System security of hybrid AC-HVDC-systems challenges and new approaches for combined security assessment,preventive optimization and curative actions

1 Introduction

The increase of renewable energy sources causes future challenges for the power system.Regarding the transmission level,the significant distances between areas of high renewable generation,storage capacities and centers of load require bulk power transmission.The existing AC grid can be supported by HVDC links based on the Voltage Source Converter (VSC)technology,realized as modular multilevel converters (MMC).Beginning with single point-to-point interconnections,these links can be merged into a meshed HVDC overlay grid within Europe [1,2] or China [3–5] in a next step.In combination with the AC transmission grid,they form a hybrid AC-HVDC-system.

In general,the operation of a power system is defined by strict rules for system security,including a compliance with N-1 security:In case of any contingency event,static aspects (operational security limits)and system stability must be considered.The operation of meshed HVDC-systems requires a decentralized DC voltage control (see subsection 2.1)Consequently,HVDC-related contingencies affect both,the HVDC- and the AC-system.Section 2.2 describes a combined security assessment for hybrid ACHVDC-systems to be implemented in the operational planning process.

The combined approach can also be applied to the process of operational planning.The active and reactive converter set-points are determined by an optimal power flow(OPF)calculation.Existing OPF concepts do not assume a combined N-1 security for the hybrid AC-HVDC-system[6–8].Therefore,a combined approach is presented in section 3.The high impact of the DC voltage control on the system security in case of HVDC-related contingencies is considered by expanding operational planning by an intelligent parameterization approach presented in section 4.

Modern MMC-based VSCs enable a fast control of active and reactive power [9,10].By general design,the HVDC-system will not adapt its set-points in case of contingency events automatically.Additional control instances are required instead.Several state-of-research approaches are described in subsection 2.3,with their primary focus on AC stability.However,the dynamic and flexible behaviour of future HVDC-systems can be also applied to address operational security limits during online operation.Such curative concepts became more interesting ithin the last years [11].Section 5 describes the automated activation of curative set-point adaptations in order to maintain a feasible system state after AC-related contingencies.

The individual concepts will improve the combined operation of hybrid AC-HVDC-systems,which is considered a promising solution to increase the use of thermal transfer capacity of the existing equipment.The concepts are evaluated by means of numerical case studies in section 6.The work is concluded in section 7.

2 System security of hybrid AC-HVDC-systems

In general,system security limits classes are operational security limits (static)and stability limits (dynamic).Operational security limits include the thermal rating of equipment and defined voltage levels.AC-system stability is divided into rotor angle,voltage and frequency stability[12],whereas the HVDC-system is characterized by DC(voltage or energy)stability as described in subsection 2.1.This is respected in a combined security assessment of hybrid AC-HVDC-systems (subsection 2.2).

2.1 DC voltage control

As described by equation (1),an imbalance of energy results in a derivation of the DC voltage.Therefore,DC stability refers to the ability of a DC-system to maintain steady voltages at all buses in the system after being subjected to a disturbance from a given initial operating condition by the ability to maintain/restore an energy equilibrium.DC stability includes small-signal (e.g.[13])and transient aspects (e.g.[14]).

A local and decentralized DC voltage control is considered state-of-the-art.The literature suggests the implementation of a control concept based on p-vcharacteristics,such as a linear droop (kDC)as applied in equation (2).[15,16] Other approaches use tripartite droops[17] or continuous functions [18].

The converters control the active power exchange between the AC- and HVDC-system.An adaptation of active power in-feed after a DC related disturbance results in a new (post-contingency)AC- and HVDC-power flow condition (see also [19]).Consequently,these effects need to be considered in system security assessment.

2.2 Security assessment of hybrid AC-HVDC-systems

A common definition for HVDC-related contingencies does not exist yet.In some cases,each HVDC-system is defined as one single element during security assessment,leading to large security margins or unexploited thermal transport capacity.A combined security assessment is supposed instead (see Table 1)which is capable to represent the additional level introduced by hybrid AC-HVDC-systems.Thereby,the assessment of stability aspects is considered obligatory (see [20,21]).

Table 1 Impact of individual contingency types

AC system(static)X X X X AC system(dyn.)X X X X DC system(stat)X X DC system(dyn.)X X impact AC branch Generator VSC(-poles)DC branch AC system (static)X X X X AC system (dyn.)X X X X DC system (stat)X X DC system (dyn.)X X

The effect of AC-related contingencies (AC branches,generation units)is limited to the security of the AC system.In contrast,HVDC-related contingencies affect both the HVDC- and the AC-system by changing the power flow condition and losing reactive power provision (VSC outages).This comprises the faulted converter and the reaction of the DC voltage control (see subsection 2.1)whose impact needs to be respected in any case.Fig.1 shows the resulting system states by means of system security.

Fig.1 Combined security assessment for hybrid AC-HVDC-systems

The advanced OPF algorithm presented in section 3 complies with the described combination of AC- and HVDC-system security.The approach respects the impact of a tuned DC voltage control (section 4)as well as the capability and potential of curative actions (section 5).

2.3 Contribution of HVDC-systems to AC-stability

In future,stability aspects are considered a limiting factor for power transfer capacity which has to be considered during security assessment [22].The fast control of active and reactive currents respectively power enables the MMCVSCs to address stability phenomena.An advanced current control guarantees a fault ride through (FRT)capability [23].The modulation of active and reactive power set-points enables a support of transient stability [24] or oscillation damping [25].The individual aspects are not described in detail by this contribution.Shah et al.provide an detailed overview in [26].In a next step,the (dynamic)stability assessment has to take the additional control structures into account,if they are applied to the individual VSCs.

3 P-SCOPF for N-1 security

For assessing the system security within an OPF calculation,the problem formulation of the OPF has to be extended to include security-related constraints [27].The general problem formulation is as follows:

subject to:

Thereby,gk and hk are representing the equality and inequality constraints.The subscript k indicates the corresponding contingency scenario,which include all contingencies (NC)within the hybrid AC-HVDC-system.The special case k = 0 represents the base case in which the preventive measure will be activated.N-1-security is maintained if every contingency k complies with the constraint functions gk and hk.This consideration results in a preventive security constraint OPF (P-SCOPF)problem.Sennewald et al.presented the first AC-HVDC-application in [28].The major aspects are described in this section.The P-SCOPF for hybrid AC-HVDC-systems utilizes set-point of the VSCs as degrees of freedom.The resulting values can be activated by the operator during a schedule update.

In (3)multiple objectives are formulated:The first objective aims to establish N-1 security by minimizing the number of critical contingencies (SCC),which includes all individual contingency scenarios k violating gk or hk.The second objective is meant to minimize the set-point deviation compared to the initial VSC set-points.

In accordance with the methodology from [28],the individual contingency scenarios are only assessed if the setpoint is valid/feasible for the base case.A detailed list of the used constraints can be found in [29],in which an OPF formulation for AC-HVDC-systems has been proposed initially.They include general aspects (AC/HVDC power balance)and security aspects (equipment loading/voltage limits).The presented approach can be expanded in the future,e.g.to incorporate the DC voltage control as outlined in section 4,an additional consideration of generators for redispatch or the availability of curative actions (section 5)for those contingencies that could not be removed by the preventive approach.These thoughts are already depicted in Fig.1.

4 Optimization of DC voltage control

Existing publications for a parameterization of DC voltage control focus on stability aspects [30]:determines droop gains in order to damp DC oscillations and maintaining AC small signal stability,which leads to an optimization of AC small signal stability [31].Eriksson et al.adapt the parameter settings to minimize the impact on transient AC stability in case of DC-disturbances.[32]However,the parameterization (droop gains)of DC voltage control determines the power flow condition after a DC-related contingency,as described in section 2.1.In order to prevent violations of operational security limits in case of a VSC outage,a preventive parameterization approach is presented in this section to be applied during the phase of operational planning.

Fig.2 depicts the optimization process initially presented in [33].The optimization of the DC voltage control is based on its parameterization,e.g.the droop gain kDC (see (8)).

The optimization objective is to find a set of DC voltage control parameters for all VSCs to prevent violations of system security limits after any VSC outage.N-1 security is established by minimizing the number of constraint violations (critical contingencies)in the individual scenario k.However,the objective function (9)is only applied to VSC-related contingencies (VSC).

Fig.2 DC voltage parameterization approach

The applied parameterization approach is feasible for all DC voltage control functions,due to the incorporated combined AC-DC power flow algorithm.In a first step,the optimization is limited to a consideration of static operational security limits,comprising the thermal limits of DC and AC lines,the power limits of the converters as well as the maximum voltage of DC nodes.They are represented as constraint functions (hk(xk,uk)).If the defined DC voltage control characteristic prevents any assessed constraint violation,the hybrid system is N-1 secure regarding VSC outages.The approach is realized by a Particle Swarm Optimization (PSO).An optimization of sigmoidal functions and comparison with linear droops is presented in [33].

5 Curative actions by HVDC-system

In a hybrid AC-HVDC-system,VSC set-points enable a direct impact on the AC power flow condition regarding equipment loading and reactive power provision.In case of contingency related congestions and violations of operational security limits,curative actions of the HVDC-system can retransfer the system to a feasible state of operation within a feasible time due to their dynamic behavior and independent control of active and reactive power.However,the concept requires a central coordination of curative set-points for the individual contingencies(subsection 5.1)and an activation strategy (subsection 5.2)in order to enhance the existing concept of preventive system security.The idea was presented first by Marten et al.in [34].

5.1 Calculation of curative actions

The calculation of curative actions for the HVDC-system becomes part of the operational planning process,subordinated to the existing static security assessment.In contrast to the P-SCOPF approach (section 3),the corrective security constrained OPF (C-SCOPF)which was first proposed by [35] focusses on a feasible solution for the kth post-contingency configuration only.

A C-SCOPF application was first described by [36]and extended for multi-terminal HVDC-systems by [6].In [37],a combined calculation for fast (VSC based)and slow (generation dispatch)corrective actions is presented,respecting the active power and DC voltage,which has been neglected before.An additional consideration of reactive power was not published yet for C-SCOPF applications but is included in the approach presented in this work,as described by (10).In order to determine a feasible curative action for the contingency scenario k,all VSC set-points are included in the C-SCOPF calculation.The fitness function is described by (12).It determines the minimal required setpoint adaptationto relieve the congestions by complying with all defined constraints,which coincide with the constraints of section 3.

5.2 Activation of curative actions

Once pre-calculated during operational planning,the contingency-related curative actions are ready to be activated in online operation.The activation process can either be based on an operator (manual)or system decision(automated).Both strategies require an identification of the referring contingency event by means of surveillance and situation awareness tools [38].The identification of events is not discussed in this work in detail but is most likely based on PMU measurements that are compared to a database of historic measurements or simulation data.The matching process of this data-driven approach is based on feature extraction methods (e.g.[39,40])and classifiers (e.g.[41]).

On one hand,the violation of operational security limits leaves enough time for a manual reaction in the range of one minute.Sass et al.describe such a manual approach in [42]:The identification of events is based on a central analysis and classification of PMU measurements,resulting in a decision support of the operator by proposing the referring curative action.The set-point changes are transmitted to the individual converter substations via the existing communication infrastructure.This enables a coordinated system wide activation.

On the other hand,an automated activation enables a fast and reliable activation even in case of great stress of the operator.In contrast to the central and manual approach,a strictly local approach is presented in [43]:Both,the contingency-related set-points and the identification database are pre-calculated (Fig.3 b)and c))and transferred to the individual converter substations subsequently.During online operation,local measurements are surveyed and analyzed(Fig.3 d)).In case of a successful identification,the curative set-points are activated.A local secondary control (see [44])enables an active power adaptation without interfering with the DC voltage control instances at each converter station.

Fig.3 General structure of method for local activation of curative actions

5.3 Enhancement of system operation by curative actions

A future application of fast curative actions as a part of system operation can complement the existing preventive security for individual contingency types.However,the activation and the impact on system security have to be guaranteed during operational planning already,regardless of the activation approach,in order to maintain system stability and prevent cascaded outages.Automated curative actions comply with the NERC definition of Remedial Action Schemes (RAS,[45])or ENTSO-E definition of Special Protection Schemes (SPS,[46]).

6 Numerical case studies

6.1 Test system

A common test system for hybrid AC-HVDC-systems did not exist initially.An application of the CIGRE HVDC test system [47] is not considered reasonable,since it focusses on an offshore HVDC grid instead of an hybrid AC-HVDC-system.Therefore,the authors proposed an open source benchmark system in [48],which is presented in Fig.4.It consists of a meshed AC system with 67 nodes and 102 lines and a meshed HVDC grid with 9 nodes and 11 lines.VSC 9 represents an offshore wind farm connected via a tap line.The system is characterized by a significant power transit between the control areas 1 and 2.The provided static model is expanded by standard type 5th order synchronous machines and VSC models complying with [49] (type 6).

In its initial state,the AC-HVDC-system is not N-1 secure by design.Considering single outages of AC or HVDC branches,AC generating units and VSCs,it comprises 17 critical contingencies,i.e.contingencies that result in impermissible post-contingency system states.These critical contingencies are marked red in Fig.4.

Fig.4 AC-HVDC-benchmark system:topology of a)AC-system,b)HVDC-system

6.2 Preventive SCOPF

The P-SCOPF approach described in section 3 is applied to the test system in order to establish N-1 security for the hybrid AC-HVDC-system.Table 2 enlists the initial ( pVSC,vDC)and the optimized preventive VSC set-pointsPlease note that the only utilized degree of freedom of VSC 9 is its DC nodal voltage as it is operated in a constant power mode.By deploying the preventive set-point,the number of critical contingencies can be reduced from 17 to 10 contingencies,comprising 4 AC line outages,3 DC line outages and 3 VSC outages.The P-SCOPF algorithm was able to reduce the number of critical contingencies,even though it was not able to establish N-1 security by only adapting the VSC set-points.A combination with conventional redispatch can be applied in the future or HVDC-related curative measures can be pre-calculated to address the AC contingencies (see section 5).

Table 2 Initial and OPF-based preventive set-points

6.3 Preventive parameterization of DC voltage control

The preventive optimization approach is tested on the AC-HVDC-benchmark system described in section 6.1.Linear droop functions are used for the DC voltage control.In their non-optimized condition the linear droop factors are equal for all VSC1-VSC8 (kDC = 140 MW/kV)under consideration of the full use of the rated voltage range (475-525 kV).The initial condition results into three critical contingencies due to VSC outages.Table 3 enlists the referring scenarios and the resulting violations of line loading (see also Fig.3).

The optimization of kDC is constrained by a lower bound (kDC,min = 20 MW/kV)to ensure the full provision of balancing power and an upper bound (kDC,max = ∞),which complies with the constant voltage control mode.The preventive optimization process minimizes the number of VSC-related contingencies to one (see Table 3).For outages of VSC 2 and VSC 5,the optimized DC voltage control results in an AC power flow condition without limit violations.Whereas the initial parameterization leads to an overloading of AC branch #43,resp.AC branch #83.

Table 3 Post-contingency line loadings

VSC2#43108.2%99.2%VSC5#83109.1%97.8%VSC8#43102.6%93.9%#47120.1%117.3%

In contrast,the critical impact of the outage of VSC 8 on system security cannot be prevented completely by the preventive approach:The overloading of AC line #47 is a result of the loss of the reactive power supply by VSC 8,which is the only source of reactive power in an area with a high demand of load.Still,the optimized control function affects a minor AC line loading in comparison to the case with a non-optimized kDC.

Fig.5 Control Characteristics of the VSCs - post contingency operating point for outage of VSC 5

In addition,Fig.5 shows the post-contingency operating points for the outage of VSC 5.The plots include the individual p-v-characteristics of each VSC.Without optimization,kDC is set equal for all VSCs,which results in an almost even distribution of balancing power for the remaining converters.The optimization leads to high droop constants for VSC 1,5 and 8.This enables an increase of their balancing power provision in comparison to the other ones while preventing constraint violations.

6.4 Curative actions

The application of the C-SCOPF algorithm presented in subsection 5.1 enables the determination of feasible curative actions for 15 of the 17 critical contingencies.All curative actions comprise both an adaptation of active and reactive power.The two remaining contingencies relate to VSC outages.The numerical case study presented in this subsection intends to describe the strict local approach as described in subsection 5.2,which is considered the maximum requirement.This means,a successful application of the local approach also proofs the feasibility of central activation of curative actions.

The presented results focus on the activation process of the pre-calculated curative actions and their impact on AC- and DC-security and stability.The details on the local identification algorithm and its restriction are described in[43].An outage of AC line 43 (3-phase short circuit and subsequent breaker tripping),which is connecting control area 1 and 2 at a load level of 80 %,causes a violation of operational security limits (AC line 42 and 83)if no curative action is activated (Fig.6 a).

Fig.6 Activation of curative action by HVDC-system:a)impact on AC line loading,b)active power provision,c)DC-voltage

The proposed approach enables a local identification of the contingency scenario at several converter stations(VSC 1-VSC 3,VSC 7)within 400 ms.This enables the activation of the pre-calculated active and reactive power set-point adaptations without any communication with the control room.Respecting a defined slew rate of 2 GW/s,the individual converters reach their new set-points within 350 ms (see Fig.6 b),thereby removing the congestion 600 ms after the line outage (see Fig.6 a).Thereby,the overall system is retransferred into a normal (feasible)state.The DC voltage profile is affected by the described process (Fig.6 c),but the converters without local identification remain at their reference active power set-points due to the local secondary control instance.The presented contingency scenario does not result in a violation of reactive power or AC voltage limits.Therefore,the curative set-point adaptation does not comprise an adaptation of qVSC.

7 Conclusions

The ongoing implementation of VSC-based HVDC-systems into the existing AC transmission system will result in a hybrid AC-HVDC-system.Such a system requires a combined approach for operational planning and system operation in order to guarantee an efficient use of the total(AC- and HVDC-)thermal transfer capacity.Consequently,this paper describes the update of existing security considerations.

This paper proposes three individual and independent concepts to address the described challenges and to enable a secure and efficient operation of hybrid AC-HVDC-systems:An improved P-SCOPF (section 3)enables an optimization of converter set-points to establish N-1 security of the hybrid AC-HVDC-system.Most likely,not all critical contingencies can be addressed by preventive optimization(see section 6).Regarding VSC outages,a preventive parameter optimization (section 4)guarantees a predicted behaviour of DC voltage control and thereby prevents the violation of operational security limits,which increases hybrid N-1 security.

In addition to the preventive approaches,HVDC-based curative actions (section 5)enable a fast and reliant power adaptation in case of critical contingencies in order to maintain system security.The presented approach focusses on static aspects and slow stability phenomena introduced by AC-related contingencies.The described methods complement existing functions of HVDC-systems and thereby support system operation and operational planning(see Fig.7).A future combination promises a reduction of today’s required redispatch efforts and an increase of the used grid capacity.

Fig.7 Mutual impact of HVDC-systems on system security of hybrid AC-HVDC-systems