8. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 4, characterized in that: The formula for calculating adjustable potential is: Where T max It is the maximum gap between load forecast and guaranteed load; The adjustable potential for user i; t1, t2, and t3 represent the morning, mid-day, and evening periods, respectively; R(T) represents the overlap between the gap period T and the early, mid, and late periods; is the baseline load of user i at time point j; X i represents the initial value of the adjustable potential of user i; n represents the number of all users in the user resource pool.

9. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 4, characterized in that: The calculation formula for the two high attributes is P1 = wh*H; Among them, H indicates whether the user has the "two high attributes", 1 means yes, 0 means no, and the weight wh is 1; The calculation formula for the ability to quickly go up and down is P2 = wf*F; Among them, F indicates whether the user has the "fast up and down ability", 1 means yes, 0 means no, and the weight wf is 1.

10. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 4, characterized in that: The formula for calculating monthly electricity consumption is: in Indicates the monthly maximum electricity consumption of user i.

11. The method for power load forecasting and supply-demand balancing considering climate impact according to claim 1, characterized in that: The calculation formula of the user's response amount is as follows: in The predicted user response volume; Ability to respond to user needs; m random is a randomly generated random number between [0,1]; m1 is the statistical probability of users responding to demand; m2 is the statistical probability that the user does not respond to the demand. A1 and A2 are random numbers between [0.8,1.2].

12. A power load forecasting and supply-demand balancing system considering climate impact, using the method according to any one of claims 1 to 11, characterized in that: The system comprises: An information acquisition module is used to obtain historical energy demand data and corresponding temperature data; An influencing factor analysis module, used for dividing the historical energy demand data into basic load data and meteorological load data according to the temperature data; The power load forecasting module is used to train the deep learning network with basic load data to obtain a basic load forecasting model, and predict the basic load forecasting results; fit the meteorological load forecasting model according to the meteorological load data and the corresponding temperature data, and predict the meteorological load forecasting results; combine the basic load forecasting results and the meteorological load forecasting results to obtain the power load forecasting results; The supply and demand balance strategy generation module is used to use the power load forecast results to predict the supply and demand situation, and generate a supply and demand balance strategy based on the recommended strategy, invite power customers to respond to demand and calculate the user's response amount when inviting power customers to respond to demand, and dynamically adjust the supply and demand balance strategy.

13. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images