Power load and electricity price collaborative prediction method and system based on deep learning

Abstract

The invention discloses an electric power load and electricity price collaborative prediction method and system based on deep learning, and relates to the technical field of electric power systems.The method comprises the steps that electric power state data are collected in real time through a sensor set arranged in electric power distribution equipment, and data processing is carried out; acquiring a power disturbance data set, calculating a voltage lag feedback index dyx, performing power supply disturbance evaluation, executing a coupling response trend prediction instruction, and acquiring a disturbance characteristic data set and calculating a synchronous response excitation index tbj according to real-time extraction and dimensionless processing of a special API (Application Program Interface); cooperative disturbance evaluation is carried out by calculating a comprehensive cooperative prediction index xty, and when it is detected that the load and the electricity price have a cooperative disturbance trend, early warning is automatically generated, and price adjustment simulation, a negative control strategy and a standby power supply scheduling mechanism are linked. The method can effectively improve the prediction precision and the regulation and control response capability of the power grid under multi-factor disturbance, and improves the operation stability and economical efficiency of the smart power grid.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power systems, in particular to a power load and price collaborative prediction method and system based on deep learning. BACKGROUND

[0002] In a modern power system, the dynamic fluctuation of power load and the timing adjustment of price mechanism present a highly complex nonlinear coupling relationship. With the access of renewable energy, the diversification of user behavior and the refinement of market price mechanism, the power system is evolving from the traditional "load and power generation matching" to "price and load response collaborative scheduling". In this process, how to accurately predict the trend of power load change and the trend of price change becomes the key to supporting the regulation and optimal operation of smart grids. As a comprehensive prediction task that integrates physical state, economic signal and disturbance response mechanism, the key to "power load and price collaborative prediction" lies in establishing a deep perception mechanism of multi-source disturbance characteristics and identifying potential collaborative response relationships, so as to realize the synchronous optimization of energy scheduling and price strategy.

[0003] In the Chinese patent application with the application publication number CN120013587A, a multi-region power spot market node price collaborative analysis method is disclosed, which relates to the technical field of power markets and includes obtaining historical data of each power market region, the historical data including node prices, supply and demand data, and preprocessing the historical data; constructing a direct current flow model to calculate the transmission network state information between nodes based on the direct current flow model to describe the power flow distribution of the power system; constructing a price prediction model based on a machine learning algorithm according to the supply and demand data and the transmission network state information to obtain node price prediction values; and optimizing power scheduling between different regions according to the price prediction model and node prices, and dynamically adjusting the price prediction model. The application provides a multi-region power spot market node price collaborative analysis method, which solves the problem of low efficiency of cross-regional power scheduling and resource allocation, realizes the collaborative optimization of node prices in the multi-region power spot market, and improves the market operation efficiency and economy.

[0004] In combination with the prior art, the above application also has the following deficiencies: The above application has cross-regional node electricity price coordination and prediction ability, but has limited performance in actual power disturbance response and load behavior evolution, lacks dynamic identification and quantitative mechanism for the coupling relationship between electricity price disturbance and load response, and is not easy to realize accurate tracking of load lag response behavior and nonlinear analysis of synchronous electricity price disturbance. In addition, the above application focuses on the structural modeling of static data, ignores key characteristics such as short-term disturbance characteristics, non-steady-state behavior, and collaborative evolution trends of complex systems, and lacks an integrated processing mechanism for multi-source disturbance factors in real-time operation state. Finally, the electricity price prediction in the above application mainly relies on the combination of node basic cost, supply and demand factors, and power flow weights, and uses linear adjustment to correct the prediction model, which is difficult to capture the mutation trend under extreme load fluctuations or high-frequency electricity price disturbances, and lacks a dynamic index system responding to asymmetric coupling characteristics. Therefore, there are still obvious deficiencies in real-time performance, robustness and system disturbance perception, which makes it difficult to provide sufficient support for peak load regulation, abnormal warning and advance pricing strategy. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a power load and electricity price collaborative prediction method and system based on deep learning, which solves the problems in the above background art.

[0006] To achieve the above object, the present application is realized by the following technical scheme: a power load and electricity price collaborative prediction method based on deep learning, comprising the following steps: S1, real-time acquisition of power state data according to the set sensor group, and transmission of the power state data to the power data center through the API interface for data processing to obtain a power disturbance data group; S2, calculation of voltage lag feedback index dyx according to the power disturbance data group, and supply disturbance evaluation with a preset load fluctuation trend threshold G, and execution of a coupled response trend prediction instruction when the supply disturbance evaluation is stable power supply; S3, execution of the coupled response trend prediction instruction, real-time extraction of dynamic disturbance data according to the special API interface, and non-dimensional processing through the power data center to obtain a disturbance characteristic data group, and then calculation of a synchronous response excitation index tbj according to the disturbance characteristic data group; S4, calculation of a comprehensive collaborative prediction index xty according to the voltage lag feedback index dyx and the synchronous response excitation index tbj, and collaborative disturbance evaluation with a preset load electricity price collaborative change risk threshold P.

[0007] Preferably, S1 comprises S11; S11, real-time acquisition of power state data according to the sensor group set in the power distribution equipment, and marking of the acquisition time stamp; The power state data includes voltage V, power distribution cabinet internal temperature wd and active power sequence P; The sensor group includes a voltage sensor, a thermal temperature sensor and a three-phase power meter; The voltage sensor is used to be installed at the transformer outgoing line end and the user access node to collect real-time voltage V in the power system; The thermal temperature sensor is used to be installed in the power distribution cabinet to collect real-time power distribution cabinet internal temperature wd; The three-phase power meter is used to be installed in the electric meter box to monitor real-time active power and form active power sequence P according to the collection time stamp, and the specific form of the active power sequence P is: P={P t , P t+1 , …, P t+n}.

[0008] Preferably, S1 further includes S12; S12 is used to establish a communication connection between the sensor group and the power data center through an API interface, and transmit the acquired power state data to the power data center for data processing in real time; The data processing is used to perform data analysis on the power state data, data preprocessing, and obtain a power disturbance data set; The data analysis includes voltage offset analysis, temperature amplitude analysis and power change analysis; The voltage offset analysis is based on the acquired real-time voltage V and the rated voltage V rated of the corresponding node to perform difference processing and obtain voltage offset value vg, and the specific formula is: vg=|V-V rated |; The temperature amplitude analysis is based on the acquired power distribution cabinet internal temperature wd and the normal temperature wb under the standard state to perform difference processing and obtain power distribution cabinet temperature abnormal amplitude tc, and the specific formula is: tc=|wd-wb|; The power change analysis is based on the acquired active power sequence P to calculate active power change rate pa, and the specific formula is: , wherein, ∆t represents the time interval between P t and P t+1 ; The data preprocessing is used to perform dimensionless processing, denoising and abnormal value processing on the parameters after data analysis; The dimensionless processing is used to process the dimension influence of the power state data by the Max-Min maximization method, the denoising is used to eliminate the noise influence in the power state data by decomposing the signals on different frequency scales by wavelet transform technology, and the abnormal value processing is used to detect and process the abnormal values in the power state data by using the quartile range method; The power disturbance data set comprises a voltage offset value vg, a temperature abnormality amplitude tc of a power distribution cabinet, and an active power change rate pa.

[0009] Preferably, S2 comprises S21; S21, calculating a voltage hysteresis feedback index dyx according to the power disturbance data set, for measuring the sensitivity of load response to power supply system disturbance, and the specific formula is as follows; ; In the formula, ln represents a logarithmic function.

[0010] Preferably, S2 further comprises S22; S22, sorting all voltage hysteresis feedback indexes dyx in the past three months from small to large according to a statistical method, and calculating a 85% quantile value to preset a load fluctuation trend threshold G, and then performing power supply disturbance evaluation with the real-time obtained voltage hysteresis feedback index dyx, and the specific evaluation scheme is as follows: when the voltage hysteresis feedback index dyx is less than or equal to the load fluctuation trend threshold G, it indicates that the current power supply is stable, and the load change is stable, and at this time, the coupling response trend prediction instruction is executed; When the voltage hysteresis feedback index dyx is greater than the load fluctuation trend threshold G, it indicates that there is a load disturbance, at this time, a load warning is generated, and after triggering the load prediction correction and executing the load soft start control strategy, iterative analysis and evaluation are performed through S21.

[0011] Preferably, S3 comprises S31; S31, for executing the coupling response trend prediction instruction when the power supply disturbance evaluation is power supply stability; The coupling response trend prediction instruction is used for extracting dynamic disturbance data in real time according to a special API interface, and transmitting the dynamic disturbance data to a power data platform for dimensionless processing to obtain a disturbance characteristic data set; The disturbance characteristic data set comprises a price switching frequency fp, a bus frequency deviation value fb, and a total harmonic distortion rate ne; The price switching frequency fp is extracted in real time according to the API interface accessing a price strategy pushing platform; The bus frequency deviation value fb is extracted in real time according to the API interface accessing a phasor measurement unit PMU; The total harmonic distortion rate ne is extracted in real time according to the API interface accessing a power quality monitoring module PQM.

[0012] Preferably, S3 further comprises S32; S32, calculating a synchronous response excitation index tbj according to the disturbance characteristic data set, for quantifying the synchronous induction effect of short-time price disturbance on load response, and the specific formula is as follows; ; wherein, cos -1 denotes the inverse cosine function, and ln denotes the logarithmic function.

[0013] Preferably, S4 comprises S41; S41, according to the real-time acquired voltage hysteresis feedback index dyx and the synchronous response excitation index tbj, performs comprehensive calculation to obtain a comprehensive cooperative prediction index xty, which is used to evaluate the coupling reinforcement relationship between hysteresis feedback and synchronous excitation, and the specific formula is as follows: wherein, cos -1 denotes the inverse cosine function, and ∈ denotes a minimum constant for preventing the denominator from being zero, and the value is 0.001.

[0014] Preferably, S4 further comprises S42; S42, according to the statistical method, sorts all comprehensive cooperative prediction indexes xty in the past three months from small to large, calculates the 95% quantile value using the quantile method to obtain a preset load price cooperative change risk threshold P, and then performs cooperative disturbance evaluation on the real-time acquired comprehensive cooperative prediction index xty, and the specific evaluation scheme is as follows: When the comprehensive cooperative prediction index xty is less than or equal to the load price cooperative change risk threshold P, it indicates that there is no cooperative disturbance between the load trend and the price fluctuation, and at this time, normal dispatching operation and monitoring are maintained; When the comprehensive cooperative prediction index xty is greater than the load price cooperative change risk threshold P, it indicates that there is a cooperative disturbance between the load trend and the price fluctuation, at this time, coupling information is generated and transmitted to relevant personnel, prompting the relevant personnel to perform advance price adjustment simulation and load control strategy, and to dispatch standby power supply in advance and start the peak load reduction plan.

[0015] A power load and price cooperative prediction system based on deep learning, comprising a data acquisition module, a power evaluation module, a coupling response analysis module and a cooperative evaluation module; The data acquisition module is used to acquire power state data in real time according to a set sensor group, and transmit the power state data to a power data center through an API interface for data processing to obtain a power disturbance data set; The power evaluation module is used to calculate a voltage hysteresis feedback index dyx according to the power disturbance data set, and perform power disturbance evaluation with a preset load fluctuation trend threshold G, and execute a coupling response trend prediction instruction when the power disturbance evaluation is power stable; The coupling response analysis module is used to execute the coupling response trend prediction instruction, extract dynamic disturbance data in real time according to a special API interface, and perform dimensionless processing through the power data center to obtain a disturbance feature data set, and then calculate a synchronous response excitation index tbj according to the disturbance feature data set; ​The cooperative evaluation module is used for calculating a comprehensive cooperative prediction index xty according to the voltage lag feedback index dyx and the synchronous response triggering index tbj, and performing cooperative disturbance evaluation with a preset load price cooperative change risk threshold P.

[0016] The application provides a power load and price cooperative prediction method and system based on deep learning. (1) The method constructs a complete technical closed loop from bottom layer power state perception to high layer risk prediction. Through S1, voltage sensors, temperature sensors and three-phase power meters are arranged at the transformer outlet end, the power distribution cabinet and the electric meter box, so that fine collection of original data such as voltage V, temperature wd and power sequence P is realized, and the original data are transmitted to the power data center in combination with an API interface. The power data center adopts a multi-dimensional data analysis method to analyze voltage offset, temperature amplitude and power change of the power state data, and effectively constructs a power disturbance data set capable of representing the power disturbance state through dimensionless, denoising and outlier elimination, thereby significantly improving the effectiveness and analyzability of the data.

[0017] (2) In S2, the method calculates the voltage lag feedback index dyx through the power disturbance data set, and performs power disturbance evaluation with the set load fluctuation trend threshold G, so as to realize preliminary quantitative evaluation of the power disturbance degree. When the voltage lag feedback index dyx is less than or equal to the load fluctuation trend threshold G, it indicates that the power is stable, and the coupled response trend prediction instruction is executed. If the voltage lag feedback index dyx is greater than the load fluctuation trend threshold G, a load warning is automatically generated, and the load prediction correction and slow start strategy are executed. In S3, the coupled response trend prediction instruction is executed, real-time dynamic disturbance data are extracted through a special API interface, and are transmitted to the power data center for dimensionless processing to obtain a disturbance feature data set for calculating the synchronous response triggering index tbj, reflecting the induction effect of short-time price disturbance on load response, and highlighting the role mechanism of disturbance frequency, frequency deviation and harmonic disturbance in the system cooperative relationship.

[0018] (3) In S4, the method further integrates the voltage hysteresis feedback index dyx and the synchronous response excitation index tbj, constructs a comprehensive cooperative prediction index xty through the inverse cosine function and normalization processing, and uses it to evaluate the coupling strength between the two. The comprehensive cooperative prediction index xty reflects the asymmetric cooperative relationship between hysteresis feedback and synchronous response, and is subjected to cooperative disturbance evaluation with a preset load price cooperative variation risk threshold P. If the comprehensive cooperative prediction index xty is greater than the load price cooperative variation risk threshold P, it is considered that there is a cooperative disturbance of load trend and price fluctuation. At this time, the coupling information is automatically generated to remind the relevant personnel to carry out advance price simulation, load control, standby power dispatching and start peak shaving plan. Finally, through the four-level process of "data acquisition, disturbance calculation, response excitation and cooperative prediction", the method comprehensively realizes the deep analysis and intelligent prediction of the cooperative relationship between power load and price fluctuation, and improves the forward-looking of dispatching decision and the stability of system operation. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A step schematic diagram of the power load and price cooperative prediction method based on deep learning according to the present application; Figure 2 A system flow schematic diagram of the power load and price cooperative prediction method based on deep learning according to the present application; Figure 3 A cooperative disturbance evaluation broken line schematic diagram according to the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0021] Embodiment 1 Please refer to Figure 1 The present application provides a power load and price cooperative prediction method based on deep learning. To achieve the above purpose, the present application is implemented through the following technical solutions: comprising the following steps: S1, real-time acquisition of power state data according to the set sensor group, and transmission of the power state data to the power data center through the API interface for data processing to obtain a power disturbance data group; S2, calculation of the voltage hysteresis feedback index dyx according to the power disturbance data group, and supply disturbance evaluation with a preset load fluctuation trend threshold G, and execution of a coupling response trend prediction instruction when the supply disturbance evaluation is stable; S3. Execute the coupled response trend prediction instruction, extract dynamic disturbance data in real time based on the dedicated API interface, perform dimensionless processing through the power data center, obtain a disturbance feature data group, and then calculate the synchronous response excitation index tbj based on the disturbance feature data group; S4. Calculate the comprehensive collaborative prediction index xty based on the voltage hysteresis feedback index dyx and the synchronous response excitation index tbj, and perform collaborative disturbance assessment with the preset load electricity price collaborative change risk threshold P.

[0022] In this embodiment, by setting up a multi-source sensor group at the power system site, including voltage sensors, thermal temperature sensors and three-phase power meters, power status data is collected in real time, and transmitted to the power data center through the API interface for data processing, and a power disturbance data group characterizing power supply stability, electrical thermal imbalance and load change behavior is constructed. This process significantly improves the response lag problem caused by traditional manual inspection or periodic meter reading, and lays a foundation for dynamic disturbance analysis. In the power supply disturbance assessment and response trend prediction stage, the voltage hysteresis feedback index dyx is calculated through the power disturbance data group, and the power supply disturbance assessment is carried out with the load fluctuation trend threshold G, realizing sensitive detection and real-time classification judgment of load disturbance events. When the power supply state is stable, the coupled response trend prediction instruction is started, and further the electricity price switching frequency fp, bus frequency deviation fb and total harmonic distortion rate ne are extracted from the electricity price strategy platform, synchronous phasor measurement unit PMU and power quality monitoring module PQM module, and the synchronous response excitation index tbj is calculated to quantify the coupled driving effect of short-term electricity price disturbance on load response. This mechanism breaks through the problem that traditional electricity price prediction mainly focuses on time series fitting and lacks disturbance modeling, and enhances the forward identification ability of sudden electricity price adjustment on system load response. Through the comprehensive collaborative prediction index xty constructed by integrating the voltage hysteresis feedback index dyx and the synchronous response excitation index tbj, and the load electricity price collaborative change risk threshold P set based on the 95% quantile method, the quantitative evaluation of the coupled disturbance relationship can be completed. When the comprehensive collaborative prediction index xty exceeds the load electricity price collaborative change risk threshold P, a number of linkage strategies such as early price adjustment simulation, load peak shaving, and standby power supply scheduling can be triggered to effectively respond to the risk of system coupling instability. Compared with the current strategy mainly based on static rules or isolated prediction models, this method realizes cross-level, multi-parameter, multi-platform data fusion and prediction collaboration, and brings significant improvements in power supply security, prediction accuracy and response efficiency, providing technical support for building an intelligent and highly resilient modern power dispatching system.

[0023] Embodiment 2 Please refer to Figure 1 , specifically: S1 includes S11; S11. Collect power status data in real time based on the sensor group installed on the power distribution equipment, and mark the collection timestamp; The power status data includes voltage V, internal temperature wd of the distribution cabinet, and active power sequence P; The sensor group includes a voltage sensor, a thermistor temperature sensor, and a three-phase power meter; The voltage sensor is used to be installed at the transformer output terminal and user access node to collect the real-time voltage V in the power system; The thermal temperature sensor is installed inside the power distribution cabinet to collect the internal temperature wd of the power distribution cabinet in real time. The three-phase power meter is installed inside the meter box to monitor active power in real time and form an active power sequence P based on the data collection timestamp. The specific form of the active power sequence P is: P = {P t P t+1 ,,,,,P t+n}

[0024] S1 also includes S12; S12 is used to establish a communication connection between the sensor group and the power data platform through the API interface, and transmit the acquired power status data to the power data platform for data processing in real time. The data processing is used to perform data preprocessing on the power status data after data analysis to obtain power disturbance data sets; The data analysis includes voltage offset analysis, temperature amplitude analysis, and power change analysis; Voltage offset analysis is based on the obtained real-time voltage V and the corresponding node's rated voltage V. rated The difference is processed to obtain the voltage offset value vg, which represents the instantaneous deviation of the actual voltage value in the power system from the rated voltage value, reflecting the stability of the power supply. The specific formula is: vg=|VV rated |; Temperature amplitude analysis is based on the obtained internal temperature wd of the distribution cabinet, and then the difference is processed with the normal temperature wb under standard conditions to obtain the abnormal temperature amplitude tc of the distribution cabinet. It represents the degree of abnormal deviation of the internal temperature value of the distribution cabinet from the historical average normal operating temperature, and is used to reflect the abnormal operation and overload conditions. The specific formula is: tc=|wd-wb|. Power change analysis calculates the active power change rate pa based on the obtained active power sequence P, which represents the magnitude of active power change per unit time and reflects load start-up and shutdown and grid dynamic behavior. The specific formula is as follows: In the formula, ∆t represents P t and P t+1 The time interval between; The data preprocessing is used to perform dimensionless processing, noise reduction, and outlier processing on the parameters after data analysis. The dimensionless processing uses the Max-Min method to eliminate the dimensional influence of the power state data. The noise reduction uses wavelet transform technology to decompose signals at different frequency scales to eliminate the noise influence in the power state data. The outlier processing uses the interquartile range method to detect and process outliers in the power state data. The power disturbance data set includes voltage offset value vg, distribution cabinet temperature anomaly amplitude tc, and active power change rate pa.

[0025] In this embodiment, voltage sensors, thermistors, and three-phase power meters are deployed at the transformer outgoing terminals, user access nodes, inside the distribution cabinet, and in the meter box, respectively, to collect real-time power status data, including voltage V, distribution cabinet internal temperature wd, and active power sequence P. The collected data is then timestamped. The data is then transmitted to the power data platform via an API interface. After data analysis, the power status data undergoes preprocessing to form a power disturbance data set. This eliminates dimensional interference and abnormal fluctuations, ensuring the stability and comparability of the disturbance data. This approach not only achieves high-precision, structured acquisition and processing of power status data, improving data quality and system response speed, but also provides a precise foundation for the subsequent calculation of the voltage hysteresis feedback index and synchronization response excitation index. This effectively enhances the entire system's ability to perceive power disturbances and the robustness of subsequent prediction models, significantly outperforming traditional low-frequency monitoring and single-index diagnostic methods.

[0026] Example 3 Please refer to Figure 1 Specifically: S2 includes S21; S21. Calculate the voltage hysteresis feedback index dyx based on the power disturbance data set. This index is used to measure the sensitivity of the load response to power supply system disturbances. The specific formula is as follows: ; In the formula, ln represents the logarithmic function. This indicates that the voltage deviation is regulated by the magnitude of temperature anomalies. If the temperature anomaly is severe, it indicates that the system is overheated and unstable, and this parameter is suppressed. The absolute value ensures that the direction of the disturbance does not affect the evaluation; only the intensity of the disturbance is considered. A logarithmic function is used to avoid numerical explosion and to preserve the nonlinear response characteristics. It is used to amplify the overall lag when the load growth rate changes drastically.

[0027] S2 also includes S22; S22, according to the statistical method, all voltage hysteresis feedback indexes dyx in the last three months are sorted from small to large, and the 85% quantile value is calculated by using the quantile method to preset the load fluctuation trend threshold G, and then the voltage hysteresis feedback index dyx obtained in real time is used for power supply disturbance evaluation, and the specific evaluation scheme is as follows: When the voltage hysteresis feedback index dyx is less than or equal to the load fluctuation trend threshold G, it indicates that the current power supply is stable, and the load change is stable, and at this time the coupling response trend prediction instruction is executed; When the voltage hysteresis feedback index dyx is greater than the load fluctuation trend threshold G, it indicates that there is a load disturbance, at this time a load warning is generated, and after triggering the load prediction correction and executing the load soft start control strategy, iterative analysis and evaluation are performed through S21.

[0028] In this embodiment, based on the collected power disturbance data set, the voltage hysteresis feedback index dyx is calculated to measure the response sensitivity of the load to the power supply disturbance. Specifically, a nonlinear adjustment model is constructed by using the voltage deviation vg and the temperature abnormal amplitude tc, and the logarithmic function is used to enhance the amplification recognition ability of high dynamic load change.

[0029] The physical meaning of the formula reflects the hysteresis feedback capability of the voltage disturbance after temperature dynamic adjustment under the background of rapid load change. The voltage deviation vg reflects the instantaneous deviation degree of the actual voltage of the power supply system relative to the rated voltage, which is a direct physical representation of the power supply stability. The abnormal amplitude tc of the power distribution cabinet temperature describes the deviation of the internal temperature of the power distribution cabinet from the standard temperature, and reflects the system overheating effect caused by load thermal disturbance. The active power change rate pa is used to describe the power disturbance caused by the dynamic start, switching or sudden rise and fall of the load. The purpose of using the logarithmic function is to compress the maximum disturbance value and avoid numerical explosion, while retaining the nonlinear growth characteristic. The effective disturbance intensity of the voltage deviation under certain thermal state is reflected. The higher the temperature, the lower the effectiveness of the disturbance. The absolute value is used to eliminate the influence of the voltage disturbance direction and only focuses on the disturbance intensity. The formula represents the amplification adjustment mechanism of the hysteresis feedback effect of the system when the load grows or switches violently, and is used to increase the weight of the entire index with the increase of the active power change rate. After square processing, the response to the dramatic change of power is enhanced, and after root processing, the growth rate is gentle, avoiding being too radical. The product of the two parts forms a composite response, which means that only when the voltage disturbance is not completely suppressed under high temperature and the power change rate is high, the voltage hysteresis feedback index dyx will be significantly increased. Subsequently, the voltage hysteresis feedback index dyx in the last three months is sorted by statistical method, and the 85% quantile value is calculated by quantile method to set the load fluctuation trend threshold G. The real-time voltage hysteresis feedback index dyx is compared with the load fluctuation trend threshold G to complete the power supply disturbance evaluation and judgment. When the voltage hysteresis feedback index dyx is lower than the load fluctuation trend threshold G, it is judged that the power supply is stable, and the subsequent coupling trend prediction is automatically executed; when the voltage hysteresis feedback index dyx exceeds the load fluctuation trend threshold G, the load warning is triggered, and the prediction correction and slow start control instructions are automatically linked. Through the implementation of this step, not only the dynamic quantification and real-time evaluation of the power system disturbance response are realized, but also the accurate identification and response control ability of the system to the load fluctuation state are effectively improved. Compared with the traditional method based on fixed threshold or experience rule, the method has stronger adaptability and prediction foresight, thereby significantly improving the stability and intelligent level of the power grid operation Embodiment 4

[0030] Please refer to Figure 1 Specifically, S3 includes S31; S31, for executing a coupling response trend prediction instruction when the power supply disturbance evaluation is power supply stability; The coupling response trend prediction instruction is used to extract dynamic disturbance data in real time according to a special API interface, and transmit the dynamic disturbance data to a power data center for dimensionless processing to obtain a disturbance feature data set; The disturbance feature data set includes a price switching frequency fp, a bus frequency deviation value fb and a total harmonic distortion rate ne; The price switching frequency fp is extracted in real time from a price strategy pushing platform according to an API interface, and represents the number of current price changes, which is read from a real-time price log; The bus frequency deviation value fb is extracted in real time from a phasor measurement unit PMU according to an API interface, and represents the difference between the current main bus frequency and the standard value; The total harmonic distortion rate ne is extracted in real time from a power quality monitoring module PQM according to an API interface, and represents the degree of current power quality disturbance.

[0031] S3 also includes S32; S32, a synchronous response excitation index tbj is calculated according to the disturbance feature data set, which is used to quantify the synchronous induction effect of short-term price disturbance on load response, and the specific formula is as follows; ; In the formula, cos -1 represents the inverse cosine function, ln represents the logarithmic function, represents the non-linear reflection of the frequency of price fluctuation, which uses a non-linear angle mapping method to amplify the representation ability of the potential influence of frequent price fluctuation on the system, and fb2 represents the amplification effect of frequency disturbance on system instability, for the harmonic complexity as a suppression term, to avoid the frequency offset and price disturbance being over-amplified.

[0032] In this embodiment, when the power supply disturbance evaluation result is determined to be power supply stable, the coupling response trend prediction instruction is automatically executed, the dynamic disturbance data is extracted in real time from the electricity price strategy pushing platform, the synchronized phasor measurement unit PMU and the power quality monitoring module PQM through the special API interface, and three disturbance characteristic data of the electricity price switching frequency fp, the bus frequency deviation value fb and the total harmonic distortion rate ne are formed respectively, and are transmitted to the power data center for non-dimensional processing to obtain the disturbance characteristic data group, which is used to calculate the synchronous response excitation index tbj. The index is a nonlinear expression composed of inverse cosine function and logarithmic function, which can comprehensively measure the short-time synchronous excitation effect of frequent electricity price fluctuation on load response, and introduces the frequency disturbance amplification factor and the harmonic suppression term to enhance the recognition robustness of the model in unstable scenarios. Through this implementation, the depth perception of the cooperative action of electricity price disturbance and frequency offset on load change is realized, the limitation of traditional method of separate modeling of load prediction and electricity price analysis is effectively broken through, the response speed, accuracy and dynamic scheduling ability to sudden electricity price fluctuation in actual operation are improved, and technical support is provided for building a highly adaptive energy regulation system.

[0033] The formula has physical meaning, and is essentially used to construct a nonlinear evaluation index reflecting the "degree of load response excitation by short-time electricity price disturbance". The core idea is that under the premise of stable power supply, if there are frequent electricity price fluctuations, frequency abnormalities and harmonic disturbances, even if the load itself does not fluctuate, it may still cause "synchronous load response behavior" due to these factors. The index is a measurement function of this behavior. The inverse cosine term is used to express the nonlinear disturbance reaction of the electricity price switching frequency fp. Its essence is to convert the frequency of electricity price change from the linear domain to the angle domain, and to strengthen the superimposed induction effect of electricity price disturbance on the power system when the electricity price is frequently switched. The smaller the value is, the more frequent the electricity price is switched, and the smaller the angle value is, the more inconsistent the direction of the disturbance source is, and the more unstable the response induction is. The amplification factor fb 2 represents the disturbance amplification effect of the main bus frequency deviation. Once the frequency deviates, it means that there is power imbalance between the power supply and the load. The square processing is used to emphasize that the system deviates from the stable frequency, and the greater the load response synchronization induction effect is. The harmonic complexity suppression term is compressed by the logarithmic function to prevent the formula from producing non-physical explosion value when the harmonic is abnormal. Add +1 to prevent the logarithm from being negative infinity when the total harmonic distortion rate ne is zero, and at the same time meet the data smoothing requirement.

[0034] Embodiment 5 Please refer toFigure 1 and Figure 3 Specifically, S4 includes S41; S41, according to the real-time acquired voltage hysteresis feedback index dyx and synchronous response excitation index tbj, comprehensive calculation is performed to obtain a comprehensive cooperative prediction index xty, which is used to evaluate the coupling strengthening relationship between hysteresis feedback and synchronous excitation, and the specific formula is as follows: ; In the formula, cos -1 represents the inverse cosine function, ∈ represents a minimum constant for preventing the denominator from being zero, and the value is 0.001, represents whether there is a deviation between hysteresis response and synchronous excitation, if the difference value is large but the value itself is large, it indicates that the coupling is strong and asymmetric, represents the structural deviation between hysteresis feedback and synchronous response, represents a normalization term for preventing numerical explosion, the coupling angle relationship between the two is described by the inverse cosine function, the smaller the angle, the stronger the response consistency.

[0035] S4 also includes S42; S42, according to the statistical method, all comprehensive cooperative prediction indexes xty in the past three months are sorted from small to large, and the 95% quantile value is calculated by using the quantile method to preset the load price cooperative change risk threshold P, and then the real-time acquired comprehensive cooperative prediction index xty is used for cooperative disturbance evaluation, and the specific evaluation scheme is as follows: When the comprehensive cooperative prediction index xty is less than or equal to the load price cooperative change risk threshold P, it indicates that there is no cooperative disturbance between the load trend and the price fluctuation, at this time, normal dispatching operation and monitoring are maintained; When the comprehensive cooperative prediction index xty is greater than the load price cooperative change risk threshold P, it indicates that there is a cooperative disturbance between the load trend and the price fluctuation, at this time, coupling information is generated and transmitted to relevant personnel, prompting the relevant personnel to perform advance price adjustment simulation and control load strategy, and advance dispatching of standby power supply and starting of peak load reduction plan.

[0036] In this embodiment, the normalized difference between the voltage hysteresis feedback index dyx and the synchronous response excitation index tbj is used to calculate the comprehensive cooperative prediction index xty, and the coupling angle is described by the inverse cosine function, so as to reveal the response consistency of the two under asymmetric disturbance. Then, based on the quantile method of the sequence of the comprehensive cooperative prediction index xty in the last three months, the 95% quantile value is extracted to construct the load price cooperative change risk threshold P, which is used to judge whether there is a cooperative disturbance risk between the load and the price in real time. When the comprehensive cooperative prediction index xty exceeds the load price cooperative change risk threshold P, the coupling early warning information is automatically generated and pushed to the relevant personnel, triggering the advance price adjustment simulation, load control strategy, standby power source scheduling and peak load reduction mechanism. This embodiment effectively solves the technical problems that the complex coupling relationship between the load and the price cannot be quantified, the response is lagging and the strategy execution lacks timeliness in the traditional dispatching system, and improves the forward-looking of the control decision, the stability of the system operation and the initiative of the energy management.

[0037] The formula has a physical meaning, which is a tool for comprehensively characterizing the nonlinear coupling strength of the voltage hysteresis feedback index dyx and the synchronous response excitation index tbj of two different dimension disturbance response factors. The voltage hysteresis feedback index dyx reflects the hysteresis of the power load to the system disturbance, that is, the degree of response lag of the current load to the power supply disturbance, and embodies the response delay sensitivity; the synchronous response excitation index tbj describes the synchronous excitation effect of the short-time price disturbance on the load, that is, whether the frequent change or frequency fluctuation of the price can quickly induce the load response in the moment, and embodies the excitation sensitivity; the two reflect different coupling characteristics in the time delay response domain and the synchronous response domain, respectively; The middle molecular term represents the square difference in intensity of the two disturbance response mechanisms, and emphasizes the absolute size of the numerical difference. The denominator term represents the square root normalization of the numerical values of the two mechanisms. If both are large but the difference is also large, it reflects the asymmetric coupling, but can maintain the ability of high-intensity response under asymmetric structure coupling. This term is essentially a modeling of the response “conflict degree” under asymmetric structure coupling; This term is equivalent to the inverse cosine of the angle between the two normalized values, which is regarded as a vector angle. The smaller the angle, the closer the direction of the two, that is, the load change direction and the price change direction remain consistent, which represents the “coupling directionality”. This is a kind of nonlinear mapping relationship identification mechanism, which makes up for the defect that the traditional difference calculation cannot identify the consistency of the direction. The comprehensive cooperative prediction index xty couples the voltage hysteresis feedback index dyx and the synchronous response excitation index tbj, describes the structural deviation and directional consistency relationship between the two, and is used as an index of the coordination degree between synchronous response and hysteresis feedback, for judging whether the current system is in a high-coupling disturbance state.

[0038] Embodiment 6 Please refer to Figure 2The application discloses a power load and price collaborative prediction system based on deep learning, which comprises a data acquisition module, a power evaluation module, a coupling response analysis module and a collaborative evaluation module. The data acquisition module is used for collecting power state data in real time according to a set sensor group, and transmitting the power state data to a power data center via an API interface for data processing, so as to obtain a power disturbance data group. The power evaluation module is used for calculating a voltage lag feedback index dyx according to the power disturbance data group, and performing power supply disturbance evaluation with a preset load fluctuation trend threshold G, and executing a coupling response trend prediction instruction when the power supply disturbance evaluation is power supply stability. The coupling response analysis module is used for executing the coupling response trend prediction instruction, extracting dynamic disturbance data in real time according to a special API interface, and performing dimensionless processing through the power data center to obtain a disturbance characteristic data group, and then calculating a synchronous response excitation index tbj according to the disturbance characteristic data group. The collaborative evaluation module is used for calculating a comprehensive collaborative prediction index xty according to the voltage lag feedback index dyx and the synchronous response excitation index tbj, and performing collaborative disturbance evaluation with a preset load price collaborative change risk threshold P.

[0039] Although the embodiments of the present application have been shown and described, it is to be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for co-predicting electricity load and electricity price based on deep learning, characterized in that: Includes the following steps: S1. Collect power status data in real time based on the set sensor group, and transmit the power status data to the power data platform for data processing through the API interface to obtain the power disturbance data group. S2. Calculate the voltage hysteresis feedback index dyx based on the power disturbance data set, and evaluate the power supply disturbance with the preset load fluctuation trend threshold G. When the power supply disturbance evaluation is that the power supply is stable, execute the coupled response trend prediction command. S3. Execute the coupled response trend prediction command, extract dynamic disturbance data in real time according to the dedicated API interface, and perform dimensionless processing through the power data platform to obtain the disturbance characteristic data set. Then, calculate the synchronization response excitation index tbj based on the disturbance characteristic data set. S4. Calculate the comprehensive collaborative prediction index xty based on the voltage hysteresis feedback index dyx and the synchronization response excitation index tbj, and conduct a collaborative disturbance assessment with the preset load electricity price collaborative change risk threshold P.

2. The method for co-predicting power load and electricity price based on deep learning according to claim 1, characterized in that: S1 includes S11; S11. Collect power status data in real time based on the sensor group installed on the power distribution equipment, and mark the collection timestamp; The power status data includes voltage V, internal temperature wd of the distribution cabinet, and active power sequence P; The sensor group includes a voltage sensor, a thermistor temperature sensor, and a three-phase power meter; The voltage sensor is used to be installed at the transformer output terminal and user access node to collect the real-time voltage V in the power system; The thermal temperature sensor is installed inside the power distribution cabinet to collect the internal temperature wd of the power distribution cabinet in real time. The three-phase power meter is installed inside the meter box to monitor active power in real time and form an active power sequence P based on the data collection timestamp. The specific form of the active power sequence P is: P = {P t P t+1 ,,,,,P t+n } 3. The method for co-predicting power load and electricity price based on deep learning according to claim 1, characterized in that: S1 also includes S12; S12 is used to establish a communication connection between the sensor group and the power data platform through the API interface, and transmit the acquired power status data to the power data platform for data processing in real time. The data processing is used to perform data preprocessing on the power status data after data analysis to obtain power disturbance data sets; The data analysis includes voltage offset analysis, temperature amplitude analysis, and power change analysis; Voltage offset analysis is based on the obtained real-time voltage V and the corresponding node's rated voltage V. rated Perform interpolation to obtain the voltage offset value vg. The specific formula is: vg = |VV| rated |; Temperature amplitude analysis is based on the obtained internal temperature wd of the distribution cabinet, and then the difference is processed with the normal temperature wb under standard conditions to obtain the abnormal temperature amplitude tc of the distribution cabinet. The specific formula is: tc=|wd-wb|. Power change analysis calculates the active power change rate pa based on the obtained active power sequence P, using the following formula: In the formula, ∆t represents P t and P t+1 The time interval between; The data preprocessing is used to perform dimensionless processing, noise reduction, and outlier processing on the parameters after data analysis. The dimensionless processing uses the Max-Min method to eliminate the dimensional influence of the power state data. The noise reduction uses wavelet transform technology to decompose signals at different frequency scales to eliminate the noise influence in the power state data. The outlier processing uses the interquartile range method to detect and process outliers in the power state data. The power disturbance data set includes voltage offset value vg, distribution cabinet temperature anomaly amplitude tc, and active power change rate pa.

4. The method for co-predicting power load and electricity price based on deep learning according to claim 3, characterized in that: S2 includes S21; S21. Calculate the voltage hysteresis feedback index dyx based on the power disturbance data set. This index is used to measure the sensitivity of the load response to power supply system disturbances. The specific formula is as follows: ； In the formula, ln represents the logarithmic function.

5. The method for co-predicting power load and electricity price based on deep learning according to claim 4, characterized in that: S2 also includes S22; S22. Based on the statistical method, sort all voltage hysteresis feedback indices dyx in the past three months from smallest to largest, and use the quantile method to calculate the 85th percentile value as the preset load fluctuation trend threshold G. Then, compare it with the real-time voltage hysteresis feedback index dyx to evaluate the power supply disturbance. The specific evaluation scheme is as follows. When the voltage hysteresis feedback exponent dyx ≤ the load fluctuation trend threshold G, it indicates that the current power supply is stable and the load change is stable. At this time, the coupled response trend prediction command is executed. When the voltage hysteresis feedback index dyx is greater than the load fluctuation trend threshold G, it indicates that there is a load disturbance. At this time, a load warning is generated, and after triggering the load prediction correction and executing the load slow start control strategy, iterative analysis and evaluation are performed through S21.

6. The method for co-predicting power load and electricity price based on deep learning according to claim 1, characterized in that: S3 includes S31; S31 is used to execute the coupled response trend prediction command when the power supply disturbance assessment is that the power supply is stable. The coupled response trend prediction instruction is used to extract dynamic disturbance data in real time based on a dedicated API interface, and transmit it to the power data platform for dimensionless processing to obtain disturbance feature data sets. The disturbance characteristic data set includes the electricity price switching frequency fp, the bus frequency deviation value fb, and the total harmonic distortion rate ne; The electricity price switching frequency fp is extracted in real time based on the electricity price strategy push platform accessed through the API interface. The bus frequency deviation value fb is extracted in real time by connecting to the synchronous phasor measurement unit (PMU) via the API interface. The total harmonic distortion (THD) is extracted in real time by connecting to the power quality monitoring module (PQM) via the API interface.

7. The method for co-predicting power load and electricity price based on deep learning according to claim 6, characterized in that: S3 also includes S32; S32. Calculate the synchronization response excitation index tbj based on the disturbance characteristic data set, which is used to quantify the synchronization-induced effect of short-term electricity price disturbances on load response. The specific formula is as follows: ； In the formula, cos -1 Let ln denote the inverse cosine function, and ln denote the logarithmic function.

8. The method for co-predicting power load and electricity price based on deep learning according to claim 7, characterized in that: S4 includes S41; S41. Based on the real-time acquired voltage hysteresis feedback index dyx and synchronization response excitation index tbj, a comprehensive cooperative prediction index xty is obtained, which is used to evaluate the coupling enhancement relationship between hysteresis feedback and synchronization excitation. The specific formula is as follows: ； In the formula, cos -1 Let represent the inverse cosine function, and ∈ represent a minimal constant to prevent the denominator from being zero, with a value of 0.

001.

9. The method for co-predicting power load and electricity price based on deep learning according to claim 8, characterized in that: S4 also includes S42; S42. Based on the statistical method, sort all comprehensive collaborative forecasting indices xty in the past three months from smallest to largest, and use the quantile method to calculate the 95th percentile value to preset the load price collaborative change risk threshold P. Then, conduct a collaborative disturbance assessment with the real-time acquired comprehensive collaborative forecasting index xty. The specific assessment scheme is as follows. When the comprehensive collaborative forecast index xty ≤ the load-price collaborative change risk threshold P, it means that there is no collaborative disturbance between load trend and price fluctuation. At this time, normal dispatching and monitoring should be maintained. When the comprehensive collaborative forecast index xty > the load-price collaborative change risk threshold P, it indicates that there is a collaborative disturbance between load trend and electricity price fluctuation. At this time, coupling information is generated and transmitted to relevant personnel to remind them to implement advance price adjustment simulation and load control strategies, and to schedule backup power sources and initiate peak load reduction plans in advance.

10. A deep learning-based power load and electricity price co-forecasting system, comprising the deep learning-based power load and electricity price co-forecasting method as described in any one of claims 1-9, characterized in that: It includes a data acquisition module, a power assessment module, a coupled response analysis module, and a collaborative assessment module; The data acquisition module is used to collect power status data in real time based on the set sensor group, and transmit the power status data to the power data platform for data processing through the API interface to obtain the power disturbance data group. The power assessment module is used to calculate the voltage hysteresis feedback index dyx based on the power disturbance data set, and to perform power supply disturbance assessment with the preset load fluctuation trend threshold G. When the power supply disturbance assessment is that the power supply is stable, the module executes the coupled response trend prediction command. The coupling response analysis module is used to execute coupling response trend prediction instructions, extract dynamic disturbance data in real time according to the dedicated API interface, and perform dimensionless processing through the power data platform to obtain disturbance feature data set, and then calculate the synchronization response excitation index tbj based on the disturbance feature data set. The collaborative evaluation module is used to calculate the comprehensive collaborative prediction index xty based on the voltage hysteresis feedback index dyx and the synchronization response excitation index tbj, and to conduct collaborative disturbance evaluation with the preset load price collaborative change risk threshold P.

Images