A power system power and energy balance analysis method based on multi-level cascading multi-process panoramic time sequence operation simulation

Abstract

The application discloses a power system power and energy balance analysis method based on multi-level cascading multi-process panoramic timing operation simulation, which comprises the following steps: 1) a flexible regulation model considering the response characteristics of equipment is constructed, and the annual daily starting mode is determined by using the unit aggregation technology; 2) a timing decoupling criterion is determined, and a multi-scenario timing production simulation model for timing decoupling is established; 3) a multi-objective multi-scenario optimization model for boundary revision is established with the highest clean energy utilization rate, the lowest load shedding and the lowest operation cost as targets; the power and energy balance analysis result is obtained by using the multi-objective multi-scenario optimization model, and the annual hourly power profit and loss curve is corrected; and it is judged whether the power and energy balance analysis index meets the power and energy balance criterion. The application proposes a timing decoupling strategy of flexible equipment, and the efficiency and accuracy of the power and energy balance analysis are greatly improved by using the multi-process parallel solving technology, and the supply and demand balance analysis is iteratively carried out.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power balance analysis, and particularly relates to a power system power balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation. BACKGROUND

[0002] At present, wind power and photovoltaic power generation are the main force of new energy power generation. However, wind power and photovoltaic resources have uncertainty and intermittency, and when they are connected to the power system, the operation of the power system is challenged. The time sequence production simulation of the power system can fully consider the time sequence characteristics of wind power, photovoltaic and other new energy output, simulate each time period, and optimize the target according to the scene and the opinion of the decision maker, such as the maximum new energy consumption, the optimal economy, the minimum construction investment cost, the optimal environment and the minimum carbon emission, to achieve the optimal power balance result. The existing researches deeply study the power system operation simulation and supply and demand balance method under the condition of high proportion of new energy and large-scale access of energy storage, but most of them are limited to a single time scale, and few consider the coordination problem under multiple time scales, so they cannot effectively use various types of regulation resources to reduce the supply and demand mismatch risk. At the same time, with the increase of the type and number of units in the power system, the randomness and volatility of new energy power generation, the access of energy storage and pumped storage power station and other types of energy storage, the traditional power system production simulation calculation cannot adapt to the actual scene of large-scale new energy power supply and the access of multiple types of units, and the calculation result cannot reflect the actual operation of the power grid, and the solving efficiency is insufficient. SUMMARY

[0003] The present application aims to provide a power system power balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation, comprising the following steps:

[0004] 1) collect and analyze power system data information and initial operation mode, define power and energy balance analysis index system and iteration termination criterion; 2) generate wind, light, water and load scene set, calculate power and energy supply and demand mismatch risk under multiple scenes combined with power and energy balance analysis index system, and obtain power and energy balance analysis result; judge whether the power and energy balance analysis result meets the power and energy balance criterion, if yes, output the power and energy balance analysis result, otherwise, go to step 3); 3) build a flexible regulation model considering the response characteristics of equipment, and determine the annual daily start mode by using unit aggregation technology; 4) determine the time sequence decoupling criterion, and establish a multi-scenario time sequence production simulation model for time sequence decoupling; obtain the power and energy balance analysis result by using the flexible regulation model considering the response characteristics of equipment, and correct the annual hourly power profit and loss curve in parallel; judge whether the power and energy balance analysis index meets the power and energy balance criterion, if yes, output the power and energy balance analysis result, otherwise, go to step 5); 5) establish a multi-objective multi-scenario optimization model for boundary revision with the highest clean energy utilization rate, the lowest load shedding and operation cost as the target; obtain the power and energy balance analysis result by using the multi-objective multi-scenario optimization model, and correct the annual hourly power profit and loss curve; judge whether the power and energy balance analysis index meets the power and energy balance criterion, if yes, output the power and energy balance analysis result, otherwise, go to step 6); 6) judge whether the iteration convergence criterion is met, if yes, output the power and energy balance analysis result, if not, return to step 2).

[0005] Further, the power system data information includes new energy power, load power, unit parameter, energy storage / pumping storage parameter.

[0006] Further, the step of generating wind, light, water and load scene set comprises:

[0007] 1) the K-means clustering technology based on RV coefficient is adopted to divide the wind, light, water and nuclear power historical data set X into K classes. Let the set X = {X1, X2, …, X N} represent the daily output sample set of N p wind, light, water and nuclear power stations and load, wherein X i is the output matrix of the i-th day, and N is the sample capacity;

[0008] The wind, light, water, nuclear and load historical data set is divided into wind, light, water, nuclear and load daily output matrix by taking matrix as clustering object;

[0009] The feature difference of different wind, light, water, nuclear and load daily output matrix is quantified by improving the RV coefficient, that is:

[0010]

[0011] In the formula: RV(X i ,Xj ) represents the sample X i and the RV coefficient of sample X j ; tr(·) is the trace of a matrix; D diag (·) is a diagonal matrix; i, j = 1, 2, …, N; is an intermediate variable; N is the number of samples; X' i is the transpose of X i ;

[0012] 2) According to the typical daily power state obtained by clustering, the Markov state transition probability matrix Pr and the Markov cumulative state transition probability matrix Pcum are calculated; then, in chronological order, the power state of 365 days is randomly extracted based on the Markov chain Monte Carlo algorithm to form a wind-solar-hydro-nuclear power generation power state transition process of a year, and a set T intra containing Ns wind-solar-hydro-nuclear power generation power state transition processes of a year is obtained.

[0013] The Markov state transition probability matrix P r is as follows:

[0014]

[0015] In the formula, p kl represents the probability that the wind-solar-hydro-nuclear daily power state is transferred from state k to state l; k, l = 1, 2, …, K; K is the total number of states;

[0016] The maximum likelihood estimation of the probability p kl is as follows:

[0017]

[0018] In the formula, n kl is the number of days in the historical data that is transferred from state k to state l;

[0019] The Markov cumulative state transition probability matrix Pcum is as follows:

[0020]

[0021] In the formula, m represents the state;

[0022] 3) Taking the daily state in each annual power generation power state transition process in the set T intra as a label, a wind-solar-hydro-nuclear multi-site daily power generation power curve of the corresponding state is generated based on the SGAN network driven by Gaussian white noise.

[0023] ​4) Connect the daily power generation curves according to the order determined by the annual power generation state transition process of wind, solar, hydro, thermal and nuclear power to obtain the annual power generation scenario set S of wind, solar, hydro, thermal and nuclear power.

[0024] Furthermore, a scaled dot product attention mechanism is introduced into the generator and discriminator of the SGAN network;

[0025] The scaled dot product attention mechanism refers to multiplying the matching weight α with the input matrix to obtain the output matrix A(x) of the attention mechanism. a ), Distinguishing matrix x a The power generation information of different energy plants in A(x) a The degree of contribution of data differentiation is used to extract the spatial correlation of power generation from multiple power plants;

[0026] Output matrix A(x) a As shown below:

[0027] A(x a )=αx a (6)

[0028]

[0029] In the formula, x a The input matrix represents the attention mechanism, W is the learnable projection matrix, and d w Let W be the dimension of the square matrix; the softmax(·) function is used to normalize the weights.

[0030] The temporal feature extraction unit of the SGAN network includes stacked residual modules, and the output of the residual modules is obtained by fusing historical power and convolution operation information.

[0031] In the SGAN network, the scene power value y at time t t As shown below:

[0032] y t =g causal (x0,x1,...,x t ),t=0,1,...,T (8)

[0033] In the formula: x t g is the input power at time t; causal (·) represents the causal convolution operation; T represents the total number of time segments.

[0034] The dilated convolution operation of the SGAN network is shown below:

[0035]

[0036] where DC(x) is the result of the dilated convolution operation of the filter on the elements in the historical power vector x, δ is the dilated convolution operator, f(i f ) represents the i f th filter, δ is the dilation rate, and k is the filter size; f is the filter; is the dilated value of the i f th filter on the cth sequence point of the historical power vector x;

[0037] The loss functions of the generator and the discriminator in the SGAN network are as follows, respectively:

[0038] L G = -E S [D(S|C)] (10)

[0039]

[0040] where E[·] represents the expected value of the corresponding random variable; D(·) is the discriminator function; L G is the generator loss function; L D is the discriminator loss function; C represents the daily generation state label; S represents the generation power scenario; P obs is the historical observed power fused with the state label; E S [·] represents the expected value of the historical observed power P obs and the generation power scenario S;

[0041] The training target of the SGAN network is as follows:

[0042]

[0043] where C represents the daily generation state label; S represents the generation power scenario.

[0044] Further, the power and electricity balance analysis index system includes maximum power gap expectation, gap period percentage expectation, and cumulative electricity gap percentage expectation.

[0045] The steps of constructing the power and electricity balance analysis index system include:

[0046] 1) The power supply and demand balance of the power system is analyzed to obtain:

[0047]

[0048] where N the,t is the number of thermal power units at time t; N wat,t is the number of hydropower units at time t; and N nuc ​N is the number of nuclear power units; N buy N is the number of nuclear power units; N sale N is the number of nuclear power units; N wind,t N is the number of nuclear power units; N pv,t N is the number of nuclear power units; N dwat,t N is the number of nuclear power units; N buy,n,t N is the number of nuclear power units; N sale,n,t N is the number of nuclear power units; N the,n,t N is the number of nuclear power units; N wat,n,t N is the number of nuclear power units; N dwat,n,t N is the number of nuclear power units; N wind,n,t N is the number of nuclear power units; N pv,n,t N is the number of nuclear power units; N nuc,n,t N is the number of nuclear power units; N x,t N is the number of nuclear power units; N gas.t N is the number of nuclear power units; N t N is the number of nuclear power units; N res,t N is the number of nuclear power units; N load,t N is the number of nuclear power units; N

[0049] 2) Establish an electricity and energy balance evaluation index;

[0050] N is the number of nuclear power units; N max As shown below:

[0051]

[0052] In the formula, π s N is the number of nuclear power units; N t s N is the number of nuclear power units; N

[0053] N is the number of nuclear power units; N N As shown below:

[0054]

[0055] In the formula, N is the number of nuclear power units; N

[0056] N is the number of nuclear power units; N Q As shown below:

[0057]

[0058] wherein, denotes the load at time t under scenario s.

[0059] Further, the power energy balance criterion is shown as follows:

[0060]

[0061] wherein, denotes the maximum power gap threshold; denotes the gap period percentage threshold; denotes the cumulative power gap percentage threshold; Q denotes the cumulative energy gap percentage expectation; N denotes the gap period percentage expectation; max denotes the maximum power gap expectation;

[0062] The iteration convergence criterion is shown as follows:

[0063]

[0064] wherein, i denotes the number of iterations, denotes the iteration termination condition; denotes the maximum power gap expectation of the i-th and i-1-th iteration; denotes the gap period percentage expectation of the i-th and i-1-th iteration; denotes the cumulative power gap percentage expectation of the i-th and i-1-th iteration.

[0065] Further, the objective function of the flexible regulation model minf considering the equipment response characteristics is shown as follows:

[0066]

[0067] wherein is the wind curtailment penalty of scenario s; is the light curtailment penalty of scenario s; is the load shedding penalty of scenario s; T is the total time length; is the predicted power of the n-th wind turbine at time t under scenario s; c cur_wind,n is the wind curtailment penalty of the n-th wind turbine; is the predicted power of the n-th photovoltaic turbine at time t under scenario s; c cur_pv,n is the light curtailment penalty of the n-th photovoltaic turbine; c cur_load is the load shedding penalty; is the power gap at time t under scenario s; is the output power of the wind turbine at time t under scenario s; Pn,s(t) represents the output power of the nth photovoltaic generator at time t; the subsequent superscript s is omitted;

[0068] The constraint conditions of the flexible regulation model considering the equipment response characteristics include power balance constraints, aggregated thermal power generator output constraints, hydropower generator capacity constraints, small hydropower generator capacity constraints under local regulation, wind power capacity constraints, photovoltaic capacity constraints, and reserve constraints.

[0069] The power balance constraints are as follows:

[0070]

[0071] In the formula, N the_j,t represents the total number of thermal power generator groups at time t; P the,j,t represents the output of the jth thermal power group at time t. N wat,t represents the number of hydropower generators at time t; N nuc represents the number of nuclear power generators; N buy , N sale represents the number of sales and receiving areas; N wind,t represents the number of wind power generators at time t; N pv,t represents the number of photovoltaic generators at time t; N dwat,t represents the number of small hydropower generators under local regulation at time t; N x represents the number of biomass power generators; P x,n,t represents the power of the nth biomass power generator at time t;

[0072] The aggregated thermal power generator output constraints are as follows:

[0073]

[0074] In the formula, s the,j,t represents the number of thermal power generators in the jth group that are started at time t; represents the minimum output of a single generator in the jth thermal power group at time t; P the_G,j,t represents the capacity of a single generator in the jth thermal power group; P the_lim,j,t represents the limited output of a single generator in the jth thermal power group at time t; n j represents the number of generators in the jth thermal power group; P the_m,j,t represents the maintenance capacity of the jth thermal power group at time t; respectively represent the minimum and maximum values of the number of thermal power generators in the jth group that are started; P the,j,t represents the output power of the jth thermal power group at time t;

[0075] The hydropower generator capacity constraints are as follows:

[0076]

[0077] In the formula, is the maximum output of the nth water turbine unit at time t; is the minimum output of the nth water turbine unit at time t; wat_G,n,t is the capacity of the nth water turbine unit; wat_lim,n,t is the limited output of the nth water turbine unit at time t; wat_m,n,t is the maintenance capacity of the nth water turbine unit at time t.

[0078] The capacity constraint of the small water turbine unit of the ground station is as follows:

[0079] 0≤P dwat,n,t ≤P pre_dwat,n,t (24)

[0080] The capacity constraint of the wind power is as follows:

[0081] 0≤P wind,n,t ≤P pre_wind,n,t (25)

[0082] The capacity constraint of the photovoltaic is as follows:

[0083] 0≤P pv,n,t ≤P pre_pv,n,t (26)

[0084] The standby constraint is as follows:

[0085]

[0086] In the formula, s g,t is the start-stop state of the nth unit at time t, is the maximum and minimum power generation of the unit at time t, P res,u , P res,d are the upper and lower standby capacities; s the,t is the number of units started at time t; P g,t is the power generation of the unit.

[0087] Further, the step of determining the time sequence decoupling criterion comprises:

[0088] 1) record the time sequence wherein the time satisfies the following conditions:

[0089]

[0090] In the formula, k=1, 2,...N z ; N z +1 is the number of optimization periods; is the power gap at time time ; and

[0091] 2) Divide the whole year into N z +1 optimization periods with the elements in time series T z as the demarcation point.

[0092] 3) Establish a time sequence decoupling criterion. If the time sequence decoupling criterion is not satisfied, the optimization period is merged with the previous optimization period; if it is satisfied, it is not merged.

[0093] The time sequence decoupling criterion is as follows:

[0094]

[0095] In the formula, t0 is the optimization period duration criterion; S z is the optimization period imbalance time criterion; Δt is the time granularity; P t (τ) is the power gap; T z,k-1 , T z,k are the start and end times of the optimization period.

[0096] If , take the following in the optimization period:

[0097]

[0098] In the formula, is the energy stored by the electric energy storage; is the energy stored by the nth pumped storage; E B,max , W n,max is the maximum electric quantity of the nth energy storage device and the maximum energy capacity of the nth reservoir.

[0099] If , take the following in the optimization period:

[0100]

[0101] In the formula E B,n,min is the minimum electric quantity of the nth energy storage device; is the minimum energy capacity of the nth reservoir.

[0102] Further, the objective function minf of the multi-scenario time sequence production simulation model is as follows:

[0103] minf = C cur_wind + C cur_pv + C cur_load (32)

[0104] In the formula, C cur_wind , C cur_pv , C cur_load are the wind curtailment penalty, light curtailment penalty, and load shedding penalty.

[0105] The constraint conditions of the multi-scenario time sequence production simulation model include power balance constraint, energy storage capacity constraint, energy storage charging and discharging power balance constraint, pumped storage capacity constraint, pumped storage unit constraint, pumped storage unit climbing constraint, thermal power unit capacity constraint, thermal power unit climbing constraint, hydropower unit capacity constraint, hydropower unit climbing constraint, local dispatching small hydropower unit capacity constraint, wind power capacity constraint, photovoltaic capacity constraint, and backup capacity constraint;

[0106] The power balance constraint is as follows:

[0107]

[0108] In the formula, P sto_cha,n,t is the charging power of the nth energy storage device at time t; P sto_dis,n,t is the discharging power of the nth energy storage device at time t; P pump_dis,n,t is the pumped storage power of the nth reservoir at time t; and P pump_cha,n,t is the power generated by the nth reservoir at time t.

[0109] The energy storage capacity constraint is as follows:

[0110]

[0111] In the formula, γ B is the self-loss coefficient of the nth electric energy storage; η B,cha,n and η B,dis,n are the charging and discharging efficiencies of the nth electric energy storage, respectively; P cha,n.t and P dis,n,t are the charging and discharging powers of the nth electric energy storage at time t, respectively.

[0112] The energy storage charging and discharging power balance constraint is as follows:

[0113]

[0114] In the formula, P cha,n,max is the maximum charging power of the nth energy storage device; P dis,n,min is the maximum discharging power of the nth energy storage device; s st_cha,t and s st_dis,t are the charging and discharging states of the electric energy storage at time t, which are 0, 1 variables, 1 when running, and 0 otherwise. N sto is the energy storage charging and discharging state threshold of all energy storage devices.

[0115] The pumped storage capacity constraint is as follows:

[0116]

[0117] In the formula, η pump_cha,n and η pump_dis,n are the pumping and discharging efficiencies of the nth pumped storage, respectively.pump_cha,n,t P pump_dis,n,t These represent the pumping and releasing power of the nth pumped storage unit at time t.

[0118] The constraints of the pumped-storage unit are as follows:

[0119]

[0120] In the formula, P pump_d,n,min and P pump_d,max P represents the minimum and maximum power output limits of the nth pumped storage unit, respectively; pump_u,n,min and P pump_u,n,max s represents the minimum and maximum pumping output limits of the nth pumped storage unit, respectively. pump_d,n,t and s pump_u,n,t These represent the power generation and pumping states of the nth pumped-storage unit, respectively.

[0121] The ramp-up constraints for pumped-storage units are as follows:

[0122]

[0123] In the formula ur pump_d,n / dr pump_d,n This represents the ramp-up / ramp-down constraint when the nth pumped-storage unit is generating electricity; ur pump_u,n / dr pump_u,n This represents the up / down ramp constraint when the nth pumped storage unit is pumping water.

[0124] The capacity constraints of thermal power units are as follows:

[0125]

[0126] The ramping constraints for thermal power units are as follows:

[0127]

[0128] In the formula ur the,j For the climbing constraints of thermal power units in group j; dr the,j For the downhill ramp constraint of thermal power units in group j;

[0129] The capacity constraints of the hydropower units are as follows:

[0130]

[0131] The gradient constraints for hydropower units are as follows:

[0132]

[0133] In the formula ur wat,n For the uphill slope constraint of the nth hydropower unit; dr wat,nThe lower ramping constraint of the nth hydroelectric generating unit.

[0134] The capacity constraint of the small hydroelectric generating unit of the local dispatching center is shown as follows:

[0135] 0≤P dwat,n,t ≤P pre_dwat,n,t (42)

[0136] The wind power capacity constraint is shown as follows:

[0137] 0≤P wind,n,t ≤P pre_wind,n,t (43)

[0138] The photovoltaic capacity constraint is shown as follows:

[0139] 0≤P pv,n,t ≤P pre_pv,n,t (44)

[0140] The reserve capacity constraint is shown as follows:

[0141]

[0142] Further, the objective function of the multi-objective multi-scenario optimization model is shown as follows:

[0143] minf=C cur_wind +C cur_pv +C cur_load (46)

[0144] The constraint conditions of the multi-objective multi-scenario optimization model include the power balance constraint, the thermal power generating unit capacity constraint, the hydroelectric maintenance duration and continuity constraint, the thermal power maintenance time window constraint, the thermal power generating unit capacity constraint, the hydroelectric maintenance duration and continuity constraint, the hydroelectric maintenance time window constraint, the reservoir dispatching constraint, the small hydroelectric generating unit capacity constraint of the local dispatching center, the wind power capacity constraint, the photovoltaic capacity constraint, and the reserve capacity constraint.

[0145] The power balance constraint is shown as follows:

[0146] The thermal power generating unit capacity constraint is shown as follows:

[0147]

[0148] In the formula, u the,n,t is the maintenance state of the jth thermal power generating unit group at time t, 0 indicates that it is in the maintenance state, and otherwise indicates that it is not in the maintenance state.

[0149] The hydroelectric maintenance duration and continuity constraint is shown as follows:

[0150]

[0151] where D the,j is the jth thermal power unit group maintenance duration.

[0152] The thermal power maintenance time window constraint is shown as follows:

[0153]

[0154] where, is the earliest maintenance time of the jth thermal power unit group; is the slowest maintenance time of the jth thermal power unit group.

[0155] The thermal power unit capacity constraint is shown as follows:

[0156]

[0157] where u wat,n,t is the maintenance state of the nth hydropower unit at time t, 0 indicates being in maintenance state, and 1 indicates not being in maintenance state; λ wat,n is the power generation efficiency of the nth hydropower unit; Q wat,n,t is the power generation water volume of the nth reservoir at time t; h wat,n,t is the water head of the nth reservoir at time t. s wat,n,t is the hydropower unit start state;

[0158] The hydropower maintenance duration and continuity constraint is shown as follows:

[0159]

[0160] where D wat,n is the maintenance duration of the nth hydropower unit.

[0161] The hydropower maintenance time window constraint is shown as follows:

[0162]

[0163] where is the earliest maintenance time of the nth hydropower unit; is the slowest maintenance time of the nth hydropower unit.

[0164] The reservoir dispatching constraint is shown as follows:

[0165]

[0166] where Q cut,n,t is the water abandonment volume of the nth reservoir at time t; f n,t is the natural inflow volume of the nth reservoir at time t; V wat,n,t is the reservoir capacity of the nth reservoir at time t; is the maximum power generation water flow of the nth reservoir; is the maximum abandoned water volume of the nth reservoir; Q cut,n-1,t is the abandoned water volume of the upstream reservoir at time t of the nth reservoir; Q wat,n-1,t is the power generation flow of the upstream reservoir at time t of the nth reservoir.

[0167] The capacity constraint of the ground regulation small hydropower unit is as follows:

[0168] 0≤P dwat,n,t ≤P pre_dwat,n,t (54)

[0169] The capacity constraint of the wind power is as follows:

[0170] 0≤P wind,n,t ≤P pre_wind,n,t (55)

[0171] The capacity constraint of the photovoltaic is as follows:

[0172] 0≤P pv,n,t ≤P pre_pv,n,t (56)

[0173] The standby capacity constraint is as follows:

[0174]

[0175] The technical effect of the present application is self-evident. The present application proposes a multi-stage cascade, multi-process hour-day-week-year panoramic time sequence simulation method. A three-stage power and energy balance analysis idea of multi-stage cascade multi-process multi-scenario checking and time sequence operation simulation iteration is proposed. The equipment with different response characteristics is regulated in stages in a "cascade" progressive manner. A time sequence decoupling strategy of flexible equipment is proposed. The "multi-process" parallel solving technology greatly improves the efficiency and accuracy of power and energy balance analysis, and the supply and demand balance analysis is iteratively carried out. BRIEF DESCRIPTION OF DRAWINGS

[0176] Figure 1 is a schematic diagram of a power system power and energy balance analysis method based on multi-stage cascade multi-process panoramic time sequence operation simulation; Figure 2 is a solving process of a power system power and energy balance analysis method based on multi-stage cascade multi-process panoramic time sequence operation simulation; Figure 3 is a medium and long-term typical scene supply and demand balance; Figure 4 is a medium and long-term typical scene energy storage / pumping storage flexible regulation analysis; Figure 5 is a medium and long-term typical scene repair / water reservoir scheduling plan revision analysis; Figure 6 is a power surplus and deficit curve of a certain regional power grid in 2021. DETAILED DESCRIPTION

[0177] The application will be further described in connection with the following examples, which should not be construed as limiting the above-mentioned subject matter of the application to the examples described below. Various replacements and modifications can be made according to ordinary technical knowledge and conventional means in the art without departing from the above-mentioned technical idea of the application, and all of them should be included in the protection scope of the application.

[0178] Example 1

[0179] Referring to Figures 1 to 5 A power system power and energy balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation, comprising the following steps:

[0180] 1) Collect and analyze power system data information and initial operation mode, define power and energy balance analysis index system and iteration termination criterion; 2) generate wind, light, water and load scene set, calculate power and energy supply and demand mismatch risk under multiple scenes combined with power and energy balance analysis index system, obtain power and energy balance analysis result; judge whether the power and energy balance analysis result meets the power and energy balance criterion, if yes, output the power and energy balance analysis result, otherwise, go to step 3); 3) build a flexible regulation and control model considering the response characteristics of equipment, and determine the annual daily start mode by using unit aggregation technology; 4) determine the time sequence decoupling criterion, and establish a multi-scenario time sequence production simulation model for time sequence decoupling; obtain the power and energy balance analysis result by using the flexible regulation and control model considering the response characteristics of equipment, and correct the annual hourly power profit and loss curve in parallel; judge whether the power and energy balance analysis index meets the power and energy balance criterion, if yes, output the power and energy balance analysis result, otherwise, go to step 5); 5) establish a boundary revised multi-objective multi-scenario optimization model with the highest clean energy utilization rate and the lowest load shedding and operation cost as the target; obtain the power and energy balance analysis result by using the multi-objective multi-scenario optimization model, and correct the annual hourly power profit and loss curve; judge whether the power and energy balance analysis index meets the power and energy balance criterion, if yes, output the power and energy balance analysis result, otherwise, go to step 6); 6) judge whether the iteration convergence criterion is met, if yes, output the power and energy balance analysis result, if not, return to step 2).

[0181] Example 2

[0182] A power system power and energy balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation, the technical content is the same as that of example 1, further, the power system data information includes new energy power, load power, unit parameter, energy storage / pumping storage parameter.

[0183] Example 3

[0184] The technical content is the same as any one of embodiments 1-2, further, the step of generating the wind, light, water, and load scene set comprises:

[0185] 1) The K-means clustering technology based on the RV coefficient is adopted to divide the wind, light, water, and nuclear power generation power history data set X into K classes.

[0186] Let the set X = {X1, X2, …, X N} represent the daily output sample set of N p wind, light, water, and nuclear power stations and loads, wherein X i is the output matrix of the i-th day, and N is the sample capacity;

[0187] The wind, light, water, and nuclear load history data set is divided by taking the matrix as the clustering object to obtain the wind, light, water, and nuclear load daily output matrix;

[0188] The characteristic difference of different wind, light, water, and nuclear load daily output matrices is quantified by improving the RV coefficient, that is:

[0189]

[0190] In the formula: RV(X i ,X j ) represents the RV coefficient of sample X i and sample X j ; tr(·) is the trace of the matrix; D diag (·) is a diagonal matrix; i, j1, 2, …, N; is an intermediate variable; N is the sample number; X′ i is the transpose of X i ;

[0191] 2) According to the typical daily power state obtained by clustering, the Markov state transition probability matrix Pr and the Markov cumulative state transition probability matrix Pcum are calculated;

[0192] Then, according to the time sequence, based on the Markov chain Monte Carlo algorithm, the daily power generation power state is randomly extracted to form an annual wind, light, water, and nuclear power generation power state transition process, and a set T intra containing Ns annual wind, light, water, and nuclear power generation power state transition processes is obtained.

[0193] The Markov state transition probability matrix P r is as follows:

[0194]

[0195] In the formula, p klProbability of state transition from state k to state l; k, l = 1, 2, …, K; K is the total number of states;

[0196] Probability p kl The maximum likelihood estimation of p is shown as follows:

[0197]

[0198] In the formula, n kl is the number of days in the historical data that transition from state k to state l;

[0199] The Markov cumulative state transition probability matrix Pcum is shown as follows:

[0200]

[0201] In the formula m represents the state;

[0202] 3) Take the daily state in the annual power generation state transition process of each set T intra as a label, drive it with Gaussian white noise, and generate the wind, light, water, nuclear, and day power curve of the corresponding state based on the SGAN network;

[0203] 4) According to the order determined by the wind, light, water, nuclear annual power generation state transition process, connect the daily power generation curve to obtain the annual wind, light, water, nuclear power generation scenario set S.

[0204] Embodiment 4:

[0205] A power system power and energy balance analysis method based on multi-level cascading multi-process panoramic time sequence operation simulation, the technical content is the same as any one of embodiments 1-3, further, the generator and discriminator of the SGAN network introduce a scaling dot product attention mechanism;

[0206] The scaling dot product attention mechanism refers to multiplying the matching degree weight α and the input matrix to obtain the output matrix A(x a ) of the attention mechanism, which distinguishes the contribution degree of the power generation information of different energy field stations in the matrix x a to the differentiation of A(x a ) data, thereby realizing the extraction of the spatial correlation of multi-field station power generation;

[0207] The output matrix A(x a ) is shown as follows:

[0208] A(x a ) = αx a (6)

[0209]

[0210] where x a represents the input matrix of attention mechanism, W is a learnable projection matrix, d w is the dimension of matrix W; the softmax(·) function is used to normalize the weights.

[0211] The time sequence feature extraction unit of the SGAN network includes stacked residual modules, and the output of the residual modules is fused with historical power and convolution operation information.

[0212] In the SGAN network, the scene power value y t at time t is as follows:

[0213] y t = g causal (x0,x1,...,x t ),t = 0,1,...,T (8)

[0214] where x t is the input power at time t; g causal (·) is a causal convolution operation; and T is the total number of time sections.

[0215] The dilated convolution operation of the SGAN network is as follows:

[0216]

[0217] where DC(x) is the result of dilated convolution operation of the filter on the elements in the historical power vector x, δ is a dilated convolution operator, f(i f ) represents the i f th filter, δ is the dilated rate, and k is the filter size; f is the filter. is the dilated value of the i f th filter on the cth sequence point of the historical power vector x;

[0218] In the SGAN network, the loss functions of the generator and the discriminator are as follows:

[0219] L G = -E S [D(S|C)] (10)

[0220]

[0221] where E[·] represents the expected value of the corresponding random variable; D(·) is the discriminator function; L G is the generator loss function; and L D is the discriminator loss function; C represents the daily power generation state label; and S represents the power generation power scene.obs a history observation power fused with the state label; E S [·] represents a history observation power P obs , an expected value corresponding to a power generation power scenario S;

[0222] a training target of the SGAN network as follows:

[0223]

[0224] In the formula, C represents a daily power generation state label; and S represents a power generation power scenario.

[0225] Embodiment 5:

[0226] A power system power and energy balance analysis method based on multi-cascaded multi-process panoramic time sequence operation simulation, the technical content is the same as any one of embodiments 1-4, further, the power and energy balance analysis index system includes maximum power gap expectation, gap time period percentage expectation, and cumulative energy gap percentage expectation;

[0227] The step of constructing the power and energy balance analysis index system includes:

[0228] 1) power system supply and demand balance is analyzed to obtain:

[0229]

[0230] Wherein, N the,t is the number of thermal power units at time t; N wat,t is the number of hydropower units at time t; N nuc is the number of nuclear power units; N buy , N sale is the number of selling and receiving regions; N wind,t is the number of wind power units at time t; N pv,t is the number of photovoltaic units at time t; N dwat,t is the number of small hydropower units in the local area at time t; P buy,n,t is the power purchased from the nth region outside at time t; P sale,n,t is the power sent to the nth region outside at time t; P the,n,t is the output power of the nth thermal power unit at time t; P wat,n,t is the output power of the nth hydropower unit at time t; P dwat,n,t is the output power of the nth small hydropower unit in the local area at time t; P wind,n,t is the output power of the wind power unit at time t; P pv,n,t is the output power of the nth photovoltaic unit at time t; P nuc,n,t is the output power of the nth nuclear power unit at time t; Px,t Pb(t) is the output power of biomass at time t; P gas.t Png(t) is the output power of natural gas units at time t; P t Pgap(t) is the power gap at time t; P res,t Pspare(t) is the spare capacity at time t; P load,t Pload(t) is the load at time t;

[0231] 2) Establish the power and energy balance evaluation index;

[0232] wherein, the maximum power gap expectation P max As shown below:

[0233]

[0234] In the formula, π s P(s) is the probability of scenario s; P t s Pgap(s, t) is the power gap at time t under scenario s;

[0235] The gap period percentage expectation P N As shown below:

[0236]

[0237] In the formula, Pgap(s, t) is the probability of scenario s under the existence of power gap, and T is the total time.

[0238] The cumulative energy gap percentage expectation P Q As shown below:

[0239]

[0240] In the formula, Pload(s, t) is the load at time t under scenario s.

[0241] Example 6:

[0242] A power system power and energy balance analysis method based on multi-level multi-process panoramic time sequence operation simulation, the technical content is the same as any one of examples 1-5, further, the power and energy balance criterion is as shown below:

[0243]

[0244] In the formula, Pmaxgap is the maximum power gap threshold; Pgapper is the gap period percentage threshold; Pcumgap is the cumulative power gap percentage threshold; P Q Pcumgap is the cumulative power gap percentage threshold; P N Pgapper is the gap period percentage threshold; P maxrepresents the maximum power gap expectation;

[0245] The iteration convergence criterion is shown as follows:

[0246]

[0247] In the formula, i represents the number of iterations, represents the iteration termination condition; represents the maximum power gap expectation of the i-th and i-1-th iterations; represents the gap period percentage expectation of the i-th and i-1-th iterations; represents the cumulative power gap percentage expectation of the i-th and i-1-th iterations.

[0248] Embodiment 7:

[0249] A power system power and energy balance analysis method based on multi-cascade multi-process panoramic timing operation simulation, the technical content is the same as any one of embodiments 1-6, further, the objective function of the flexible regulation model minf considering the equipment response characteristic is shown as follows:

[0250]

[0251]

[0252] In the formula is the wind curtailment penalty of the s scene; is the light curtailment penalty of the s scene; is the load shedding penalty of the s scene; T is the total time length; is the predicted power of the n-th wind turbine at the t time of the s scene; c cur_wind,n is the wind curtailment penalty of the n-th wind turbine; is the predicted power of the n-th photovoltaic turbine at the t time of the s scene; c cur_pv,n is the light curtailment penalty of the n-th photovoltaic turbine; c cur_load is the load shedding penalty; P t s is the power gap at the t time of the s scene; is the output power of the wind turbine at the t time of the s scene; is the output power of the n-th photovoltaic turbine at the t time of the s scene; the subsequent superscript s is omitted;

[0253] The constraint condition of the flexible regulation model considering the equipment response characteristic includes a power balance constraint, an aggregated thermal power turbine output constraint, a hydropower turbine capacity constraint, a local small hydropower turbine capacity constraint, a wind power capacity constraint, a photovoltaic capacity constraint, and a reserve constraint;

[0254] The power balance constraint is shown as follows:

[0255]

[0256] N the_j,t N the,j,t P wat,t N nuc N buy N sale N wind,t N pv,t N dwat,t N x P x,n,t P

[0257] The aggregate thermal power unit output constraint is shown as follows:

[0258]

[0259] s the,j,t N P the_G,j,t P the_lim,j,t P j N the_m,j,t N N the,j,t P

[0260] The hydroelectric unit capacity constraint is shown as follows:

[0261]

[0262] n P P wat_G,n,t P wat_lim,n,t P wat_m,n,t N

[0263] The small local hydroelectric unit capacity constraint is shown as follows:

[0264] 0≤P dwat,n,t ≤P pre_dwat,n,t (24)

[0265] The wind power capacity constraint is shown as follows:

[0266] 0≤P wind,n,t ≤P pre_wind,n,t (25)

[0267] The photovoltaic capacity constraint is shown as follows:

[0268] 0≤P pv,n,t ≤P pre_pv,n,t (26)

[0269] The reserve constraint is shown as follows:

[0270]

[0271] In the formula, s g,t is the start-stop state of the nth unit at time t, is the maximum and minimum power generation of the unit at time t, P res,u , P res,d are the upper and lower reserve capacities; s the,t is the number of units of the thermal power unit started at time t; and P g,t is the power generation of the unit.

[0272] Embodiment 8:

[0273] A power system power and energy balance analysis method based on multi-level multi-process panoramic time sequence operation simulation, the technical content is the same as any one of embodiments 1-7, further, the step of determining the time sequence decoupling criterion comprises:

[0274] 1) record the time sequence wherein the time satisfies the following conditions:

[0275]

[0276] In the formula, k=1, 2,...N z ; N z +1 is the number of optimization periods; is the power gap at time time ;

[0277] 2) divide the whole year into N z +1 optimization periods z with the elements in the time sequence T

[0278] 3) Establish a timing decoupling criterion, if the timing decoupling criterion is not satisfied, the optimization period is merged with the last optimization period; if it is satisfied, it is not merged;

[0279] The timing decoupling criterion is as follows:

[0280]

[0281] In the formula, t0 is the optimization period duration criterion; S z is the unbalanced time criterion of the optimization period; Δt is the time granularity; P t (τ) is the power gap; T z,k-1 , T z,k is the start and end time of the optimization period;

[0282] If , the optimization period is taken as:

[0283]

[0284] In the formula, is the energy stored by the energy storage; is the energy stored by the nth pumped storage; E B,max , W n,max is the maximum power of the nth energy storage device and the maximum energy capacity of the nth reservoir;

[0285] If , the optimization period is taken as:

[0286]

[0287] In the formula E B,n,min is the minimum power of the nth energy storage device; W n min is the minimum energy capacity of the nth reservoir.

[0288] Embodiment 9:

[0289] A power system power and energy balance analysis method based on multi-level multi-process panoramic timing operation simulation, the technical content is the same as any one of embodiments 1-8, further, the objective function minf of the multi-scenario timing production simulation model is as follows:

[0290] minf=C cur_wind +C cur_pv +C cur_load (32)

[0291] In the formula, C cur_wind , C cur_pv , C cur_load is the penalty for abandoning wind, the penalty for abandoning light, and the penalty for cutting load;

[0292] The constraint conditions of the multi-scenario time sequence production simulation model include power balance constraint, energy storage capacity constraint, energy storage charging and discharging power balance constraint, pumped storage capacity constraint, pumped storage unit constraint, pumped storage unit climbing constraint, thermal power unit capacity constraint, thermal power unit climbing constraint, hydropower unit capacity constraint, hydropower unit climbing constraint, local dispatching small hydropower unit capacity constraint, wind power capacity constraint, photovoltaic capacity constraint, and backup capacity constraint;

[0293] The power balance constraint is as follows:

[0294]

[0295] wherein P sto_cha,n,t is the charging power of the nth energy storage device at time t; P sto_dis,n,t is the discharging power of the nth energy storage device at time t; P pump_dis,n,t is the pumped storage power of the nth reservoir at time t; and P pump_cha,n,t is the power generated by the nth reservoir at time t.

[0296] The energy storage capacity constraint is as follows:

[0297]

[0298] wherein γ B is the self-loss coefficient of the nth electric energy storage; η B,cha,n and η B,dis,n are the charging and discharging efficiencies of the nth electric energy storage, respectively; P cha,n.t and P dis,n,t are the charging and discharging powers of the nth electric energy storage at time t, respectively.

[0299] The energy storage charging and discharging power balance constraint is as follows:

[0300]

[0301] wherein P cha,n,max is the maximum charging power of the nth energy storage device; P dis,n,min is the maximum discharging power of the nth energy storage device; s st_cha,t and s st_dis,t are the charging and discharging states of the electric energy storage at time t, which are 0, 1 variables, 1 when running, and 0 otherwise. N sto is the charging and discharging state threshold of all energy storage devices;

[0302] The pumped storage capacity constraint is as follows:

[0303]

[0304] wherein η pump_cha,n and η pump_dis,n are the pumping and discharging efficiencies of the nth pumped storage, respectively.pump_cha,n,t , P pump_dis,n,t are the nth pumped storage pumping and releasing power at time t, respectively. n,t is the reservoir storage capacity; are the upper and lower limits of the reservoir storage capacity;

[0305] The pumped storage unit constraints are shown as follows:

[0306]

[0307] In the formula, P pump_d,n,min and P pump_d,max represent the minimum and maximum power generation output limits of the nth pumped storage unit, respectively; P pump_u,n,min and P pump_u,n,max represent the minimum and maximum pumping output limits of the nth pumped storage unit, respectively. pump_d,n,t and s pump_u,n,t represent the state of power generation and pumping of the nth pumped storage unit, respectively.

[0308] The pumped storage unit ramping constraints are shown as follows:

[0309]

[0310] In the formula, ur pump_d,n / dr pump_d,n represent the upper / lower ramping constraints when the nth pumped storage unit generates power; ur pump_u,n / dr pump_u,n represent the upper / lower ramping constraints when the nth pumped storage unit pumps water.

[0311] The thermal power unit capacity constraints are shown as follows:

[0312]

[0313] The thermal power unit ramping constraints are shown as follows:

[0314]

[0315] In the formula, ur the,j is the upper ramping constraint of the jth group of thermal power units; dr the,j is the lower ramping constraint of the jth group of thermal power units.

[0316] The hydropower unit capacity constraints are shown as follows:

[0317]

[0318] The hydropower unit ramping constraints are shown as follows:

[0319]

[0320] In the formula, urwat,n is the up ramp constraint of the nth hydroelectric generating unit; dr wat,n is the down ramp constraint of the nth hydroelectric generating unit.

[0321] The capacity constraint of the small hydroelectric generating unit of the ground station is as follows:

[0322] 0≤P dwat,n,t ≤P pre_dwat,n,t (42)

[0323] The capacity constraint of the wind power is as follows:

[0324] 0≤P wind,n,t ≤P pre_wind,n,t (43)

[0325] The capacity constraint of the photovoltaic is as follows:

[0326] 0≤P pv,n,t ≤P pre_pv,n,t (44)

[0327] The reserve capacity constraint is as follows:

[0328]

[0329] Embodiment 10:

[0330] A power system power and energy balance analysis method based on multi-level multi-process panoramic time sequence operation simulation, the technical content is the same as any one of embodiments 1-9, further, the objective function of the multi-objective multi-scenario optimization model is as follows:

[0331] min f=C cur_wind +C cur_pv +C cur_load (46)

[0332] The constraint conditions of the multi-objective multi-scenario optimization model include power balance constraint, thermal power unit capacity constraint, hydroelectric maintenance duration and continuity constraint, thermal power maintenance time window constraint, thermal power unit capacity constraint, hydroelectric maintenance duration and continuity constraint, hydroelectric maintenance time window constraint, reservoir scheduling constraint, small hydroelectric generating unit capacity constraint of the ground station, wind power capacity constraint, photovoltaic capacity constraint, reserve capacity constraint;

[0333] The power balance constraint is as follows:

[0334] The capacity constraint of the thermal power unit is as follows:

[0335]

[0336] In the formula, u the,n,tis the maintenance state of the jth thermal power group at time t, 0 represents being in the maintenance state, and otherwise represents not being in the maintenance state.

[0337] The water power maintenance duration and continuity constraints are as follows:

[0338]

[0339] In the formula, D the,j is the maintenance duration of the jth thermal power group.

[0340] The thermal power maintenance time window constraint is as follows:

[0341]

[0342] In the formula, is the earliest maintenance time of the jth thermal power group; is the slowest maintenance time of the jth thermal power group.

[0343] The thermal power unit capacity constraint is as follows:

[0344]

[0345] In the formula, u wat,n,t is the maintenance state of the nth water power unit at time t, 0 represents being in the maintenance state, and 1 represents not being in the maintenance state; λ wat,n is the power generation efficiency of the nth water power unit; Q wat,n,t is the water quantity generated by the nth reservoir at time t; h wat,n,t is the water head of the nth reservoir at time t. s wat,n,t is the water power unit start state;

[0346] The water power maintenance duration and continuity constraints are as follows:

[0347]

[0348] In the formula, D wat,n is the maintenance duration of the nth water power unit.

[0349] The water power maintenance time window constraint is as follows:

[0350]

[0351] In the formula is the earliest maintenance time of the nth water power unit; is the slowest maintenance time of the nth water power unit.

[0352] The reservoir scheduling constraint is as follows:

[0353]

[0354] Q = Q + Q cut,n,t is the nth reservoir t time of the abandoned water; f n,t is the nth reservoir t time of the natural inflow; V wat,n,t is the nth reservoir t time of the reservoir capacity; is the nth reservoir of the maximum power water flow; is the nth reservoir of the maximum abandoned water; Q cut,n-1,t is the nth reservoir upstream of the reservoir t time of the abandoned water; Q wat,n-1,t is the nth reservoir upstream of the reservoir t time of the power flow.

[0355] The capacity constraints of the local small hydropower unit are as follows:

[0356] 0≤P dwat,n,t ≤P pre_dwat,n,t (54)

[0357] The wind power capacity constraint, photovoltaic capacity constraint, and standby capacity constraint are as follows:

[0358] 0≤P wind,n,t ≤P pre_wind,n,t (55)

[0359] 0≤P pv,n,t ≤P pre_pv,n,t (56)

[0360]

[0361] Example 11:

[0362] A power system power balance analysis method based on multi-level multi-process panoramic time sequence operation simulation, the steps are as follows:

[0363] Step 1, collect and analyze power system data information and initial operation mode, define power balance analysis index system and iteration termination criterion; Step 2, generate wind light water load scene set in one stage, calculate power supply and demand mismatch risk under multiple scenarios combined with power balance analysis index system, get power balance analysis result; Step 3, use unit aggregation technology to determine the annual daily start mode in the second stage; Step 4, determine the time sequence decoupling criterion, establish the multi-scenario time sequence production simulation model of time sequence decoupling, and further get the power balance analysis result; Step 5, the third stage, propose the boundary revision method of maintenance / reservoir plan to further get the power balance analysis result, judge whether the result meets the iteration termination criterion, if it meets, output the power balance analysis result, if it does not meet, return to step 2.

[0364] Example 12:

[0365] A power system power and energy balance analysis method based on multi-level cascade multi-process panoramic time sequence operation simulation, the steps are as follows:

[0366] 1) The three-stage power and energy balance analysis idea of "multi-level cascade multi-process" multi-scenario checking and time sequence operation simulation iteration is proposed, see Figure 1 . The devices with different response characteristics are regulated in stages in a "multi-level cascade" progressive manner; the time sequence decoupling strategy of flexible devices is proposed, and the "multi-process" parallel solving technology greatly improves the efficiency and accuracy of power and energy balance analysis. The uncertainty of wind, light and water is analyzed by using multi-scenario time sequence simulation, and the power and energy supply and demand balance analysis is iteratively carried out until the convergence criterion is met.

[0367] 2) Collect and analyze new energy power, load power, unit parameter, energy storage / pumping storage parameter and other data information and initially determined operation mode. Establish the power and energy balance analysis index system including maximum power gap expectation, gap period percentage expectation, cumulative power gap percentage expectation, etc. Define the multi-scenario checking and time sequence operation simulation iterative convergence criterion including power and energy balance analysis index threshold and power and energy balance analysis index convergence.

[0368] 3) In the first stage, the scene generation method based on sequential generative adversarial network (SGAN) and the daily state transition simulation method based on Markov chain are used to generate annual hourly wind, light, water and load scene sets, providing data basis for power and energy balance analysis. Calculate the power and energy supply and demand mismatch risk under multiple scenarios combined with the power and energy balance analysis index system to obtain the power and energy balance analysis result, and judge whether the convergence criterion is met. If it is met, output the power and energy balance analysis result, otherwise enter the second stage.

[0369] 4) In the second stage, a flexible regulation model considering the response characteristics of the device is proposed. Analyze the time sequence response characteristics of flexible devices including units, energy storage / pumping storage, etc., establish time sequence decoupling criterion, and explore the flexibility coordination ability among multiple devices. Use the unit aggregation technology considering unit type and electrical distance to determine the annual daily start-up mode. Establish a multi-scenario time sequence production simulation model for time sequence decoupling, further obtain the power and energy balance analysis result, and correct the annual hourly power profit and loss curve in parallel, and judge whether the power and energy balance analysis index meets the convergence criterion. If it is met, output the power and energy balance analysis result, otherwise enter the third stage.

[0370] 5) The third stage, the maintenance / reservoir plan boundary revision method is proposed. According to the maintenance requirements and reservoir inflow prediction information, a multi-objective multi-scenario optimization model for boundary revision is established, aiming at the highest utilization rate of clean energy, the lowest load and the lowest operation cost. The annual hourly power profit and loss curve is corrected, and it is judged whether the power and energy balance analysis index meets the convergence criterion. If it meets the criterion, the power and energy balance analysis result is output, otherwise the first stage is returned to recalculate the power profit and loss curve until the convergence criterion is met.

[0371] The main steps of the first stage of SGAN-based scene generation method and Markov chain day state transition simulation method to generate annual hourly wind, light, load scene set are as follows:

[0372] 1) The K-means clustering technology based on RV coefficient is used to divide the wind, light, water, and fire nuclear daily power generation historical data set X into K classes.

[0373] Let the set X = {X1, X2, …, X N} represent the daily output sample set of N p wind, light, water, and fire nuclear power stations and loads, wherein X i is the output matrix of the i-th day, and N is the sample capacity. In order to fully consider the spatial correlation characteristics of multi-station power, the wind, light, water, and fire nuclear load historical data set is divided into matrix clustering objects. The RV coefficient is an excellent statistical index for measuring matrix correlation, and the RV coefficient is introduced to quantify the feature difference of different wind, light, water, and fire nuclear load daily output matrices:

[0374]

[0375] The RV coefficient is used as a measure of the distance between the sample matrix X i and the clustering center in the K-means clustering technology, and by iterative updating of the clustering center, the matrices in the set X can be divided into K representative daily power states.

[0376] 2) According to the typical daily power state obtained by clustering, the matrix Pr and Pcum are calculated. Then, according to the time sequence, a random annual wind, light, water, and fire nuclear power state transition process is extracted based on Markov Chain Monte Carlo (MCMC) algorithm, and repeated sampling can obtain a set Tintra containing Ns annual wind, light, water, and fire nuclear power state transition processes.

[0377] According to the daily power state and its frequency obtained by clustering, the Markov state transition probability matrix P r is calculated, which is defined as follows:

[0378]

[0379] where element p kl The maximum likelihood estimation of the probability that the state of wind, light, water, and nuclear power changes from state k to state l (k, l 1, 2,..., K) is represented by state k to state l:

[0380]

[0381] According to Pr, the Markov cumulative state transition probability matrix Pcum can be further obtained:

[0382]

[0383] 3) The daily state in the annual power generation state transition process of each set T intra is taken as a label, driven by Gaussian white noise, and the corresponding state of wind, light, water, and nuclear power daily power curve is generated based on SGAN.

[0384] In view of the deficiency of GAN in annual wind and light power time sequence feature extraction, this section introduces a scaling dot product attention mechanism and TCN in the generator and discriminator based on CGAN to construct SGAN to generate daily power curve under the supervision of daily generation state label. Through the scaling dot product attention mechanism, the extraction ability of SGAN to the spatial features of multiple stations is enhanced. With the help of the time sequence information processing advantage of TCN, the time sequence features and key output events in wind, light, water and nuclear power generation are fully mined, and the simulation accuracy of daily wind, light, water and nuclear power generation is improved.

[0385] The scaling dot product attention mechanism is adopted in the application, and the calculation formula of the matching degree weight a is:

[0386]

[0387] The matching degree weight a is multiplied by the input matrix to obtain the output matrix A(x a ) of the attention mechanism, that is, the contribution degree of the power generation information of different energy stations in the matrix x a to the differentiation of A(x a ) data can be distinguished, so as to realize the extraction of the spatial correlation of the power generation of multiple stations:

[0388] A(x a )=αx a (7)

[0389] The TCN is taken as the main structure to construct the time sequence feature extraction unit of SGAN. Figure 2 The causal expansion convolution with 2 layers of hidden layers is shown in the schematic diagram, and the data before 0 time (initial time) is 0. The causal convolution makes the data between TCN layers have time sequence correlation, and the information transmission direction is as follows:Figure 2 The scene power value y t is determined only by the historical power information from 0 to t time, that is,

[0390] y t = g causal (x0, x1,..., x t ), t = 0, 1,..., T (8)

[0391] If the complete sequence information of the long-range historical power generation needs to be extracted, the number of network layers or the size of the convolution kernel will increase sharply, which will cause problems such as gradient disappearance and low efficiency in network training. The dilated convolution adds a hole in the standard convolution kernel, thereby expanding the range of the receptive field, so that the TCN can extract complete historical information without being too deep. The dilated convolution operation can be represented as:

[0392]

[0393] On this basis, the causal dilated convolution uses residual connection to form a residual module, and the output of the residual module is fused with the historical power and the convolution operation information, thereby improving the expression ability of the scene generation algorithm to the wind, light, water and fire power generation features. The stacked residual modules can form a time sequence feature extraction unit.

[0394] Based on the spatial and temporal feature extraction unit, the SGAN for generating intra-day multi-scene power sequences is shown in FIG. 2, which enhances the learning and generation ability of the network to the specified type of data through the game of two relatively independent deep neural networks (i.e., discriminator D and generator G). Figure 3 Figure 3 The green box is a residual module, each residual module is composed of 3 identical TCN modules, and the structure is shown in the left expansion diagram. The residual module is stacked N D and N G times in the discriminator and the generator respectively, to form a deep TCN network, and is numbered as l1,2,…,N D (N G ) in the direction from the input end to the far end. The input of the generator is Gaussian white noise Z and the daily power generation state label C based on One-Hot coding, and the output data is the multi-scene power generation power scene S, where d z is the noise dimension; the input of the discriminator is the historical observed power P obs fused with the state label and the generated scene S. By analyzing the characteristics of the input data, the discriminator gives the discrimination result of the authenticity of the input data.

[0395] Considering that the structures of the generator and the discriminator are relatively independent, the loss functions of the two networks are introduced:

[0396] L G = -E​S [D(S|C)] (10)

[0397]

[0398] where E[·] represents the expectation value of the corresponding random variable; D(·) is the discriminator function. According to the purpose of both generator and discriminator, the generator wants to generate more realistic scenes to make the discriminator misjudge, while the discriminator needs to distinguish the observation power and the generated scene with higher accuracy. Therefore, the training process of SGAN can be regarded as a minimax game:

[0399]

[0400] 4) According to the order determined by the wind, light, water and fire nuclear annual power generation state transition process, the daily power generation curve is connected, that is, the annual wind, light, water and fire nuclear power generation scene set S is obtained.

[0401] The first stage establishes the maximum power gap expectation, gap time period percentage expectation, cumulative power gap percentage and other power and energy balance analysis index system. Define the multi-scenario checking and time sequence operation simulation iteration convergence criterion, and the power and energy balance analysis index threshold value entering the next stage, the main steps are as follows:

[0402] 1) Establish the power and energy balance analysis index system: according to the given operation information, analyze the supply and demand balance of the power system:

[0403]

[0404] Establish three power and energy balance evaluation indexes: maximum power gap expectation, gap time period percentage expectation, and cumulative power gap percentage:

[0405]

[0406] 3) Determine the balance analysis convergence criterion

[0407] a) Power and energy balance criterion

[0408]

[0409] b) Convergence criterion

[0410] Use the relative error sum of squares of the maximum power gap expectation, gap time period percentage expectation, and cumulative power gap percentage to establish the iteration termination criterion as shown in the following formula

[0411]

[0412] When the power balance criterion or the convergence criterion is satisfied, the iteration is terminated, and the power balance analysis result is output; otherwise, the power balance analysis of the next stage is entered.

[0413] The second stage determines the daily start-up mode of the year by using the unit aggregation technology considering the unit type and electrical distance, and the main steps are as follows:

[0414] 1) The daily start-up mode takes minimizing the power gap loss and the cost of abandoned wind and light as the target, and the function is as follows:

[0415]

[0416] 2) Unit aggregation

[0417] Unit aggregation is to aggregate units of the same type in a power system, and units with the same or similar characteristic parameters are aggregated into a group. All binary variables indicating the operating state of the unit in the group can be replaced by a single integer variable, which can greatly reduce the number of integer variables in the model and reduce the search space of the variables.

[0418] 3) Consider the constraints of unit aggregation, including power balance constraints, aggregated thermal power unit output constraints, hydropower unit capacity constraints, ground control small hydropower unit capacity constraints, wind power capacity constraints, photovoltaic capacity constraints, and reserve constraints;

[0419] 3) Calculate the power gap standard.

[0420] The second stage establishes a time sequence decoupling criterion to explore the flexibility coordination capability among multiple devices. A multi-scenario time sequence production simulation model is established under the time sequence decoupling, and the annual hour-level power profit and loss curve is corrected in parallel to perform the main steps of the hour-level production simulation as follows:

[0421] 1) Determine the parallel optimization period of time sequence decoupling

[0422] Record the time sequence Where The following conditions are met:

[0423]

[0424] The sequence T z is divided into N z +1 optimization periods with the elements in the sequence as the dividing points. An optimization period merging criterion is established:

[0425]

[0426] In the formula, t0 is the optimization period duration criterion; S zTo optimize the time period imbalance time criterion; Δt is the time granularity. Each optimization time period is decoupled in time and can be optimized in parallel, improving computational efficiency.

[0427] If then take

[0428]

[0429] where T z,s is the start time of the optimization time period; E Bn,t is the energy stored in the electrical storage at time t; E B,n,min is the minimum electrical quantity of the nth energy storage device; W n,t is the energy stored in the nth pumped storage at time t; W n,t is the energy stored in the nth pumped storage at time t.

[0430] If then take

[0431]

[0432] where E B,n,min is the minimum electrical quantity of the nth energy storage device; W is the minimum energy capacity of the nth reservoir.

[0433] 2) Pumped storage / energy storage flexible regulation model

[0434] The objective of the pumped storage / energy storage flexible regulation model is to minimize the multi-cut load penalty, wind and light water load penalty cost when considering energy storage and pumped storage. The objective function is as follows:

[0435] minf = C cur_wind + C cur_pv + C cur_load (25)

[0436] The constraints of the pumped storage / energy storage flexible regulation model include power balance constraints, energy storage capacity constraints, energy storage charge and discharge power balance constraints, pumped storage capacity constraints, pumped storage unit constraints, pumped storage unit ramping constraints, thermal power unit capacity constraints, thermal power unit ramping constraints, hydropower unit capacity constraints, hydropower unit ramping constraints, small hydropower unit capacity constraints, wind power capacity constraints, photovoltaic capacity constraints, and reserve capacity constraints.

[0437] Based on the second stage corrected annual hour-level power profit and loss curve, the main steps of the maintenance / reservoir plan boundary revision method are as follows:

[0438] 1) The optimization objective is to minimize the multi-cut load penalty, wind and light water load penalty cost, and the function is as follows

[0439] minf = C cur_wind+C cur_pv +C cur_load (26)

[0440] 2) Constraints

[0441] a) Power balance constraint

[0442]

[0443] b) Thermal power unit capacity constraint

[0444]

[0445] c) Hydropower maintenance duration and continuity constraint

[0446]

[0447] d) Thermal power maintenance time window constraint

[0448]

[0449] e) Hydropower unit capacity constraint

[0450]

[0451] f) Hydropower maintenance duration and continuity constraint

[0452]

[0453] g) Hydropower maintenance time window constraint

[0454]

[0455] h) Reservoir scheduling constraint

[0456]

[0457] i) Local dispatching small hydropower unit capacity constraint

[0458] 0 ≤ P dwat,n,t ≤ P pre_dwat,n,t (34)

[0459] j) Wind power capacity constraint

[0460] 0 ≤ P wind,n,t ≤ P pre_wind,n,t (35)

[0461] k) Photovoltaic capacity constraint

[0462] 0 ≤ P pv,n,t ≤ P pre_pv,n,t (36)

[0463] l) spare capacity constraints

[0464]

[0465] Example 13:

[0466] A verification of a power system power and energy balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation, the content is as follows:

[0467] Taking the wind, light, water and load data of a certain region from 2019 to 2021 as an example, the scene generation method based on SGAN and the scene generation method of Markov chain daily state transition simulation method proposed by the application are verified, and the data are normalized.

[0468] Example 14:

[0469] An application of a power system power and energy balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation, the content is as follows:

[0470] Based on the wind, light, water and load typical scene obtained in example 13, the remaining operation boundary data refers to the initial operation mode of a certain regional power grid. Among them, considering the time sequence correlation of wind, light, water and load, a plurality of typical scenes are obtained, wherein the long-term time sequence supply and demand balance relationship is shown in Figure 3 .

[0471] The power grid operation scene 1 of a certain region in 2021 has power surplus in March to June and July to September, and has a shortage in the remaining months, the maximum shortage is 7028MW in January, and the maximum surplus is 12483MW in September; scene 2 can guarantee power surplus in February, May and July, and has a shortage in the remaining months, the maximum shortage is 8654MW in December, and the maximum surplus is 15021MW in February; scene 3 has power surplus in February, April, August and October, and has a shortage in the remaining months, the maximum shortage is 8249MW in January, and the maximum surplus is 13863MW in September.

[0472] Example 15:

[0473] An application of a power system power and energy balance analysis method based on multi-cascade multi-process panoramic time sequence operation simulation, the content is as follows:

[0474] On the basis of medium and long term time sequence supply and demand balance analysis, this embodiment analyzes the influence of energy storage / pumping storage flexible regulation and control on the power supply and demand balance of a certain regional power grid in 2021, which is shown in Figure 4 .

[0475] After the flexible regulation of energy storage / pumping storage, the power surplus of the regional power grid in scenario 1 in March to June and August to October in 2021, and the power shortage in the remaining months, the maximum shortage of 6954 MW in January, and the maximum surplus of 12483 MW in September; Scenario 2 can guarantee power surplus in February and July, and power shortage in the remaining months, the maximum shortage of 8654 MW in December, and the maximum surplus of 15021 MW in February; Scenario 3 has power surplus in February, April, August and October, and power shortage in the remaining months, the maximum shortage of 8249 MW in January, and the maximum surplus of 13863 MW in September.

[0476] Example 16

[0477] An application of a power system power and energy balance analysis method based on multi-cascading multi-process panoramic time sequence operation simulation, the content is as follows:

[0478] On the basis of medium and long term time sequence supply and demand balance analysis, this embodiment analyzes the influence of maintenance / reservoir scheduling plan revision on the power supply and demand balance of the regional power grid in 2021. According to the maintenance arrangement principle, the maintenance time window of each unit is shifted to a reasonable time period, and the reservoir water level is adjusted, and then the time sequence supply and demand balance analysis and the flexible regulation of energy storage / pumping storage are used to maximize the power supply capacity of the regional power grid. The results are shown in Figure 5 .

[0479] After the revision of the maintenance / reservoir plan, the power surplus of the regional power grid in scenario 1 in March to June and August to October in 2021, and the power shortage in the remaining months, the maximum shortage of 6954 MW in January, and the maximum surplus of 10183 MW in September; Scenario 2 can guarantee power surplus in February and July, and power shortage in the remaining months, the maximum shortage of 8654 MW in December, and the maximum surplus of 15021 MW in February; Scenario 3 has power surplus in February, April, August and October to November, and power shortage in the remaining months, the maximum shortage of 8249 MW in January, and the maximum surplus of 13863 MW in September.

[0480] Example 17

[0481] An application of a power system power and energy balance analysis method based on multi-cascading multi-process panoramic time sequence operation simulation, the content is as follows:

[0482] Because there is a certain reserve capacity in the regional power grid, the demand threshold of the cumulative energy gap percentage is set to the ratio of the reserve capacity to the load, which is 1.5%; The gap period demand is at least to meet the power load demand for ten months, that is, the threshold is set to 16.6%. When each index corresponding to the threshold is exceeded, the next adjustment stage is entered. After 16 typical scene checking and time sequence operation simulation iteration, the fine supply and demand balance relationship of the regional power grid in 2021 is obtained, as shown in Figure 6The power gap information of the operation mode of the power grid in a certain region in 2021 is shown in Table 1 in combination with the balance check index.

[0483] Table 1 Analysis table of power gap in operation mode in 2021

[0484]

[0485] In Figure 6 , stage 1 (blue curve) represents the power surplus / deficit expectation value based on multiple refined wind, light and water annual scenarios. It can be seen from Figure 6 that there is a large power gap and power gap in January and December, and the maximum power gap is 7138.12 MW, which occurs in December. Except for February and September, there is a small power gap in the remaining months, and the overall power is in a surplus state.

[0486] After the flexible regulation of energy storage / pumping storage in stage 2 (red curve), the number of gap periods and the amount of gap in the remaining months are greatly reduced except for January and December with large power gap. As shown in Table 1, the flexible regulation of energy storage / pumping storage reduces 50.60% of the power gap period and 41.56% of the power gap, which provides an effective guarantee for reducing the risk of supply and demand mismatch. However, for the scheduling period with dense gap and large gap amount (e.g. January and December), the flexible regulation effect is not obvious, because energy storage / pumping storage needs surplus power to charge, so it cannot achieve flexible regulation for scheduling periods with dense gap.

[0487] After the revision of the maintenance / reservoir plan in stage 3 (yellow curve), the power gap of the power grid in a certain region is also greatly improved, reducing 51.71% of the power gap period and 42.61% of the power gap. Compared with the flexible regulation of energy storage / pumping storage, the effect of the revision of the maintenance / reservoir plan is not significantly improved, only reducing 1.11% of the power gap period and 1.05% of the power gap. This is because the original formulated thermal power unit maintenance plan is mainly concentrated in February-April, July, September and other periods with large power surplus. According to the maintenance principle of the power grid in a certain region, thermal power maintenance is not arranged in January and December, which are the peak load periods. Therefore, the revision of the thermal power maintenance plan has no obvious effect on reducing the power gap. The hydropower maintenance is concentrated in the dry season, and the original maintenance plan involves few periods with power gap, so the window shifting space of the hydropower maintenance plan is narrow.

[0488] In summary, the original maintenance plan of the power grid in a certain region in 2021 is reasonable, so the revision of the maintenance plan has no obvious improvement effect on reducing the power gap.

Claims

1. A power system power flow balancing analysis method based on multi-level multi-process panoramic time sequence operation simulation, characterized in that, The method comprises the following steps: 1) collecting and analyzing power system data information and initial operation mode, defining power and energy balance analysis index system and iteration termination criterion; 2) generating wind, light, water and load scene set, combining power and energy balance analysis index system to calculate power and energy supply and demand mismatch risk under multiple scenes, and obtaining power and energy balance analysis result; determining whether the power and energy balance analysis result meets the power and energy balance criterion, if yes, outputting the power and energy balance analysis result, otherwise, entering step 3); 3) constructing a flexible regulation model considering equipment response characteristics, and determining annual daily start mode by using unit aggregation technology; 4) determining time sequence decoupling criterion, and establishing time sequence decoupling multi-scene time sequence production simulation model; obtaining the power and energy balance analysis result by using the flexible regulation model considering equipment response characteristics, and correcting annual hourly power profit and loss curve in parallel; determining whether the power and energy balance analysis index meets the power and energy balance criterion, if yes, outputting the power and energy balance analysis result, otherwise, entering step 5); 5) establishing a multi-objective multi-scene optimization model for boundary revision with the highest clean energy utilization rate and the lowest load shedding and operation cost as the target; obtaining the power and energy balance analysis result by using the multi-objective multi-scene optimization model, and correcting the annual hourly power profit and loss curve; determining whether the power and energy balance analysis index meets the power and energy balance criterion, if yes, outputting the power and energy balance analysis result, otherwise, entering step 6); 6) determining whether the iteration convergence criterion is met, if yes, outputting the power and energy balance analysis result, if not, returning to step 2); The power system data information includes new energy power, load power, unit parameters and energy storage / pumping storage parameters; The step of determining the time sequence decoupling criterion comprises: 4.1) Time series wherein time satisfies the following conditions: where k = 1, 2,... N z ; N z + 1 is the number of optimization periods; is the time time power gap; 4.2) dividing the year into N z +1 optimization periods with the elements in time series T z +1 as a break point 4.3) establishing a time sequence decoupling criterion, if the time sequence decoupling criterion is not met, the optimization period is combined with the previous optimization period; if it is met, it is not combined; The time sequence decoupling criterion is as follows: where t0 is an optimized time period duration criterion; S z is an optimized time period imbalance time criterion; Δt is a time granularity; P t (τ) is a power gap; T z,k-1 , T z,k is an optimized time period start and end time; If then the optimization period is taken as: In the formula, is the energy stored by the electrical energy storage; is the energy stored by the nth pumped hydro storage;E B,max , W n,max is the maximum electrical energy of the nth energy storage device, the maximum energy capacity of the nth reservoir. If then the optimization period is taken as: In the formula, E B,n,min is the minimum energy of the nth energy storage device; is the minimum energy capacity of the nth reservoir.

2. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 1, characterized in that, The step of generating wind, light, water and load scene set comprises: 1) adopting K-means clustering technology based on RV coefficient to divide wind, light, water and nuclear load historical data set X into K classes; Let the set X = {X1, X2, …, X N} represent the daily output sample set of N p wind, light, water, and fire nuclear power stations and loads, where X i is the output matrix of the i-th day, and N is the sample capacity. The wind, light, water and nuclear load historical data set is divided into wind, light, water and nuclear load daily output matrix by taking matrix as the clustering object; The feature difference of different wind, light, water and nuclear load daily output matrix is quantified by improving RV coefficient, that is: where RV(X i ,X j ) denotes the RV coefficient of sample X i and sample X j ; tr(·) is the trace of a matrix; D diag (·) is a diagonal matrix; i,j1,2,…,N; is an intermediate variable; N is the number of samples; X i ' is the transpose of X i ; 2) according to the typical daily power state obtained by clustering, calculating Markov state transition probability matrix Pr and Markov cumulative state transition probability matrix Pcum; Then, in time sequence, based on Markov chain Monte Carlo algorithm, 365 days of power generation state are randomly extracted to form a annual wind-solar-hydro-nuclear power generation state transition process, and a set T containing Ns annual wind-solar-hydro-nuclear power generation state transition processes is obtained intra ; Markov state transition probability matrix P r As shown below: where p kl represents the probability of the wind, light, water, and fire nuclear power and day light output state transferring from state k to state l; k, l = 1, 2, …, K; K is the total number of states; probability p kl The maximum likelihood estimate is shown below: wherein n kl is the number of days in the historical data that transitioned from state k to state l; The Markov cumulative state transition probability matrix Pcum is as follows: In the formulae m represents a state; 3) Set T intra The daily state during the power generation state transition process of each year is used as a label. Driven by Gaussian white noise, the daily power generation curves of wind, solar, hydro, thermal and nuclear power plants in the corresponding state are generated based on the SGAN network. 4) connecting the daily power curve according to the order determined by the wind, light, water and nuclear annual power state transition process to obtain the annual wind, light, water and nuclear power generation scene set S.

3. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 2, characterized in that, The scaling dot attention mechanism is introduced into the generator and discriminator of the SGAN network; The scaled dot product attention mechanism refers to multiplying the matching weight α with the input matrix to obtain the output matrix A(x) of the attention mechanism. a ), Distinguishing matrix x a The power generation information of different energy plants in A(x) a The degree of contribution of data differentiation is used to extract the spatial correlation of power generation from multiple power plants; Output matrix A(x a ) is as follows: A(x a ) = ax a (6) where x a input matrix representing the attention mechanism, W is a learnable projection matrix, d w dimension of the matrix W; softmax(·) function is used to normalize the weights; The time sequence feature extraction unit of the SGAN network comprises a stacked residual module, and the output of the residual module is fused with historical power and convolution operation information; In the SGAN network, the scene power value y at time t t As follows: y t = g causal (x0, x1,..., x t ), t = 0, 1,..., T (8) where: x t is the input power at time t; g causal (·) is a causal convolution operation; T is the total number of time sections; The dilated convolution operation of the SGAN network is as follows: where DC(x) is the result of the dilated convolution operation of the filter on the elements c in the history power vector x, δ is the dilated convolution operator, f(i f ) represents the i f th filter, δ is the dilation rate, k is the filter size; f is the filter; is the dilated value of the i f th filter on the cth sequence point of the history power vector x; In the SGAN network, the loss functions of the generator and the discriminator are as follows: L G = -E S [D(S|C)] (10) where D(·) is the discriminator function; L G is the generator loss function; L D is the discriminator loss function; C represents the daily generation state label; S represents the generation power scenario; P obs is the historical observed power fused with the state label; E S [·] represents the expected value corresponding to the historical observed power P obs and the generation power scenario S. Training objective of the SGAN network As shown below: In the formula, C represents the daily generation state label; S represents the generation power scenario.

4. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 1, characterized in that, The power and energy balance analysis index system includes maximum power gap expectation, gap period percentage expectation, and cumulative energy gap percentage expectation. The steps of constructing the power and energy balance analysis index system include: 1) performing supply and demand balance analysis on the power system to obtain: N the,t is the number of thermal power units at time t; N wat,t is the number of hydroelectric power units at time t; N nuc is the number of nuclear power units; N buy , N sale is the number of sales and receiving areas; N wind,t is the number of wind power units at time t; N pv,t is the number of photovoltaic units at time t; N dwat,t is the number of small hydropower units in the local area at time t; P buy,n,t is the power purchased from the nth area outside at time t; P sale,n,t is the power sent to the nth area outside at time t; P the,n,t is the output power of the nth thermal power unit at time t; P wat,n,t is the output power of the nth hydroelectric power unit at time t; P dwat,n,t is the output power of the nth small hydropower unit in the local area at time t; P wind,n,t is the output power of the wind power unit at time t; P pv,n,t is the output power of the nth photovoltaic unit at time t; P nuc,n,t is the output power of the nth nuclear power unit at time t; P x,t is the output power of biomass power generation at time t; P gas.t represents the output power of the natural gas unit at time t; P t is the power gap at time t; P res,t represents the reserve capacity at time t; P load,t is the load at time t; 2) establishing power and energy balance evaluation indexes; where the maximum power gap expectation P max As follows: wherein π s denotes the probability of the scenario s; P t s is the power gap at time t for the scenario s. Notch period percentage expected P N As follows: In the formula, denotes the time in which a power gap exists under the scenario s, T is the total time; Cumulative energy gap percentage expectation P Q As follows: In the formula, represents the load at time t under scenario s.

5. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 1, characterized in that, The power and energy balance criterion is as follows: wherein, represents a maximum power gap threshold; represents a gap period percentage threshold; represents a cumulative power gap percentage threshold; P Q represents a cumulative power gap percentage expectation; P N represents a gap period percentage expectation; P max represents a maximum power gap expectation; The iteration convergence criterion is as follows: where i represents the number of iterations, represents an iteration termination condition; represents the maximum power gap expectation for the i-th, i-1-th iteration; represents the gap period percentage expectation for the i-th, i-1-th iteration; represents the cumulative power gap percentage expectation for the i-th, i-1-th iteration.

6. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 1, characterized in that, The objective function of the flexible regulation model minf considering the response characteristics of equipment is as follows: wherein is the curtailment penalty for wind for scenario s; is the curtailment penalty for solar for scenario s; is the load shedding penalty for scenario s; T is the total time length; is the predicted power of the nth wind turbine at time t for scenario s; c cur_wind,n is the curtailment penalty for the nth wind turbine; is the predicted power of the nth solar PV unit at time t for scenario s; c cur_pv,n is the curtailment penalty for the nth solar PV unit; c cur_load is the load shedding penalty; P t s is the power gap at time t for scenario s; is the power output of the wind turbine at time t for scenario s; is the power output of the nth solar PV unit at time t for scenario s; The constraint conditions of the flexible regulation model considering the response characteristics of equipment include power balance constraint, aggregated thermal power unit output constraint, hydropower unit capacity constraint, ground regulation small hydropower unit capacity constraint, wind power capacity constraint, photovoltaic capacity constraint, and reserve constraint. The power balance constraint is as follows: N the_j,t represents the total number of thermal power groups at time t; P the,j,t represents the output of the jth type of thermal power group at time t; N wat,t represents the number of hydropower units at time t; N nuc represents the number of nuclear power units; N buy , N sale represents the number of sales and receiving areas; N wind,t represents the number of wind power units at time t; N pv,t represents the number of photovoltaic units at time t; N dwat,t represents the number of small hydropower units in the local area at time t; N x represents the number of biomass power units; P x,n,t represents the power of the nth biomass power unit at time t; The aggregated thermal power unit output constraint is as follows: where s the,j,t is the number of j group of thermal power units at time t; is the minimum output of a single unit in the jth group of thermal power units at time t; P the_G,j,t is the capacity of a single unit in the jth group of thermal power units; P the_lim,j,t is the limited output of a single unit in the jth group of thermal power units at time t; n j represents the number of units in the jth group of thermal power units; P the_m,j,t is the maintenance capacity of the jth group of thermal power units at time t; respectively represent the minimum and maximum values of the number of j group of thermal power units; P the,j,t is the output power of the j group of thermal power units at time t; The hydropower unit capacity constraint is as follows: In the formula is the maximum output of the nth water turbine unit at time t; is the minimum output of the nth water turbine unit at time t; P wat_G,n,t is the capacity of the nth water turbine unit; P wat_lim,n,t is the limited output of the nth water turbine unit at time t; P wat_m,n,t is the maintenance capacity of the nth water turbine unit at time t; The ground regulation small hydropower unit capacity constraint is as follows: 0 < P dwat,n,t ≤ P pre_dwat,n,t (24) The wind power capacity constraint is as follows: 0 < P wind,n,t ≤ P pre_wind,n,t (25) The photovoltaic capacity constraint is as follows: 0 < P pv,n,t ≤ P pre_pv,n,t (26) The reserve constraint is as follows: In the formula, s g,t is the start-stop state of the nth unit t, is the maximum and minimum power generation of the unit t, P res,u , P res,d is the upper and lower standby capacity; s the,t is the number of units started at time t; P g,t is the power generation of the unit.

7. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 1, characterized in that, The objective function minf of the multi-scenario time sequence production simulation model is as follows: minf = C cur_wind +C cur_pv +C cur_load (32) In the formula, C cur_wind , C cur_pv , C cur_load is a wind curtailment penalty, a light curtailment penalty, a cut load penalty; The constraint conditions of the multi-scenario time sequence production simulation model include power balance constraint, energy storage capacity constraint, energy storage charging and discharging power balance constraint, pumped storage capacity constraint, pumped storage unit constraint, pumped storage unit ramping constraint, thermal power unit capacity constraint, thermal power unit ramping constraint, hydropower unit capacity constraint, hydropower unit ramping constraint, ground regulation small hydropower unit capacity constraint, wind power capacity constraint, photovoltaic capacity constraint, and reserve capacity constraint. The power balance constraint is as follows: In the formula, P sto_cha,n,t is the charging power of the nth energy storage device at time t; P sto_dis,n,t is the discharging power of the nth energy storage device at time t; P pump_dis,n,t is the pumping and storage power of the nth reservoir at time t; P pump_cha,n,t is the power generation of the nth reservoir at time t; The energy storage capacity constraint is as follows: wherein γ B is the self-loss coefficient of the nth electric energy storage; η B,cha,n is the charging efficiency of the nth electric energy storage; η B,dis,n is the discharging efficiency of the nth electric energy storage; P cha,n.t is the charging power of the nth electric energy storage at time t; and P dis,n,t is the discharging power of the nth electric energy storage at time t. The energy storage charging and discharging power balance constraint is as follows: In the formula, P cha,n,max is the maximum charging power of the nth energy storage device; P dis,n,min is the maximum discharging power of the nth energy storage device; s st_cha,t , s st_dis,t are the charging and discharging states of the energy storage at time t, and are 0, 1 variables, 1 in operation, and 0 otherwise; N sto is the charging and discharging state threshold of all energy storage devices; The pumped storage capacity constraint is as follows: wherein η pump_cha,n , η pump_dis,n are the pumping and releasing efficiencies of the nth pumped storage respectively; P pump_cha,n,t , P pump_dis,n,t are the pumping and releasing powers of the nth pumped storage at time t respectively; W n,t is the reservoir storage capacity; is the upper and lower limits of the reservoir storage capacity; The pumped storage unit constraint is as follows: where P pump_d,n,min and P pump_d,max represent the minimum and maximum generation output limits of the nth pumped storage unit, respectively; P pump_u,n,min and P pump_u,n,max represent the minimum and maximum pumping output limits of the nth pumped storage unit, respectively; s pump_d,n,t and s pump_u,n,t represent the state of generation and pumping of the nth pumped storage unit, respectively. The pumped storage unit ramping constraint is as follows: where ur pump_d,n / dr pump_d,n denotes the up / down ramping constraint for the nth pumped storage unit when generating electricity; ur pump_u,n / dr pump_u,n denotes the up / down ramping constraint for the nth pumped storage unit when pumping water; The thermal power unit capacity constraint is as follows: The thermal power unit ramping constraint is as follows: where ur the,j is the up ramp constraint for the jth group of thermal units; dr the,j is the down ramp constraint for the jth group of thermal units; The hydropower unit capacity constraint is as follows: The hydropower unit ramping constraint is as follows: where ur wat,n is the upper ramping constraint of the nth hydroelectric generating unit; dr wat,n is the lower ramping constraint of the nth hydroelectric generating unit; The ground regulation small hydropower unit capacity constraint is as follows: 0 < P dwat,n,t ≤ P pre_dwat,n,t (42) The wind power capacity constraint is as follows: 0 < P wind,n,t ≤ P pre_wind,n,t (43) The photovoltaic capacity constraint is as follows: 0 < P pv,n,t ≤ P pre_pv,n,t (44) The reserve capacity constraint is as follows:

8. The power flow and energy balance analysis method of power system based on multi-level cascaded multi-process panoramic time sequence operation simulation according to claim 1, characterized in that, The objective function of the multi-objective multi-scenario optimization model is as follows: minf = C cur_wind +C cur_pv +C cur_load (46) The constraint conditions of the multi-objective multi-scenario optimization model include power balance constraint, thermal power unit capacity constraint, hydropower maintenance duration and continuity constraint, thermal power maintenance time window constraint, thermal power unit capacity constraint, hydropower maintenance duration and continuity constraint, hydropower maintenance time window constraint, reservoir scheduling constraint, ground regulation small hydropower unit capacity constraint, wind power capacity constraint, photovoltaic capacity constraint, and reserve capacity constraint. The power balance constraint is as follows: The capacity constraints of the thermal power units are as follows: In the formula, u the,n,t is the repair state of the jth thermal power unit group at time t, 0 indicates that it is in repair state, otherwise it indicates that it is not in repair state; The hydropower maintenance duration and continuity constraint is as follows: In the formula, D the,j is the jth thermal power unit group maintenance duration; The thermal power maintenance time window constraint is as follows: In the formula, is the earliest maintenance time of the jth thermal power unit group; is the slowest maintenance time of the jth thermal power unit group; The thermal power unit capacity constraint is as follows: wherein u wat,n,t is the repair state of the nth water turbine at time t, 0 indicating being in repair state and 1 indicating not being in repair state; λ wat,n is the power generation efficiency of the nth water turbine; Q wat,n,t is the water quantity generated by the nth reservoir at time t; h wat,n,t is the water head of the nth reservoir at time t; s wat,n,t is the start state of the water turbine; The hydropower maintenance duration and continuity constraint is as follows: In the formula, D wat,n is the nth water turbine unit maintenance duration; The hydropower maintenance time window constraint is as follows: In the formula is the earliest maintenance time of the nth water turbine unit; is the slowest maintenance time of the nth water turbine unit; The reservoir scheduling constraint is as follows: wherein Q cut,n,t is the abandoned water volume of the nth reservoir at time t; f n,t is the natural inflow of the nth reservoir at time t; V wat,n,t is the storage capacity of the nth reservoir at time t; is the maximum power generation water flow of the nth reservoir; is the maximum abandoned water volume of the nth reservoir; Q cut,n-1,t is the abandoned water volume of the upstream reservoir of the nth reservoir at time t; Q wat,n-1,t is the power generation flow of the upstream reservoir of the nth reservoir at time t; The ground regulation small hydropower unit capacity constraint is as follows: 0 < P dwat,n,t ≤P pre_dwat,n,t (54) The wind power capacity constraint is as follows: 0 < P wind,n,t ≤ P pre_wind,n,t (55) The photovoltaic capacity constraint is as follows: 0 < P pv,n,t ≤ P pre_pv,n,t (56) The reserve capacity constraint is shown below:

Images